The check_to_create_blob() function used to check only the case
where we are applying to the working tree. Rename the function to
check_to_create() and make it also responsible for checking the case
where we apply to the index. Also make its caller responsible for
issuing an error message.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
load_preimage() is very specific to grab the current contents for
the path given by patch->old_name. Split the logic that grabs the
contents for a path out of it into a separate load_patch_target()
function.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code to grab the result of application of a previous patch in the
input was mixed with error message generation for a case where a later
patch tries to modify contents of a path that has been removed.
The same code is duplicated elsewhere in the code. Introduce a helper
to clarify what is going on.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Given a patch for a single path, the function apply_data() reads the
preimage in core, and applies the change represented in the patch.
Separate out the first part that reads the preimage into a separate
helper function load_preimage().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a patch wants to touch a path, if the path exists in the index
but is missing in the working tree, "git apply --index" checks out
the file to the working tree from the index automatically and then
applies the patch.
Split this logic out to a separate helper function.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reading a blob out of the object store does not have to require that the
caller has a cache entry for it.
Create a read_blob_object() helper function that takes the object name and
mode, and use it to reimplement the original function as a thin wrapper to
it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The clear_image() function did not clear the line table in the image
structure; this does not matter for the current callers, as the function
is only called from the codepaths that deal with binary patches where the
line table is never populated, and the codepaths that do populate the line
table free it themselves.
But it will start to matter when we introduce a codepath to retry a failed
patch, so make sure it clears and frees everything.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code is littered with to_be_deleted() whose purpose is not so clear.
Describe where it matters. Also remove an extra space before "#define"
that snuck in by mistake at 7fac0ee (builtin-apply: keep information about
files to be deleted, 2009-04-11).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This check is not only about type-change (for which it would be
sufficient to check only was_deleted()) but is also about a swap
rename. Otherwise to_be_deleted() check is not justified.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
More message strings marked for i18n.
By Nguyễn Thái Ngọc Duy (10) and Jonathan Nieder (1)
* nd/i18n:
help: replace underlining "help -a" headers using hyphens with a blank line
i18n: bundle: mark strings for translation
i18n: index-pack: mark strings for translation
i18n: apply: update say_patch_name to give translators complete sentence
i18n: apply: mark strings for translation
i18n: remote: mark strings for translation
i18n: make warn_dangling_symref() automatically append \n
i18n: help: mark strings for translation
i18n: mark relative dates for translation
strbuf: convenience format functions with \n automatically appended
Makefile: feed all header files to xgettext
Valgrind reports quite a lot of discarded memory inside apply.
Fix them, audit and document the buffer ownership rules.
By Junio C Hamano (8) and Jared Hance (1)
* jh/apply-free-patch:
apply: document buffer ownership rules across functions
apply: tighten constness of line buffer
apply: drop unused macro
apply: free unused fragments for submodule patch
apply: free patch->result
apply: release memory for fn_table
apply: free patch->{def,old,new}_name fields
apply: rename free_patch() to free_patch_list()
apply: do not leak patches and fragments
In general, the private functions in this file were not very
much documented; even though what each of them do is reasonably
self explanatory, the ownership rules for various buffers and
data structures were not very obvious.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These point into a single line in the patch text we read from
the input, and they are not used to modify it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We simply discarded the fragments that we are not going to use upon seeing
a patch to update the submodule commit bound at path that we have not
checked out.
Free these fragments, not to leak them.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is by far the largest piece of data, much larger than the patch and
fragment structures or the three name fields in the patch structure.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The fn_table is used to record the result of earlier patch application in
case a hand-crafted input file contains multiple patches to the same file.
Both its string key (filename) and the contents are borrowed from "struct
patch" that represents the previous application in the same apply_patch()
call, and they do not leak, but the table itself was not freed properly.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These were all allocated in the heap by parsing the header parts of the
patch, but we did not bother to free them. Some used to share the memory
(e.g. copying def_name to old_name) so this is not just the matter of
adding three calls to free().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the while loop inside apply_patch, patch and fragments are
dynamically allocated with a calloc. However, only unused patches
are actually freed and the rest are left to leak. Since a list is
actively built up consisting of the used patches, they can simply be
iterated and freed at the end of the function.
