1
0
Fork 0
mirror of https://github.com/git/git.git synced 2024-11-05 08:47:56 +01:00
git/diff.c

1738 lines
42 KiB
C
Raw Normal View History

/*
* Copyright (C) 2005 Junio C Hamano
*/
#include <sys/types.h>
#include <sys/wait.h>
#include <signal.h>
#include "cache.h"
#include "quote.h"
#include "diff.h"
#include "diffcore.h"
#include "xdiff-interface.h"
static int use_size_cache;
int diff_rename_limit_default = -1;
int git_diff_config(const char *var, const char *value)
{
if (!strcmp(var, "diff.renamelimit")) {
diff_rename_limit_default = git_config_int(var, value);
return 0;
}
return git_default_config(var, value);
}
static char *quote_one(const char *str)
{
int needlen;
char *xp;
if (!str)
return NULL;
needlen = quote_c_style(str, NULL, NULL, 0);
if (!needlen)
return strdup(str);
xp = xmalloc(needlen + 1);
quote_c_style(str, xp, NULL, 0);
return xp;
}
static char *quote_two(const char *one, const char *two)
{
int need_one = quote_c_style(one, NULL, NULL, 1);
int need_two = quote_c_style(two, NULL, NULL, 1);
char *xp;
if (need_one + need_two) {
if (!need_one) need_one = strlen(one);
if (!need_two) need_one = strlen(two);
xp = xmalloc(need_one + need_two + 3);
xp[0] = '"';
quote_c_style(one, xp + 1, NULL, 1);
quote_c_style(two, xp + need_one + 1, NULL, 1);
strcpy(xp + need_one + need_two + 1, "\"");
return xp;
}
need_one = strlen(one);
need_two = strlen(two);
xp = xmalloc(need_one + need_two + 1);
strcpy(xp, one);
strcpy(xp + need_one, two);
return xp;
}
static const char *external_diff(void)
{
Rename environment variables. H. Peter Anvin mentioned that using SHA1_whatever as an environment variable name is not nice and we should instead use names starting with "GIT_" prefix to avoid conflicts. Here is what this patch does: * Renames the following environment variables: New name Old Name GIT_AUTHOR_DATE AUTHOR_DATE GIT_AUTHOR_EMAIL AUTHOR_EMAIL GIT_AUTHOR_NAME AUTHOR_NAME GIT_COMMITTER_EMAIL COMMIT_AUTHOR_EMAIL GIT_COMMITTER_NAME COMMIT_AUTHOR_NAME GIT_ALTERNATE_OBJECT_DIRECTORIES SHA1_FILE_DIRECTORIES GIT_OBJECT_DIRECTORY SHA1_FILE_DIRECTORY * Introduces a compatibility macro, gitenv(), which does an getenv() and if it fails calls gitenv_bc(), which in turn picks up the value from old name while giving a warning about using an old name. * Changes all users of the environment variable to fetch environment variable with the new name using gitenv(). * Updates the documentation and scripts shipped with Linus GIT distribution. The transition plan is as follows: * We will keep the backward compatibility list used by gitenv() for now, so the current scripts and user environments continue to work as before. The users will get warnings when they have old name but not new name in their environment to the stderr. * The Porcelain layers should start using new names. However, just in case it ends up calling old Plumbing layer implementation, they should also export old names, taking values from the corresponding new names, during the transition period. * After a transition period, we would drop the compatibility support and drop gitenv(). Revert the callers to directly call getenv() but keep using the new names. The last part is probably optional and the transition duration needs to be set to a reasonable value. Signed-off-by: Junio C Hamano <junkio@cox.net>
2005-05-10 02:57:56 +02:00
static const char *external_diff_cmd = NULL;
static int done_preparing = 0;
if (done_preparing)
return external_diff_cmd;
external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
done_preparing = 1;
return external_diff_cmd;
}
#define TEMPFILE_PATH_LEN 50
static struct diff_tempfile {
const char *name; /* filename external diff should read from */
char hex[41];
char mode[10];
char tmp_path[TEMPFILE_PATH_LEN];
} diff_temp[2];
static int count_lines(const char *data, int size)
{
int count, ch, completely_empty = 1, nl_just_seen = 0;
count = 0;
while (0 < size--) {
ch = *data++;
if (ch == '\n') {
count++;
nl_just_seen = 1;
completely_empty = 0;
}
else {
nl_just_seen = 0;
completely_empty = 0;
}
}
if (completely_empty)
return 0;
if (!nl_just_seen)
count++; /* no trailing newline */
return count;
}
static void print_line_count(int count)
{
switch (count) {
case 0:
printf("0,0");
break;
case 1:
printf("1");
break;
default:
printf("1,%d", count);
break;
}
}
static void copy_file(int prefix, const char *data, int size)
{
int ch, nl_just_seen = 1;
while (0 < size--) {
ch = *data++;
if (nl_just_seen)
putchar(prefix);
putchar(ch);
if (ch == '\n')
nl_just_seen = 1;
else
nl_just_seen = 0;
}
if (!nl_just_seen)
printf("\n\\ No newline at end of file\n");
}
static void emit_rewrite_diff(const char *name_a,
const char *name_b,
struct diff_filespec *one,
struct diff_filespec *two)
{
int lc_a, lc_b;
diff_populate_filespec(one, 0);
diff_populate_filespec(two, 0);
lc_a = count_lines(one->data, one->size);
lc_b = count_lines(two->data, two->size);
printf("--- %s\n+++ %s\n@@ -", name_a, name_b);
print_line_count(lc_a);
printf(" +");
print_line_count(lc_b);
printf(" @@\n");
if (lc_a)
copy_file('-', one->data, one->size);
if (lc_b)
copy_file('+', two->data, two->size);
}
static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
{
if (!DIFF_FILE_VALID(one)) {
mf->ptr = ""; /* does not matter */
mf->size = 0;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return 0;
}
else if (diff_populate_filespec(one, 0))
return -1;
mf->ptr = one->data;
mf->size = one->size;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
return 0;
}
struct emit_callback {
const char **label_path;
};
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
static int fn_out(void *priv, mmbuffer_t *mb, int nbuf)
{
int i;
struct emit_callback *ecbdata = priv;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
if (ecbdata->label_path[0]) {
printf("--- %s\n", ecbdata->label_path[0]);
printf("+++ %s\n", ecbdata->label_path[1]);
ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
}
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
for (i = 0; i < nbuf; i++)
if (!fwrite(mb[i].ptr, mb[i].size, 1, stdout))
return -1;
return 0;
}
struct diffstat_t {
struct xdiff_emit_state xm;
int nr;
int alloc;
struct diffstat_file {
char *name;
unsigned is_unmerged:1;
unsigned is_binary:1;
unsigned int added, deleted;
} **files;
};
static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
const char *name)
{
struct diffstat_file *x;
x = xcalloc(sizeof (*x), 1);
if (diffstat->nr == diffstat->alloc) {
diffstat->alloc = alloc_nr(diffstat->alloc);
diffstat->files = xrealloc(diffstat->files,
diffstat->alloc * sizeof(x));
}
diffstat->files[diffstat->nr++] = x;
x->name = strdup(name);
return x;
}
static void diffstat_consume(void *priv, char *line, unsigned long len)
{
struct diffstat_t *diffstat = priv;
struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
if (line[0] == '+')
x->added++;
else if (line[0] == '-')
x->deleted++;
}
static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
static const char minuses[]= "----------------------------------------------------------------------";
static void show_stats(struct diffstat_t* data)
{
char *prefix = "";
int i, len, add, del, total, adds = 0, dels = 0;
int max, max_change = 0, max_len = 0;
int total_files = data->nr;
if (data->nr == 0)
return;
for (i = 0; i < data->nr; i++) {
struct diffstat_file *file = data->files[i];
len = strlen(file->name);
if (max_len < len)
max_len = len;
if (file->is_binary || file->is_unmerged)
continue;
if (max_change < file->added + file->deleted)
max_change = file->added + file->deleted;
}
for (i = 0; i < data->nr; i++) {
char *name = data->files[i]->name;
int added = data->files[i]->added;
int deleted = data->files[i]->deleted;
if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
char *qname = xmalloc(len + 1);
quote_c_style(name, qname, NULL, 0);
free(name);
data->files[i]->name = name = qname;
}
/*
* "scale" the filename
*/
len = strlen(name);
max = max_len;
if (max > 50)
max = 50;
if (len > max) {
char *slash;
prefix = "...";
max -= 3;
name += len - max;
slash = strchr(name, '/');
if (slash)
name = slash;
}
len = max;
/*
* scale the add/delete
*/
max = max_change;
if (max + len > 70)
max = 70 - len;
