This commit changes the rules for resolving refspecs to match the
rules for resolving refs in rev-parse. git-rev-parse uses clear rules
to resolve a short ref to its full name, which are well documented.
The rules for resolving refspecs documented in git-send-pack were
less strict and harder to understand. This commit replaces them by
the rules of git-rev-parse.
The unified rules are easier to understand and better resolve ambiguous
cases. You can now push from a repository containing several branches
ending on the same short name.
Note, this may break existing setups. For example, "master" will no longer
resolve to "origin/master" even when there is no other "master" elsewhere.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We use at least two rulesets for matching abbreviated refnames with
full refnames (starting with 'refs/'). git-rev-parse and git-fetch
use slightly different rules.
This commit introduces a new function refname_match
(const char *abbrev_name, const char *full_name, const char **rules).
abbrev_name is expanded using the rules and matched against full_name.
If a match is found the function returns true. rules is a NULL-terminate
list of format patterns with "%.*s", for example:
const char *ref_rev_parse_rules[] = {
"%.*s",
"refs/%.*s",
"refs/tags/%.*s",
"refs/heads/%.*s",
"refs/remotes/%.*s",
"refs/remotes/%.*s/HEAD",
NULL
};
Asterisks are included in the format strings because this is the form
required in sha1_name.c. Sharing the list with the functions there is
a good idea to avoid duplicating the rules. Hopefully this
facilitates unified matching rules in the future.
This commit makes the rules used by rev-parse for resolving refs to
sha1s available for string comparison. Before this change, the rules
were buried in get_sha1*() and dwim_ref().
A follow-up commit will refactor the rules used by fetch.
refname_match() will be used for matching refspecs in git-send-pack.
Thanks to Daniel Barkalow <barkalow@iabervon.org> for pointing
out that ref_matches_abbrev in remote.c solves a similar problem
and care should be taken to avoid confusion.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This teaches "push <remote> HEAD" to resolve HEAD on the local
side to its real branch name, e.g. master, and then act as if
the real branch name was specified. So we have a shorthand for
pushing the current branch. Besides HEAD, no other symbolic ref
is resolved.
Thanks to Daniel Barkalow <barkalow@iabervon.org> for suggesting
this implementation, which is much simpler than the
implementation proposed before.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
... by changing git-tar-tree reference to git-archive and removing
seemingly unrelevant footnote about git-ssh-{fetch,upload}.
Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Previously we tested if the "$GIT_DIR/refs/bisect" directory
existed, to check if we were bisecting.
Now with packed refs, it is simpler to check if the file
"$GIT_DIR/BISECT_NAMES" exists, as it is already created when
starting bisection and removed when reseting bisection.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We should mention the use of the "..." syntax for git-diff here. The
note about the difference between diff and the combined output of
git-format-patch then no longer fits so well, so remove it. Add a
reference to the git-format-patch[1] manpage.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
* lt/rev-list-interactive:
Fix parent rewriting in --early-output
revision walker: mini clean-up
Enhance --early-output format
Add "--early-output" log flag for interactive GUI use
Simplify topo-sort logic
* ph/diffopts:
Reorder diff_opt_parse options more logically per topics.
Make the diff_options bitfields be an unsigned with explicit masks.
Use OPT_BIT in builtin-pack-refs
Use OPT_BIT in builtin-for-each-ref
Use OPT_SET_INT and OPT_BIT in builtin-branch
parse-options new features.
This commit adds a discussion of the challenge of bisecting
merge commits to the user manual. The original author is
Junio C Hamano <gitster@pobox.com>, who posted the text to
the mailing list <http://marc.info/?l=git&m=119403257315527&w=2>.
His email was adapted for the manual.
The discussion is added to "Rewriting history and maintainig
patch series". The text added requires good understanding of
merging and rebasing. Therefore it should not be placed too
early in the manual. Right after the section on "Problems with
rewriting history", the discussion of bisect gives another reason
for linearizing as much of the history as possible.
The text includes suggestions and fixes by
Ralf Wildenhues <Ralf.Wildenhues@gmx.de> and
Benoit Sigoure <tsuna@lrde.epita.fr>.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Define NO_MKDTEMP for all variants of SunOS; Solaris 10 does not
have mkdtemp() and all the other versions our Makefile knows
about don't have it either.