In addition, the text in fragments were not freed, primarily because
they mostly point into a patch text that is freed as a whole. But
there are some cases where new piece of memory is allocated and
pointed by a fragment (namely, when handling a binary patch).
Introduce a free_patch bitfield to mark each fragment if its text
needs to be freed, and free patches, fragments and fragment text
that need to be freed when we are done with the input.
Signed-off-by: Jared Hance <jaredhance@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git diff --stat" and "git apply --stat" now learn to print the line
"%d files changed, %d insertions(+), %d deletions(-)" in singular form
whenever applicable. "0 insertions" and "0 deletions" are also omitted
unless they are both zero.
This matches how versions of "diffstat" that are not prehistoric produced
their output, and also makes this line translatable.
[jc: with help from Thomas Dickey in archaeology of "diffstat"]
[jc: squashed Jonathan's updates to illustrations in tutorials and a test]
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint-1.7.7:
Git 1.7.7.5
Git 1.7.6.5
blame: don't overflow time buffer
fetch: create status table using strbuf
checkout,merge: loosen overwriting untracked file check based on info/exclude
cast variable in call to free() in builtin/diff.c and submodule.c
apply: get rid of useless x < 0 comparison on a size_t type
Conflicts:
Documentation/git.txt
GIT-VERSION-GEN
RelNotes
builtin/fetch.c
When git apply is passed something that is not a patch, it does not produce
an error message or exit with a non-zero status if it was not actually
"applying" the patch i.e. --check or --numstat etc were supplied on the
command line.
Fix this by producing an error when apply fails to find any hunks whatsoever
while parsing the patch.
This will cause some of the output formats (--numstat, --diffstat, etc) to
produce an error when they formerly would have reported zero changes and
exited successfully. That seems like the correct behavior though. Failure
to recognize the input as a patch should be an error.
Plus, add a test.
Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
According to the C standard size_t is always unsigned, therefore the
comparison "n1 < 0 || n2 < 0" when n1 and n2 are size_t will always be
false.
This was raised by clang 2.9 which throws this warning when compiling
apply.c:
builtin/apply.c:253:9: warning: comparison of unsigned expression < 0 is always false [-Wtautological-compare]
if (n1 < 0 || n2 < 0)
~~ ^ ~
builtin/apply.c:253:19: warning: comparison of unsigned expression < 0 is always false [-Wtautological-compare]
if (n1 < 0 || n2 < 0)
~~ ^ ~
This check was originally added in v1.6.5-rc0~53^2 by Giuseppe Bilotta
while adding an option to git-apply to ignore whitespace differences.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I noticed this when "git am CORRUPTED" unexpectedly failed with an
odd diagnostic, and even removed one of the files it was supposed
to have patched.
Reproduce with any valid old/new patch from which you have removed
the "+++ b/FILE" line. You'll see a diagnostic like this
fatal: unable to write file '(null)' mode 100644: Bad address
and you'll find that FILE has been removed.
The above is on glibc-based systems. On other systems, rather than
getting "null", you may provoke a segfault as git tries to
dereference the NULL file name.
Signed-off-by: Jim Meyering <meyering@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The size of objects we read from the repository and data we try to put
into the repository are represented in "unsigned long", so that on larger
architectures we can handle objects that weigh more than 4GB.
But the interface defined in zlib.h to communicate with inflate/deflate
limits avail_in (how many bytes of input are we calling zlib with) and
avail_out (how many bytes of output from zlib are we ready to accept)
fields effectively to 4GB by defining their type to be uInt.
In many places in our code, we allocate a large buffer (e.g. mmap'ing a
large loose object file) and tell zlib its size by assigning the size to
avail_in field of the stream, but that will truncate the high octets of
the real size. The worst part of this story is that we often pass around
z_stream (the state object used by zlib) to keep track of the number of
used bytes in input/output buffer by inspecting these two fields, which
practically limits our callchain to the same 4GB limit.