if (data->files[i]->is_binary) {
printf(" %s%-*s | Bin\n", prefix, len, name);
goto free_diffstat_file;
}
else if (data->files[i]->is_unmerged) {
printf(" %s%-*s | Unmerged\n", prefix, len, name);
goto free_diffstat_file;
}
else if (added + deleted == 0) {
total_files--;
goto free_diffstat_file;
}
add = added;
del = deleted;
total = add + del;
adds += add;
dels += del;
if (max_change > 0) {
total = (total * max + max_change / 2) / max_change;
add = (add * max + max_change / 2) / max_change;
del = total - add;
}
printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
len, name, added + deleted,
add, pluses, del, minuses);
free_diffstat_file:
free(data->files[i]->name);
free(data->files[i]);
}
free(data->files);
printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
total_files, adds, dels);
}
#define FIRST_FEW_BYTES 8000
static int mmfile_is_binary(mmfile_t *mf)
{
long sz = mf->size;
if (FIRST_FEW_BYTES < sz)
sz = FIRST_FEW_BYTES;
if (memchr(mf->ptr, 0, sz))
return 1;
return 0;
}
static void builtin_diff(const char *name_a,
const char *name_b,
struct diff_filespec *one,
struct diff_filespec *two,
const char *xfrm_msg,
int complete_rewrite)
{
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
mmfile_t mf1, mf2;
const char *lbl[2];
char *a_one, *b_two;
a_one = quote_two("a/", name_a);
b_two = quote_two("b/", name_b);
lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
printf("diff --git %s %s\n", a_one, b_two);
if (lbl[0][0] == '/') {
/* /dev/null */
printf("new file mode %06o\n", two->mode);
if (xfrm_msg && xfrm_msg[0])
puts(xfrm_msg);
}
else if (lbl[1][0] == '/') {
printf("deleted file mode %06o\n", one->mode);
if (xfrm_msg && xfrm_msg[0])
puts(xfrm_msg);
}
else {
if (one->mode != two->mode) {
printf("old mode %06o\n", one->mode);
printf("new mode %06o\n", two->mode);
}
if (xfrm_msg && xfrm_msg[0])
puts(xfrm_msg);
/*
* we do not run diff between different kind
* of objects.
*/
if ((one->mode ^ two->mode) & S_IFMT)
goto free_ab_and_return;
if (complete_rewrite) {
emit_rewrite_diff(name_a, name_b, one, two);
goto free_ab_and_return;
}
}
if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
die("unable to read files to diff");
if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
printf("Binary files %s and %s differ\n", lbl[0], lbl[1]);
else {
/* Crazy xdl interfaces.. */
const char *diffopts = getenv("GIT_DIFF_OPTS");
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xpparam_t xpp;
xdemitconf_t xecfg;
xdemitcb_t ecb;
struct emit_callback ecbdata;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
ecbdata.label_path = lbl;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xpp.flags = XDF_NEED_MINIMAL;
xecfg.ctxlen = 3;
xecfg.flags = XDL_EMIT_FUNCNAMES;
if (!diffopts)
;
else if (!strncmp(diffopts, "--unified=", 10))
xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
else if (!strncmp(diffopts, "-u", 2))
xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
ecb.outf = fn_out;
ecb.priv = &ecbdata;
Use a *real* built-in diff generator This uses a simplified libxdiff setup to generate unified diffs _without_ doing fork/execve of GNU "diff". This has several huge advantages, for example: Before: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m24.818s user 0m13.332s sys 0m8.664s After: [torvalds@g5 linux]$ time git diff v2.6.16.. > /dev/null real 0m4.563s user 0m2.944s sys 0m1.580s and the fact that this should be a lot more portable (ie we can ignore all the issues with doing fork/execve under Windows). Perhaps even more importantly, this allows us to do diffs without actually ever writing out the git file contents to a temporary file (and without any of the shell quoting issues on filenames etc etc). NOTE! THIS PATCH DOES NOT DO THAT OPTIMIZATION YET! I was lazy, and the current "diff-core" code actually will always write the temp-files, because it used to be something that you simply had to do. So this current one actually writes a temp-file like before, and then reads it into memory again just to do the diff. Stupid. But if this basic infrastructure is accepted, we can start switching over diff-core to not write temp-files, which should speed things up even further, especially when doing big tree-to-tree diffs. Now, in the interest of full disclosure, I should also point out a few downsides: - the libxdiff algorithm is different, and I bet GNU diff has gotten a lot more testing. And the thing is, generating a diff is not an exact science - you can get two different diffs (and you will), and they can both be perfectly valid. So it's not possible to "validate" the libxdiff output by just comparing it against GNU diff. - GNU diff does some nice eye-candy, like trying to figure out what the last function was, and adding that information to the "@@ .." line. libxdiff doesn't do that. - The libxdiff thing has some known deficiencies. In particular, it gets the "\No newline at end of file" case wrong. So this is currently for the experimental branch only. I hope Davide will help fix it. That said, I think the huge performance advantage, and the fact that it integrates better is definitely worth it. But it should go into a development branch at least due to the missing newline issue. Technical note: this is based on libxdiff-0.17, but I did some surgery to get rid of the extraneous fat - stuff that git doesn't need, and seriously cutting down on mmfile_t, which had much more capabilities than the diff algorithm either needed or used. In this version, "mmfile_t" is just a trivial <pointer,length> tuple. That said, I tried to keep the differences to simple removals, so that you can do a diff between this and the libxdiff origin, and you'll basically see just things getting deleted. Even the mmfile_t simplifications are left in a state where the diffs should be readable. Apologies to Davide, whom I'd love to get feedback on this all from (I wrote my own "fill_mmfile()" for the new simpler mmfile_t format: the old complex format had a helper function for that, but I did my surgery with the goal in mind that eventually we _should_ just do mmfile_t mf; buf = read_sha1_file(sha1, type, &size); mf->ptr = buf; mf->size = size; .. use "mf" directly .. which was really a nightmare with the old "helpful" mmfile_t, and really is that easy with the new cut-down interfaces). [ Btw, as any hawk-eye can see from the diff, this was actually generated with itself, so it is "self-hosting". That's about all the testing it has gotten, along with the above kernel diff, which eye-balls correctly, but shows the newline issue when you double-check it with "git-apply" ] Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-03-25 05:13:22 +01:00
xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
}
free_ab_and_return:
free(a_one);
free(b_two);
return;
}
static void builtin_diffstat(const char *name_a, const char *name_b,
struct diff_filespec *one, struct diff_filespec *two,
struct diffstat_t *diffstat)
{
mmfile_t mf1, mf2;
struct diffstat_file *data;
data = diffstat_add(diffstat, name_a ? name_a : name_b);
if (!one || !two) {
data->is_unmerged = 1;
return;
}
if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
die("unable to read files to diff");
if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
data->is_binary = 1;
else {
/* Crazy xdl interfaces.. */
xpparam_t xpp;
xdemitconf_t xecfg;
xdemitcb_t ecb;
xpp.flags = XDF_NEED_MINIMAL;
xecfg.ctxlen = 0;
xecfg.flags = 0;
ecb.outf = xdiff_outf;
ecb.priv = diffstat;
xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
}
}
struct diff_filespec *alloc_filespec(const char *path)
{
int namelen = strlen(path);
struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
memset(spec, 0, sizeof(*spec));
spec->path = (char *)(spec + 1);
memcpy(spec->path, path, namelen+1);
return spec;
}
void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
unsigned short mode)
{
if (mode) {
spec->mode = canon_mode(mode);
memcpy(spec->sha1, sha1, 20);
spec->sha1_valid = !!memcmp(sha1, null_sha1, 20);
}
}
/*
* Given a name and sha1 pair, if the dircache tells us the file in
* the work tree has that object contents, return true, so that
* prepare_temp_file() does not have to inflate and extract.