NO_{SETENV,UNSETENV,C99_FORMAT,STRTOUMAX} definitions cannot be
unified across versions. Beginning with Solaris 10, the C-library
provides unsetenv(), setenv() and strtoumax(). Also 'z'/'t' formats
are supported. However, older versions of Solaris do not support
these.
Signed-off-by: Guido Ostkamp <git@ostkamp.fastmail.fm>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This has always been slightly inaccurate, since it used the
new_refs counter, which really meant "did we send any
objects," so deletions were not counted.
It has gotten even worse with recent patches, since we no
longer look at the 'ret' value, meaning we would say "up to
date" if non-ff pushes were rejected.
Instead, we now claim up to date iff every ref is either
unmatched or up to date. Any other case should already have
generated a status line.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Previously, we set all ref pushes to 'OK', and then marked
them as errors if the remote reported so. This has the
problem that if the remote dies or fails to report a ref, we
just assume it was OK.
Instead, we use a new non-OK state to indicate that we are
expecting status (if the remote doesn't support the
report-status feature, we fall back on the old behavior).
Thus we can flag refs for which we expected a status, but
got none (conversely, we now also print a warning for refs
for which we get a status, but weren't expecting one).
This also allows us to simplify the receive_status exit
code, since each ref is individually marked with failure
until we get a success response. We can just print the usual
status table, so the user still gets a sense of what we were
trying to do when the failure happened.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
cache.h:503: warning: type of bit-field 'force' is a GCC extension
cache.h:504: warning: type of bit-field 'merge' is a GCC extension
cache.h:505: warning: type of bit-field 'nonfastforward' is a GCC extension
cache.h:506: warning: type of bit-field 'deletion' is a GCC extension
So we change it to an 'unsigned int' which is not a GCC extension.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* ph/parseopt-sh:
git-quiltimport.sh fix --patches handling
git-am: -i does not take a string parameter.
sh-setup: don't let eval output to be shell-expanded.
git-sh-setup: fix parseopt `eval` string underquoting
Give git-am back the ability to add Signed-off-by lines.
git-rev-parse --parseopt
scripts: Add placeholders for OPTIONS_SPEC
Migrate git-repack.sh to use git-rev-parse --parseopt
Migrate git-quiltimport.sh to use git-rev-parse --parseopt
Migrate git-checkout.sh to use git-rev-parse --parseopt --keep-dashdash
Migrate git-instaweb.sh to use git-rev-parse --parseopt
Migrate git-merge.sh to use git-rev-parse --parseopt
Migrate git-am.sh to use git-rev-parse --parseopt
Migrate git-clone to use git-rev-parse --parseopt
Migrate git-clean.sh to use git-rev-parse --parseopt.
Update git-sh-setup(1) to allow transparent use of git-rev-parse --parseopt
Add a parseopt mode to git-rev-parse to bring parse-options to shell scripts.
When building command line to invoke external grep, the
arguments to -A/-B/-C options were placd in randarg[] buffer,
but the code forgot that snprintf() does not count terminating
NUL in its return value. This caused "git grep -A1 -B2" to
invoke external grep with "-B21 -A1".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
get_pack() receives a pair of file descriptors that communicate with
upload-pack at the remote end. In order to support the case where the
side-band demultiplexer runs in a thread, and, hence, in the same process
as the main routine, we must not close the readable file descriptor early.
The handling of the readable fd is changed in the case where upload-pack
supports side-band communication: The old code closed the fd after it was
inherited to the side-band demultiplexer process. Now we do not close it.
The caller (do_fetch_pack) will close it later anyway. The demultiplexer
is the only reader, it does not matter that the fd remains open in the
main process as well as in unpack-objects/index-pack, which inherits it.
The writable fd is not needed in get_pack(), hence, the old code closed
the fd. For symmetry with the readable fd, we now do not close it; the
caller (do_fetch_pack) will close it later anyway. Therefore, the new
behavior is that the channel now remains open during the entire
conversation, but this has no ill effects because upload-pack does not read
from it once it has begun to send the pack data. For the same reason it
does not matter that the writable fd is now inherited to the demultiplexer
and unpack-objects/index-pack processes.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://git.bogomips.org/git-svn:
git-svn: Fix a typo and add a comma in an error message in git-svn
git-svn log: handle unreachable revisions like "svn log"
git-svn log: include commit log for the smallest revision in a range
git-svn log: fix ascending revision ranges
git-svn's dcommit must use subversion's config
git-svn: add tests for command-line usage of init and clone commands
When unreachable revisions are given to "svn log", it displays all commit
logs in the given range that exist in the current tree. (If no commit
logs are found in the current tree, it simply prints a single commit log
separator.) This patch makes "git-svn log" behave the same way.