Wrap z_stream in another structure git_zstream that can express avail_in
and avail_out in unsigned long. For now, just die() when the caller gives
a size that cannot be given to a single zlib call. In later patches in the
series, we would make git_inflate() and git_deflate() internally loop to
give callers an illusion that our "improved" version of zlib interface can
operate on a buffer larger than 4GB in one go.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jc/maint-add-p-overlapping-hunks:
t3701: add-p-fix makes the last test to pass
"add -p": work-around an old laziness that does not coalesce hunks
add--interactive.perl: factor out repeated --recount option
t3701: Editing a split hunk in an "add -p" session
add -p: 'q' should really quit
Since 0beee4c (git-add--interactive: remove hunk coalescing, 2008-07-02),
"git add--interactive" behaves lazily and passes overlapping hunks to the
underlying "git apply" without coalescing. This was partially corrected
by 7a26e65 (its partial revert, 2009-05-16), but overlapping hunks are
still passed when the patch is edited.
Teach "git apply" the --allow-overlap option that disables a safety
feature that avoids misapplication of patches by not applying patches
to overlapping hunks, and pass this option form "add -p" codepath.
Do not even advertise the option, as this is merely a workaround, and the
correct fix is to make "add -p" correctly coalesce adjacent patch hunks.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the line number the patch intended to touch does not match
the line in the version being patched, GNU patch reports that
it applied the hunk at a different line number, with how big an
offset.
Teach "git apply" to do the same under --verbose option.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When looking for a place to apply a hunk, we used to check lines that
match the preimage of it, starting from the line that the patch wants to
apply the hunk at, looking forward and backward with increasing offsets
until we find a match.
Colin Guthrie found an interesting case where this misapplied a patch that
wanted to touch a preimage that consists of
}
}
return 0;
}
which is a rather unfortunately common pattern.
The target version of the file originally had only one such location, but
the hunk immediately before that created another instance of such block of
lines, and find_pos() happily reported that the preimage of the hunk
matched what it wanted to modify.
Oops.
By marking the lines application of earlier hunks touched and preventing
match_fragment() from considering them as a match with preimage of other
hunks, we can reduce such an accident.
I also considered to teach apply_one_fragment() to take the offset we have
found while applying the previous hunk into account when looking for a
match with find_pos(), but dismissed that approach, because it would
sometimes work better but sometimes worse, depending on the difference
between the version the patch was created against and the version the
patch is being applied.
This does _not_ prevent misapplication of patches to a file that has many
similar looking blocks of lines and a preimage cannot identify which one
of them should be applied. For that, we would need to scan beyond the
first match in find_pos(), and issue a warning (or error out). That will
be a separate topic.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allows better help text to be defined than "be verbose". Also make use
of the macro in places that already had a different description. No
object code changes intended.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some patches have a timezone formatted like '-08:00' instead of
'-0800' in their ---/+++ lines (e.g. http://lwn.net/Articles/131729/).
Take this into account when searching for the start of the timezone
(which is the end of the filename).
This does not actually affect the outcome of patching unless (1) a
file being patched has a non-' ' whitespace character (e.g., tab) in
its filename, or (2) the patch is whitespace-damaged, so the tab
between filename and timestamp has been replaced with spaces.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix the case when the patch is a rename or mode-change only
and -p is used with a value greater than one.
The git_header_name function did not remove more than one path
component.
Signed-off-by: Federico Cuello <fedux@lugmen.org.ar>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Usually when applying a binary diff generated without
--binary, it will be rejected early, as we don't even have
the full sha1 of the pre- and post-images.
However, if the diff is generated with --full-index (but not
--binary), then we will actually try to apply it. If we have
the postimage blob, then we can take a shortcut and never
even look at the binary diff at all (e.g., this can happen
when rebasing changes within a repository).
If we don't have the postimage blob, though, we try to look
at the actual fragments, of which there are none, and get a
segfault. This patch checks explicitly for that case and
complains to the user instead of segfaulting. We need to
keep the check at a low level so that the "shortcut" case
above continues to work.
We also add a test that demonstrates the segfault. While
we're at it, let's also explicitly test the shortcut case.
Reported-by: Rafaël Carré <rafael.carre@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>