*/
static int work_tree_matches(const char *name, const unsigned char *sha1)
{
struct cache_entry *ce;
struct stat st;
int pos, len;
/* We do not read the cache ourselves here, because the
* benchmark with my previous version that always reads cache
* shows that it makes things worse for diff-tree comparing
* two linux-2.6 kernel trees in an already checked out work
* tree. This is because most diff-tree comparisons deal with
* only a small number of files, while reading the cache is
* expensive for a large project, and its cost outweighs the
* savings we get by not inflating the object to a temporary
* file. Practically, this code only helps when we are used
* by diff-cache --cached, which does read the cache before
* calling us.
*/
if (!active_cache)
return 0;
len = strlen(name);
pos = cache_name_pos(name, len);
if (pos < 0)
return 0;
ce = active_cache[pos];
if ((lstat(name, &st) < 0) ||
!S_ISREG(st.st_mode) || /* careful! */
"Assume unchanged" git This adds "assume unchanged" logic, started by this message in the list discussion recently: <Pine.LNX.4.64.0601311807470.7301@g5.osdl.org> This is a workaround for filesystems that do not have lstat() that is quick enough for the index mechanism to take advantage of. On the paths marked as "assumed to be unchanged", the user needs to explicitly use update-index to register the object name to be in the next commit. You can use two new options to update-index to set and reset the CE_VALID bit: git-update-index --assume-unchanged path... git-update-index --no-assume-unchanged path... These forms manipulate only the CE_VALID bit; it does not change the object name recorded in the index file. Nor they add a new entry to the index. When the configuration variable "core.ignorestat = true" is set, the index entries are marked with CE_VALID bit automatically after: - update-index to explicitly register the current object name to the index file. - when update-index --refresh finds the path to be up-to-date. - when tools like read-tree -u and apply --index update the working tree file and register the current object name to the index file. The flag is dropped upon read-tree that does not check out the index entry. This happens regardless of the core.ignorestat settings. Index entries marked with CE_VALID bit are assumed to be unchanged most of the time. However, there are cases that CE_VALID bit is ignored for the sake of safety and usability: - while "git-read-tree -m" or git-apply need to make sure that the paths involved in the merge do not have local modifications. This sacrifices performance for safety. - when git-checkout-index -f -q -u -a tries to see if it needs to checkout the paths. Otherwise you can never check anything out ;-). - when git-update-index --really-refresh (a new flag) tries to see if the index entry is up to date. You can start with everything marked as CE_VALID and run this once to drop CE_VALID bit for paths that are modified. Most notably, "update-index --refresh" honours CE_VALID and does not actively stat, so after you modified a file in the working tree, update-index --refresh would not notice until you tell the index about it with "git-update-index path" or "git-update-index --no-assume-unchanged path". This version is not expected to be perfect. I think diff between index and/or tree and working files may need some adjustment, and there probably needs other cases we should automatically unmark paths that are marked to be CE_VALID. But the basics seem to work, and ready to be tested by people who asked for this feature. Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-02-09 06:15:24 +01:00
ce_match_stat(ce, &st, 0) ||
memcmp(sha1, ce->sha1, 20))
return 0;
/* we return 1 only when we can stat, it is a regular file,
* stat information matches, and sha1 recorded in the cache
* matches. I.e. we know the file in the work tree really is
* the same as the <name, sha1> pair.
*/
return 1;
}
static struct sha1_size_cache {
unsigned char sha1[20];
unsigned long size;
} **sha1_size_cache;
static int sha1_size_cache_nr, sha1_size_cache_alloc;
static struct sha1_size_cache *locate_size_cache(unsigned char *sha1,
int find_only,
unsigned long size)
{
int first, last;
struct sha1_size_cache *e;
first = 0;
last = sha1_size_cache_nr;
while (last > first) {
int cmp, next = (last + first) >> 1;
e = sha1_size_cache[next];
cmp = memcmp(e->sha1, sha1, 20);
if (!cmp)
return e;
if (cmp < 0) {
last = next;
continue;
}
first = next+1;
}
/* not found */
if (find_only)
return NULL;
/* insert to make it at "first" */
if (sha1_size_cache_alloc <= sha1_size_cache_nr) {
sha1_size_cache_alloc = alloc_nr(sha1_size_cache_alloc);
sha1_size_cache = xrealloc(sha1_size_cache,
sha1_size_cache_alloc *
sizeof(*sha1_size_cache));
}
sha1_size_cache_nr++;
if (first < sha1_size_cache_nr)
memmove(sha1_size_cache + first + 1, sha1_size_cache + first,
(sha1_size_cache_nr - first - 1) *
sizeof(*sha1_size_cache));
e = xmalloc(sizeof(struct sha1_size_cache));
sha1_size_cache[first] = e;
memcpy(e->sha1, sha1, 20);
e->size = size;
return e;
}
/*
* While doing rename detection and pickaxe operation, we may need to
* grab the data for the blob (or file) for our own in-core comparison.
* diff_filespec has data and size fields for this purpose.
*/
int diff_populate_filespec(struct diff_filespec *s, int size_only)
{
int err = 0;
if (!DIFF_FILE_VALID(s))
die("internal error: asking to populate invalid file.");
if (S_ISDIR(s->mode))
return -1;
if (!use_size_cache)
size_only = 0;
if (s->data)
return err;
if (!s->sha1_valid ||
work_tree_matches(s->path, s->sha1)) {
struct stat st;
int fd;
if (lstat(s->path, &st) < 0) {
if (errno == ENOENT) {
err_empty:
err = -1;
empty:
s->data = "";
s->size = 0;
return err;
}
}
s->size = st.st_size;
if (!s->size)
goto empty;
if (size_only)
return 0;
if (S_ISLNK(st.st_mode)) {
int ret;
s->data = xmalloc(s->size);
s->should_free = 1;
ret = readlink(s->path, s->data, s->size);
if (ret < 0) {
free(s->data);
goto err_empty;
}
return 0;
}
fd = open(s->path, O_RDONLY);
if (fd < 0)
goto err_empty;
s->data = mmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
close(fd);
if (s->data == MAP_FAILED)
goto err_empty;
s->should_munmap = 1;
}
else {
char type[20];
struct sha1_size_cache *e;
if (size_only) {
e = locate_size_cache(s->sha1, 1, 0);
if (e) {
s->size = e->size;
return 0;
}
if (!sha1_object_info(s->sha1, type, &s->size))
locate_size_cache(s->sha1, 0, s->size);
}
else {
s->data = read_sha1_file(s->sha1, type, &s->size);
s->should_free = 1;
}
}
return 0;
}
void diff_free_filespec_data(struct diff_filespec *s)
{
if (s->should_free)
free(s->data);
else if (s->should_munmap)
munmap(s->data, s->size);
s->should_free = s->should_munmap = 0;
s->data = NULL;
free(s->cnt_data);
s->cnt_data = NULL;
}
static void prep_temp_blob(struct diff_tempfile *temp,
void *blob,
unsigned long size,
const unsigned char *sha1,
int mode)
{
int fd;
fd = git_mkstemp(temp->tmp_path, TEMPFILE_PATH_LEN, ".diff_XXXXXX");
if (fd < 0)
die("unable to create temp-file");
if (write(fd, blob, size) != size)
die("unable to write temp-file");
close(fd);
temp->name = temp->tmp_path;
strcpy(temp->hex, sha1_to_hex(sha1));
temp->hex[40] = 0;
sprintf(temp->mode, "%06o", mode);
}
static void prepare_temp_file(const char *name,
struct diff_tempfile *temp,
struct diff_filespec *one)
{
if (!DIFF_FILE_VALID(one)) {
not_a_valid_file:
/* A '-' entry produces this for file-2, and
* a '+' entry produces this for file-1.
*/
temp->name = "/dev/null";
strcpy(temp->hex, ".");
strcpy(temp->mode, ".");
return;
}
if (!one->sha1_valid ||
work_tree_matches(name, one->sha1)) {
struct stat st;
if (lstat(name, &st) < 0) {
if (errno == ENOENT)
goto not_a_valid_file;
die("stat(%s): %s", name, strerror(errno));
}
if (S_ISLNK(st.st_mode)) {
int ret;
char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
if (sizeof(buf) <= st.st_size)
die("symlink too long: %s", name);
ret = readlink(name, buf, st.st_size);
if (ret < 0)
die("readlink(%s)", name);
prep_temp_blob(temp, buf, st.st_size,
(one->sha1_valid ?
one->sha1 : null_sha1),
(one->sha1_valid ?
one->mode : S_IFLNK));
}
else {
/* we can borrow from the file in the work tree */
temp->name = name;
if (!one->sha1_valid)
strcpy(temp->hex, sha1_to_hex(null_sha1));
else
strcpy(temp->hex, sha1_to_hex(one->sha1));
/* Even though we may sometimes borrow the
* contents from the work tree, we always want
* one->mode. mode is trustworthy even when
* !(one->sha1_valid), as long as
* DIFF_FILE_VALID(one).
*/
sprintf(temp->mode, "%06o", one->mode);
}
return;
}
else {
if (diff_populate_filespec(one, 0))
die("cannot read data blob for %s", one->path);
prep_temp_blob(temp, one->data, one->size,
one->sha1, one->mode);
}
}
static void remove_tempfile(void)
{
int i;
for (i = 0; i < 2; i++)
if (diff_temp[i].name == diff_temp[i].tmp_path) {
unlink(diff_temp[i].name);
diff_temp[i].name = NULL;
}
}
static void remove_tempfile_on_signal(int signo)
{
remove_tempfile();
signal(SIGINT, SIG_DFL);
raise(signo);
}
static int spawn_prog(const char *pgm, const char **arg)
{
pid_t pid;
int status;
fflush(NULL);
pid = fork();
if (pid < 0)
die("unable to fork");
if (!pid) {
execvp(pgm, (char *const*) arg);
exit(255);
}
while (waitpid(pid, &status, 0) < 0) {
if (errno == EINTR)
continue;
return -1;
}
/* Earlier we did not check the exit status because
* diff exits non-zero if files are different, and
* we are not interested in knowing that. It was a
* mistake which made it harder to quit a diff-*
* session that uses the git-apply-patch-script as
* the GIT_EXTERNAL_DIFF. A custom GIT_EXTERNAL_DIFF
* should also exit non-zero only when it wants to
* abort the entire diff-* session.
*/
if (WIFEXITED(status) && !WEXITSTATUS(status))
return 0;
return -1;
}
/* An external diff command takes:
*
* diff-cmd name infile1 infile1-sha1 infile1-mode \
* infile2 infile2-sha1 infile2-mode [ rename-to ]
*
*/
static void run_external_diff(const char *pgm,
const char *name,
const char *other,
struct diff_filespec *one,
struct diff_filespec *two,
const char *xfrm_msg,
int complete_rewrite)
{
const char *spawn_arg[10];
struct diff_tempfile *temp = diff_temp;
int retval;
static int atexit_asked = 0;
const char *othername;
const char **arg = &spawn_arg[0];
othername = (other? other : name);
if (one && two) {
prepare_temp_file(name, &temp[0], one);
prepare_temp_file(othername, &temp[1], two);
if (! atexit_asked &&
(temp[0].name == temp[0].tmp_path ||
temp[1].name == temp[1].tmp_path)) {
atexit_asked = 1;
atexit(remove_tempfile);
}
signal(SIGINT, remove_tempfile_on_signal);
}
if (one && two) {
*arg++ = pgm;
*arg++ = name;
*arg++ = temp[0].name;
*arg++ = temp[0].hex;
*arg++ = temp[0].mode;
*arg++ = temp[1].name;
*arg++ = temp[1].hex;
*arg++ = temp[1].mode;
if (other) {
*arg++ = other;
*arg++ = xfrm_msg;
}
} else {
*arg++ = pgm;
*arg++ = name;
}
*arg = NULL;
retval = spawn_prog(pgm, spawn_arg);
remove_tempfile();
if (retval) {
fprintf(stderr, "external diff died, stopping at %s.\n", name);
exit(1);
}
}
static void run_diff_cmd(const char *pgm,
const char *name,
const char *other,
struct diff_filespec *one,
struct diff_filespec *two,
const char *xfrm_msg,
int complete_rewrite)
{
if (pgm) {
run_external_diff(pgm, name, other, one, two, xfrm_msg,
complete_rewrite);
return;
}
if (one && two)
builtin_diff(name, other ? other : name,
one, two, xfrm_msg, complete_rewrite);
else
printf("* Unmerged path %s\n", name);
}
static void diff_fill_sha1_info(struct diff_filespec *one)
{
if (DIFF_FILE_VALID(one)) {
if (!one->sha1_valid) {
struct stat st;
if (lstat(one->path, &st) < 0)
die("stat %s", one->path);
if (index_path(one->sha1, one->path, &st, 0))
die("cannot hash %s\n", one->path);
}
}
else
memset(one->sha1, 0, 20);
}
static void run_diff(struct diff_filepair *p, struct diff_options *o)
{
const char *pgm = external_diff();
char msg[PATH_MAX*2+300], *xfrm_msg;
struct diff_filespec *one;
struct diff_filespec *two;
const char *name;
const char *other;
char *name_munged, *other_munged;
int complete_rewrite = 0;
int len;
if (DIFF_PAIR_UNMERGED(p)) {
/* unmerged */
run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, 0);
return;
}
name = p->one->path;
other = (strcmp(name, p->two->path) ? p->two->path : NULL);
name_munged = quote_one(name);
other_munged = quote_one(other);
one = p->one; two = p->two;
diff_fill_sha1_info(one);
diff_fill_sha1_info(two);
len = 0;
switch (p->status) {
case DIFF_STATUS_COPIED:
len += snprintf(msg + len, sizeof(msg) - len,
"similarity index %d%%\n"
"copy from %s\n"
"copy to %s\n",
(int)(0.5 + p->score * 100.0/MAX_SCORE),
name_munged, other_munged);
break;
case DIFF_STATUS_RENAMED:
len += snprintf(msg + len, sizeof(msg) - len,
"similarity index %d%%\n"
"rename from %s\n"
"rename to %s\n",
(int)(0.5 + p->score * 100.0/MAX_SCORE),
name_munged, other_munged);
break;
case DIFF_STATUS_MODIFIED:
if (p->score) {
len += snprintf(msg + len, sizeof(msg) - len,
"dissimilarity index %d%%\n",
(int)(0.5 + p->score *
100.0/MAX_SCORE));
complete_rewrite = 1;
break;
}
/* fallthru */
default:
/* nothing */
;
}
if (memcmp(one->sha1, two->sha1, 20)) {
char one_sha1[41];
int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
memcpy(one_sha1, sha1_to_hex(one->sha1), 41);
len += snprintf(msg + len, sizeof(msg) - len,
"index %.*s..%.*s",
abbrev, one_sha1, abbrev,
sha1_to_hex(two->sha1));
if (one->mode == two->mode)
len += snprintf(msg + len, sizeof(msg) - len,
" %06o", one->mode);
len += snprintf(msg + len, sizeof(msg) - len, "\n");
}
if (len)
msg[--len] = 0;
xfrm_msg = len ? msg : NULL;
if (!pgm &&
DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
(S_IFMT & one->mode) != (S_IFMT & two->mode)) {
/* a filepair that changes between file and symlink
* needs to be split into deletion and creation.
*/
struct diff_filespec *null = alloc_filespec(two->path);
run_diff_cmd(NULL, name, other, one, null, xfrm_msg, 0);
free(null);
null = alloc_filespec(one->path);
run_diff_cmd(NULL, name, other, null, two, xfrm_msg, 0);
free(null);
}
else
run_diff_cmd(pgm, name, other, one, two, xfrm_msg,
complete_rewrite);
free(name_munged);
free(other_munged);
}
static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
struct diffstat_t *diffstat)
{
const char *name;
const char *other;
if (DIFF_PAIR_UNMERGED(p)) {
/* unmerged */
builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat);
return;
}
name = p->one->path;
other = (strcmp(name, p->two->path) ? p->two->path : NULL);
diff_fill_sha1_info(p->one);
diff_fill_sha1_info(p->two);
builtin_diffstat(name, other, p->one, p->two, diffstat);
}
void diff_setup(struct diff_options *options)
{
memset(options, 0, sizeof(*options));
options->output_format = DIFF_FORMAT_RAW;
options->line_termination = '\n';
options->break_opt = -1;
options->rename_limit = -1;
options->change = diff_change;
options->add_remove = diff_addremove;
}
int diff_setup_done(struct diff_options *options)
{
if ((options->find_copies_harder &&
options->detect_rename != DIFF_DETECT_COPY) ||
(0 <= options->rename_limit && !options->detect_rename))
return -1;
/*
* These cases always need recursive; we do not drop caller-supplied
* recursive bits for other formats here.
*/
if ((options->output_format == DIFF_FORMAT_PATCH) ||
(options->output_format == DIFF_FORMAT_DIFFSTAT) ||
(options->with_stat))
options->recursive = 1;
if (options->detect_rename && options->rename_limit < 0)
options->rename_limit = diff_rename_limit_default;
if (options->setup & DIFF_SETUP_USE_CACHE) {
if (!active_cache)
/* read-cache does not die even when it fails
* so it is safe for us to do this here. Also
* it does not smudge active_cache or active_nr
* when it fails, so we do not have to worry about
* cleaning it up ourselves either.
*/
read_cache();
}
if (options->setup & DIFF_SETUP_USE_SIZE_CACHE)
use_size_cache = 1;
if (options->abbrev <= 0 || 40 < options->abbrev)
options->abbrev = 40; /* full */
return 0;
}
int diff_opt_parse(struct diff_options *options, const char **av, int ac)
{
const char *arg = av[0];
if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
options->output_format = DIFF_FORMAT_PATCH;
else if (!strcmp(arg, "--patch-with-raw")) {
options->output_format = DIFF_FORMAT_PATCH;
options->with_raw = 1;
}
else if (!strcmp(arg, "--stat"))
options->output_format = DIFF_FORMAT_DIFFSTAT;
else if (!strcmp(arg, "--patch-with-stat")) {
options->output_format = DIFF_FORMAT_PATCH;
options->with_stat = 1;
}
else if (!strcmp(arg, "-z"))
options->line_termination = 0;
else if (!strncmp(arg, "-l", 2))
options->rename_limit = strtoul(arg+2, NULL, 10);
else if (!strcmp(arg, "--full-index"))
options->full_index = 1;
else if (!strcmp(arg, "--name-only"))
options->output_format = DIFF_FORMAT_NAME;
else if (!strcmp(arg, "--name-status"))
options->output_format = DIFF_FORMAT_NAME_STATUS;
else if (!strcmp(arg, "-R"))
options->reverse_diff = 1;
else if (!strncmp(arg, "-S", 2))
options->pickaxe = arg + 2;
else if (!strcmp(arg, "-s"))
options->output_format = DIFF_FORMAT_NO_OUTPUT;
else if (!strncmp(arg, "-O", 2))
options->orderfile = arg + 2;
else if (!strncmp(arg, "--diff-filter=", 14))
options->filter = arg + 14;
else if (!strcmp(arg, "--pickaxe-all"))
options->pickaxe_opts = DIFF_PICKAXE_ALL;
else if (!strcmp(arg, "--pickaxe-regex"))
options->pickaxe_opts = DIFF_PICKAXE_REGEX;
else if (!strncmp(arg, "-B", 2)) {
if ((options->break_opt =
diff_scoreopt_parse(arg)) == -1)
return -1;
}
else if (!strncmp(arg, "-M", 2)) {
if ((options->rename_score =
diff_scoreopt_parse(arg)) == -1)
return -1;
options->detect_rename = DIFF_DETECT_RENAME;
}
else if (!strncmp(arg, "-C", 2)) {
if ((options->rename_score =
diff_scoreopt_parse(arg)) == -1)
return -1;
options->detect_rename = DIFF_DETECT_COPY;
}
else if (!strcmp(arg, "--find-copies-harder"))
options->find_copies_harder = 1;
else if (!strcmp(arg, "--abbrev"))
options->abbrev = DEFAULT_ABBREV;
else if (!strncmp(arg, "--abbrev=", 9)) {
options->abbrev = strtoul(arg + 9, NULL, 10);
if (options->abbrev < MINIMUM_ABBREV)
options->abbrev = MINIMUM_ABBREV;
else if (40 < options->abbrev)
options->abbrev = 40;
}
else
return 0;
return 1;
}
static int parse_num(const char **cp_p)
{
unsigned long num, scale;
int ch, dot;
const char *cp = *cp_p;
num = 0;
scale = 1;
dot = 0;
for(;;) {
ch = *cp;
if ( !dot && ch == '.' ) {
scale = 1;
dot = 1;
} else if ( ch == '%' ) {
scale = dot ? scale*100 : 100;
cp++; /* % is always at the end */
break;
} else if ( ch >= '0' && ch <= '9' ) {
if ( scale < 100000 ) {
scale *= 10;
num = (num*10) + (ch-'0');
}
} else {
break;
}
cp++;
}
*cp_p = cp;
/* user says num divided by scale and we say internally that
* is MAX_SCORE * num / scale.
*/
return (num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale);
}
int diff_scoreopt_parse(const char *opt)
{
[PATCH] diff: Update -B heuristics. As Linus pointed out on the mailing list discussion, -B should break a files that has many inserts even if it still keeps enough of the original contents, so that the broken pieces can later be matched with other files by -M or -C. However, if such a broken pair does not get picked up by -M or -C, we would want to apply different criteria; namely, regardless of the amount of new material in the result, the determination of "rewrite" should be done by looking at the amount of original material still left in the result. If you still have the original 97 lines from a 100-line document, it does not matter if you add your own 13 lines to make a 110-line document, or if you add 903 lines to make a 1000-line document. It is not a rewrite but an in-place edit. On the other hand, if you did lose 97 lines from the original, it does not matter if you added 27 lines to make a 30-line document or if you added 997 lines to make a 1000-line document. You did a complete rewrite in either case. This patch introduces a post-processing phase that runs after diffcore-rename matches up broken pairs diffcore-break creates. The purpose of this post-processing is to pick up these broken pieces and merge them back into in-place modifications. For this, the score parameter -B option takes is changed into a pair of numbers, and it takes "-B99/80" format when fully spelled out. The first number is the minimum amount of "edit" (same definition as what diffcore-rename uses, which is "sum of deletion and insertion") that a modification needs to have to be broken, and the second number is the minimum amount of "delete" a surviving broken pair must have to avoid being merged back together. It can be abbreviated to "-B" to use default for both, "-B9" or "-B9/" to use 90% for "edit" but default (80%) for merge avoidance, or "-B/75" to use default (99%) "edit" and 75% for merge avoidance. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-03 10:40:28 +02:00
int opt1, opt2, cmd;
if (*opt++ != '-')
return -1;
cmd = *opt++;
if (cmd != 'M' && cmd != 'C' && cmd != 'B')
return -1; /* that is not a -M, -C nor -B option */
opt1 = parse_num(&opt);
[PATCH] diff: Update -B heuristics. As Linus pointed out on the mailing list discussion, -B should break a files that has many inserts even if it still keeps enough of the original contents, so that the broken pieces can later be matched with other files by -M or -C. However, if such a broken pair does not get picked up by -M or -C, we would want to apply different criteria; namely, regardless of the amount of new material in the result, the determination of "rewrite" should be done by looking at the amount of original material still left in the result. If you still have the original 97 lines from a 100-line document, it does not matter if you add your own 13 lines to make a 110-line document, or if you add 903 lines to make a 1000-line document. It is not a rewrite but an in-place edit. On the other hand, if you did lose 97 lines from the original, it does not matter if you added 27 lines to make a 30-line document or if you added 997 lines to make a 1000-line document. You did a complete rewrite in either case. This patch introduces a post-processing phase that runs after diffcore-rename matches up broken pairs diffcore-break creates. The purpose of this post-processing is to pick up these broken pieces and merge them back into in-place modifications. For this, the score parameter -B option takes is changed into a pair of numbers, and it takes "-B99/80" format when fully spelled out. The first number is the minimum amount of "edit" (same definition as what diffcore-rename uses, which is "sum of deletion and insertion") that a modification needs to have to be broken, and the second number is the minimum amount of "delete" a surviving broken pair must have to avoid being merged back together. It can be abbreviated to "-B" to use default for both, "-B9" or "-B9/" to use 90% for "edit" but default (80%) for merge avoidance, or "-B/75" to use default (99%) "edit" and 75% for merge avoidance. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-03 10:40:28 +02:00
if (cmd != 'B')
opt2 = 0;
else {
if (*opt == 0)
opt2 = 0;
else if (*opt != '/')
return -1; /* we expect -B80/99 or -B80 */
else {
opt++;
opt2 = parse_num(&opt);
}
}
if (*opt != 0)
return -1;
[PATCH] diff: Update -B heuristics. As Linus pointed out on the mailing list discussion, -B should break a files that has many inserts even if it still keeps enough of the original contents, so that the broken pieces can later be matched with other files by -M or -C. However, if such a broken pair does not get picked up by -M or -C, we would want to apply different criteria; namely, regardless of the amount of new material in the result, the determination of "rewrite" should be done by looking at the amount of original material still left in the result. If you still have the original 97 lines from a 100-line document, it does not matter if you add your own 13 lines to make a 110-line document, or if you add 903 lines to make a 1000-line document. It is not a rewrite but an in-place edit. On the other hand, if you did lose 97 lines from the original, it does not matter if you added 27 lines to make a 30-line document or if you added 997 lines to make a 1000-line document. You did a complete rewrite in either case. This patch introduces a post-processing phase that runs after diffcore-rename matches up broken pairs diffcore-break creates. The purpose of this post-processing is to pick up these broken pieces and merge them back into in-place modifications. For this, the score parameter -B option takes is changed into a pair of numbers, and it takes "-B99/80" format when fully spelled out. The first number is the minimum amount of "edit" (same definition as what diffcore-rename uses, which is "sum of deletion and insertion") that a modification needs to have to be broken, and the second number is the minimum amount of "delete" a surviving broken pair must have to avoid being merged back together. It can be abbreviated to "-B" to use default for both, "-B9" or "-B9/" to use 90% for "edit" but default (80%) for merge avoidance, or "-B/75" to use default (99%) "edit" and 75% for merge avoidance. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-03 10:40:28 +02:00
return opt1 | (opt2 << 16);
}
struct diff_queue_struct diff_queued_diff;
void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
{
if (queue->alloc <= queue->nr) {
queue->alloc = alloc_nr(queue->alloc);
queue->queue = xrealloc(queue->queue,
sizeof(dp) * queue->alloc);
}
queue->queue[queue->nr++] = dp;
}
struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
struct diff_filespec *one,
struct diff_filespec *two)
{
struct diff_filepair *dp = xmalloc(sizeof(*dp));
dp->one = one;
dp->two = two;
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
dp->score = 0;
dp->status = 0;
dp->source_stays = 0;
dp->broken_pair = 0;
if (queue)
diff_q(queue, dp);
return dp;
}
void diff_free_filepair(struct diff_filepair *p)
{
diff_free_filespec_data(p->one);
diff_free_filespec_data(p->two);
free(p->one);
free(p->two);
free(p);
}
/* This is different from find_unique_abbrev() in that
* it stuffs the result with dots for alignment.
*/
const char *diff_unique_abbrev(const unsigned char *sha1, int len)
{
int abblen;
const char *abbrev;
if (len == 40)
return sha1_to_hex(sha1);
abbrev = find_unique_abbrev(sha1, len);
if (!abbrev)
return sha1_to_hex(sha1);
abblen = strlen(abbrev);
if (abblen < 37) {
static char hex[41];
if (len < abblen && abblen <= len + 2)
sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
else
sprintf(hex, "%s...", abbrev);
return hex;
}
return sha1_to_hex(sha1);
}
static void diff_flush_raw(struct diff_filepair *p,
int line_termination,
int inter_name_termination,
struct diff_options *options,
int output_format)
{
int two_paths;
char status[10];
int abbrev = options->abbrev;
const char *path_one, *path_two;
path_one = p->one->path;
path_two = p->two->path;
if (line_termination) {
path_one = quote_one(path_one);
path_two = quote_one(path_two);
}
if (p->score)
sprintf(status, "%c%03d", p->status,
(int)(0.5 + p->score * 100.0/MAX_SCORE));
else {
status[0] = p->status;
status[1] = 0;
}
switch (p->status) {
case DIFF_STATUS_COPIED:
case DIFF_STATUS_RENAMED:
two_paths = 1;
break;
case DIFF_STATUS_ADDED:
case DIFF_STATUS_DELETED:
two_paths = 0;
break;
default:
two_paths = 0;
break;
}
if (output_format != DIFF_FORMAT_NAME_STATUS) {
printf(":%06o %06o %s ",
p->one->mode, p->two->mode,
diff_unique_abbrev(p->one->sha1, abbrev));
printf("%s ",
diff_unique_abbrev(p->two->sha1, abbrev));
}
printf("%s%c%s", status, inter_name_termination, path_one);
if (two_paths)
printf("%c%s", inter_name_termination, path_two);
putchar(line_termination);
if (path_one != p->one->path)
free((void*)path_one);
if (path_two != p->two->path)
free((void*)path_two);
}
static void diff_flush_name(struct diff_filepair *p,
int inter_name_termination,
int line_termination)
{
char *path = p->two->path;
if (line_termination)
path = quote_one(p->two->path);
else
path = p->two->path;
printf("%s%c", path, line_termination);
if (p->two->path != path)
free(path);
}
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
int diff_unmodified_pair(struct diff_filepair *p)
{
/* This function is written stricter than necessary to support
* the currently implemented transformers, but the idea is to
* let transformers to produce diff_filepairs any way they want,
* and filter and clean them up here before producing the output.
*/
struct diff_filespec *one, *two;
if (DIFF_PAIR_UNMERGED(p))
return 0; /* unmerged is interesting */
one = p->one;
two = p->two;
/* deletion, addition, mode or type change
* and rename are all interesting.
*/
if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
DIFF_PAIR_MODE_CHANGED(p) ||
strcmp(one->path, two->path))
return 0;
/* both are valid and point at the same path. that is, we are
* dealing with a change.
*/
if (one->sha1_valid && two->sha1_valid &&
!memcmp(one->sha1, two->sha1, sizeof(one->sha1)))
return 1; /* no change */
if (!one->sha1_valid && !two->sha1_valid)
return 1; /* both look at the same file on the filesystem. */
return 0;
}
static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
{
if (diff_unmodified_pair(p))
return;
if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
(DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
return; /* no tree diffs in patch format */
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
run_diff(p, o);
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
}
static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
struct diffstat_t *diffstat)
{
if (diff_unmodified_pair(p))
return;
if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
(DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
return; /* no tree diffs in patch format */
run_diffstat(p, o, diffstat);
}
int diff_queue_is_empty(void)
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
{
struct diff_queue_struct *q = &diff_queued_diff;
int i;
for (i = 0; i < q->nr; i++)
if (!diff_unmodified_pair(q->queue[i]))
return 0;
return 1;
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
}
#if DIFF_DEBUG
void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
{
fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
x, one ? one : "",
s->path,
DIFF_FILE_VALID(s) ? "valid" : "invalid",
s->mode,
s->sha1_valid ? sha1_to_hex(s->sha1) : "");
fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
x, one ? one : "",
s->size, s->xfrm_flags);
}
void diff_debug_filepair(const struct diff_filepair *p, int i)
{
diff_debug_filespec(p->one, i, "one");
diff_debug_filespec(p->two, i, "two");
fprintf(stderr, "score %d, status %c stays %d broken %d\n",
p->score, p->status ? p->status : '?',
p->source_stays, p->broken_pair);
}
void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
{
int i;
if (msg)
fprintf(stderr, "%s\n", msg);
fprintf(stderr, "q->nr = %d\n", q->nr);
for (i = 0; i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
diff_debug_filepair(p, i);
}
}
#endif
static void diff_resolve_rename_copy(void)
{
int i, j;
struct diff_filepair *p, *pp;
struct diff_queue_struct *q = &diff_queued_diff;
diff_debug_queue("resolve-rename-copy", q);
for (i = 0; i < q->nr; i++) {
p = q->queue[i];
p->status = 0; /* undecided */
if (DIFF_PAIR_UNMERGED(p))
p->status = DIFF_STATUS_UNMERGED;
else if (!DIFF_FILE_VALID(p->one))
p->status = DIFF_STATUS_ADDED;
else if (!DIFF_FILE_VALID(p->two))
p->status = DIFF_STATUS_DELETED;
else if (DIFF_PAIR_TYPE_CHANGED(p))
p->status = DIFF_STATUS_TYPE_CHANGED;
/* from this point on, we are dealing with a pair
* whose both sides are valid and of the same type, i.e.
* either in-place edit or rename/copy edit.
*/
else if (DIFF_PAIR_RENAME(p)) {
if (p->source_stays) {
p->status = DIFF_STATUS_COPIED;
continue;
}
/* See if there is some other filepair that
* copies from the same source as us. If so
* we are a copy. Otherwise we are either a
* copy if the path stays, or a rename if it
* does not, but we already handled "stays" case.
*/
for (j = i + 1; j < q->nr; j++) {
pp = q->queue[j];
if (strcmp(pp->one->path, p->one->path))
continue; /* not us */
if (!DIFF_PAIR_RENAME(pp))
continue; /* not a rename/copy */
/* pp is a rename/copy from the same source */
p->status = DIFF_STATUS_COPIED;
break;
}
if (!p->status)
p->status = DIFF_STATUS_RENAMED;
}
else if (memcmp(p->one->sha1, p->two->sha1, 20) ||
p->one->mode != p->two->mode)
p->status = DIFF_STATUS_MODIFIED;
else {
/* This is a "no-change" entry and should not
* happen anymore, but prepare for broken callers.
*/
error("feeding unmodified %s to diffcore",
p->one->path);
p->status = DIFF_STATUS_UNKNOWN;
}
}
diff_debug_queue("resolve-rename-copy done", q);
}
static void flush_one_pair(struct diff_filepair *p,
int diff_output_format,
struct diff_options *options,
struct diffstat_t *diffstat)
{
int inter_name_termination = '\t';
int line_termination = options->line_termination;
if (!line_termination)
inter_name_termination = 0;
switch (p->status) {
case DIFF_STATUS_UNKNOWN:
break;
case 0:
die("internal error in diff-resolve-rename-copy");
break;
default:
switch (diff_output_format) {
case DIFF_FORMAT_DIFFSTAT:
diff_flush_stat(p, options, diffstat);
break;
case DIFF_FORMAT_PATCH:
diff_flush_patch(p, options);
break;
case DIFF_FORMAT_RAW:
case DIFF_FORMAT_NAME_STATUS:
diff_flush_raw(p, line_termination,
inter_name_termination,
options, diff_output_format);
break;
case DIFF_FORMAT_NAME:
diff_flush_name(p,
inter_name_termination,
line_termination);
break;
case DIFF_FORMAT_NO_OUTPUT:
break;
}
}
}
void diff_flush(struct diff_options *options)
{
struct diff_queue_struct *q = &diff_queued_diff;
int i;
int diff_output_format = options->output_format;
struct diffstat_t *diffstat = NULL;
if (diff_output_format == DIFF_FORMAT_DIFFSTAT || options->with_stat) {
diffstat = xcalloc(sizeof (struct diffstat_t), 1);
diffstat->xm.consume = diffstat_consume;
}
if (options->with_raw) {
for (i = 0; i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
flush_one_pair(p, DIFF_FORMAT_RAW, options, NULL);
}
putchar(options->line_termination);
}
if (options->with_stat) {
for (i = 0; i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
flush_one_pair(p, DIFF_FORMAT_DIFFSTAT, options,
diffstat);
}
show_stats(diffstat);
free(diffstat);
diffstat = NULL;
putchar(options->line_termination);
}
for (i = 0; i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
flush_one_pair(p, diff_output_format, options, diffstat);
diff_free_filepair(p);
}
if (diffstat) {
show_stats(diffstat);
free(diffstat);
}
free(q->queue);
q->queue = NULL;
q->nr = q->alloc = 0;
}
static void diffcore_apply_filter(const char *filter)
{
int i;
struct diff_queue_struct *q = &diff_queued_diff;
struct diff_queue_struct outq;
outq.queue = NULL;
outq.nr = outq.alloc = 0;
if (!filter)
return;
if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
int found;
for (i = found = 0; !found && i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
if (((p->status == DIFF_STATUS_MODIFIED) &&
((p->score &&
strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
(!p->score &&
strchr(filter, DIFF_STATUS_MODIFIED)))) ||
((p->status != DIFF_STATUS_MODIFIED) &&
strchr(filter, p->status)))
found++;
}
if (found)
return;
/* otherwise we will clear the whole queue
* by copying the empty outq at the end of this
* function, but first clear the current entries
* in the queue.
*/
for (i = 0; i < q->nr; i++)
diff_free_filepair(q->queue[i]);
}
else {
/* Only the matching ones */
for (i = 0; i < q->nr; i++) {
struct diff_filepair *p = q->queue[i];
if (((p->status == DIFF_STATUS_MODIFIED) &&
((p->score &&
strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
(!p->score &&
strchr(filter, DIFF_STATUS_MODIFIED)))) ||
((p->status != DIFF_STATUS_MODIFIED) &&
strchr(filter, p->status)))
diff_q(&outq, p);
else
diff_free_filepair(p);
}
}
free(q->queue);
*q = outq;
}
void diffcore_std(struct diff_options *options)
{
if (options->break_opt != -1)
diffcore_break(options->break_opt);
if (options->detect_rename)
diffcore_rename(options);
if (options->break_opt != -1)
[PATCH] diff: Update -B heuristics. As Linus pointed out on the mailing list discussion, -B should break a files that has many inserts even if it still keeps enough of the original contents, so that the broken pieces can later be matched with other files by -M or -C. However, if such a broken pair does not get picked up by -M or -C, we would want to apply different criteria; namely, regardless of the amount of new material in the result, the determination of "rewrite" should be done by looking at the amount of original material still left in the result. If you still have the original 97 lines from a 100-line document, it does not matter if you add your own 13 lines to make a 110-line document, or if you add 903 lines to make a 1000-line document. It is not a rewrite but an in-place edit. On the other hand, if you did lose 97 lines from the original, it does not matter if you added 27 lines to make a 30-line document or if you added 997 lines to make a 1000-line document. You did a complete rewrite in either case. This patch introduces a post-processing phase that runs after diffcore-rename matches up broken pairs diffcore-break creates. The purpose of this post-processing is to pick up these broken pieces and merge them back into in-place modifications. For this, the score parameter -B option takes is changed into a pair of numbers, and it takes "-B99/80" format when fully spelled out. The first number is the minimum amount of "edit" (same definition as what diffcore-rename uses, which is "sum of deletion and insertion") that a modification needs to have to be broken, and the second number is the minimum amount of "delete" a surviving broken pair must have to avoid being merged back together. It can be abbreviated to "-B" to use default for both, "-B9" or "-B9/" to use 90% for "edit" but default (80%) for merge avoidance, or "-B/75" to use default (99%) "edit" and 75% for merge avoidance. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-03 10:40:28 +02:00
diffcore_merge_broken();
if (options->pickaxe)
diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
if (options->orderfile)
diffcore_order(options->orderfile);
diff_resolve_rename_copy();
diffcore_apply_filter(options->filter);
}
void diffcore_std_no_resolve(struct diff_options *options)
{
if (options->pickaxe)
diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
if (options->orderfile)
diffcore_order(options->orderfile);
diffcore_apply_filter(options->filter);
}
void diff_addremove(struct diff_options *options,
int addremove, unsigned mode,
const unsigned char *sha1,
const char *base, const char *path)
{
char concatpath[PATH_MAX];
struct diff_filespec *one, *two;
/* This may look odd, but it is a preparation for
* feeding "there are unchanged files which should
* not produce diffs, but when you are doing copy
* detection you would need them, so here they are"
* entries to the diff-core. They will be prefixed
* with something like '=' or '*' (I haven't decided
* which but should not make any difference).
[PATCH] Rename/copy detection fix. The rename/copy detection logic in earlier round was only good enough to show patch output and discussion on the mailing list about the diff-raw format updates revealed many problems with it. This patch fixes all the ones known to me, without making things I want to do later impossible, mostly related to patch reordering. (1) Earlier rename/copy detector determined which one is rename and which one is copy too early, which made it impossible to later introduce diffcore transformers to reorder patches. This patch fixes it by moving that logic to the very end of the processing. (2) Earlier output routine diff_flush() was pruning all the "no-change" entries indiscriminatingly. This was done due to my false assumption that one of the requirements in the diff-raw output was not to show such an entry (which resulted in my incorrect comment about "diff-helper never being able to be equivalent to built-in diff driver"). My special thanks go to Linus for correcting me about this. When we produce diff-raw output, for the downstream to be able to tell renames from copies, sometimes it _is_ necessary to output "no-change" entries, and this patch adds diffcore_prune() function for doing it. (3) Earlier diff_filepair structure was trying to be not too specific about rename/copy operations, but the purpose of the structure was to record one or two paths, which _was_ indeed about rename/copy. This patch discards xfrm_msg field which was trying to be generic for this wrong reason, and introduces a couple of fields (rename_score and rename_rank) that are explicitly specific to rename/copy logic. One thing to note is that the information in a single diff_filepair structure _still_ does not distinguish renames from copies, and it is deliberately so. This is to allow patches to be reordered in later stages. (4) This patch also adds some tests about diff-raw format output and makes sure that necessary "no-change" entries appear on the output. Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-23 06:26:09 +02:00
* Feeding the same new and old to diff_change()
* also has the same effect.
* Before the final output happens, they are pruned after
* merged into rename/copy pairs as appropriate.
*/
if (options->reverse_diff)
addremove = (addremove == '+' ? '-' :
addremove == '-' ? '+' : addremove);
if (!path) path = "";
sprintf(concatpath, "%s%s", base, path);
one = alloc_filespec(concatpath);
two = alloc_filespec(concatpath);
if (addremove != '+')
fill_filespec(one, sha1, mode);
if (addremove != '-')
fill_filespec(two, sha1, mode);
diff_queue(&diff_queued_diff, one, two);
}
void diff_change(struct diff_options *options,
unsigned old_mode, unsigned new_mode,
const unsigned char *old_sha1,
const unsigned char *new_sha1,
const char *base, const char *path)
{
char concatpath[PATH_MAX];
struct diff_filespec *one, *two;
if (options->reverse_diff) {
unsigned tmp;
const unsigned char *tmp_c;
tmp = old_mode; old_mode = new_mode; new_mode = tmp;
tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
}
if (!path) path = "";
sprintf(concatpath, "%s%s", base, path);
one = alloc_filespec(concatpath);
two = alloc_filespec(concatpath);
fill_filespec(one, old_sha1, old_mode);
fill_filespec(two, new_sha1, new_mode);
diff_queue(&diff_queued_diff, one, two);
}
void diff_unmerge(struct diff_options *options,
const char *path)
{
struct diff_filespec *one, *two;
one = alloc_filespec(path);
two = alloc_filespec(path);
diff_queue(&diff_queued_diff, one, two);
}