Ten tests added to t/t9116-git-svn-log.sh.
Signed-off-by: David D Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
The "svn log -rM:N" command shows commit logs inclusive in the range [M,N].
Previously "git-svn log" always excluded the commit log for the smallest
revision in a range, whether the range was ascending or descending. With
this patch, the smallest revision in a range is always shown.
Updated tests for ascending and descending revision ranges.
Signed-off-by: David D Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
Fixed typo in Git::SVN::Log::git_svn_log_cmd(). Previously a command like
"git-svn log -r1:4" would only show a commit log separator.
Added tests for ascending and descending revision ranges.
Signed-off-by: David D Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
When doing dcommit git-svn must use subversion's config or newly created
files will not include svn's properties
(defined in [auto-props] with 'enable-auto-props = yes').
Signed-off-by: Konstantin V. Arkhipov <voxus@onphp.org>
Acked-by: Eric Wong <normalperson@yhbt.net>
This lets us show remote errors (e.g., a denied hook) along
with the usual push output.
There is a slightly clever optimization in receive_status
that bears explanation. We need to correlate the returned
status and our ref objects, which naively could be an O(m*n)
operation. However, since the current implementation of
receive-pack returns the errors to us in the same order that
we sent them, we optimistically look for the next ref to be
looked up to come after the last one we have found. So it
should be an O(m+n) merge if the receive-pack behavior
holds, but we fall back to a correct but slower behavior if
it should change.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Previously, we manually checked the 'NONE' and 'UPTODATE'
conditions. Now that we have ref->status, we can easily
say "only update if we pushed successfully".
This adds a test for and fixes a regression introduced in
ed31df31 where deleted refs did not have their tracking
branches removed. This was due to a bogus per-ref error test
that is superseded by the more accurate ref->status flag.
Signed-off-by: Jeff King <peff@peff.net>
Completely-Acked-By: Alex "Sleepy" Riesen <raa.lkml@gmail.com>
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of keeping the 'ret' variable, we instead have a
status flag for each ref that tracks what happened to it.
We then print the ref status after all of the refs have
been examined.
This paves the way for three improvements:
- updating tracking refs only for non-error refs
- incorporating remote rejection into the printed status
- printing errors in a different order than we processed
(e.g., consolidating non-ff errors near the end with
a special message)
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Alex Riesen <raa.lkml@gmail.com>
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In bisect_visualize, "cd $GIT_DIR/refs && echo bisect/good-*" was
still used instead of "git for-each-ref". This patch fix it.
We now pass "refs/bisect/bad" and "--not refs/bisect/good-<rev>"
instead of "bisect/bad" and "--not bisect/good-<rev>" to gitk,
but it seems to work.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Brown paper bag fix to avoid using non portable sed syntax. The
test by itself didn't catch what it was supposed to, anyways.
The new test first checks if git-tag correctly errors out when
the user exited the editor without editing the file. Then it
checks if what the user was presented in the editor was any
useful, which we define as the following:
* It begins with a single blank line, where the invoked editor
would typically place the editing curser at, so that the user
can immediately start typing;
* It has some instruction but that comes after that initial
blank line, all lines prefixed with "#". We specifically do
not check for the wording of this instruction.
* And it has nothing else, as the expected behaviour is "Hey
you did not leave any message".
Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function mark_tree_uninteresting() assumed that the tree entries
are blob when they are not trees. This is not so. Since we do
not traverse into submodules (yet), the gitlinks should be ignored.
In general, we should try to start moving away from using the
"S_ISLNK()" like things for internal git state. It was a mistake to
just assume the numbers all were same across all systems in the first
place. This implementation converts to the "object_type", and then
uses a case statement.
Noticed by Ilari on IRC.
Test script taken from an earlier version by Dscho.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In "dir_struct", each exclusion element in the exclusion stack records a
base string (pointer to the beginning with length) so that we can tell
where it came from, but this pointer is just pointing at the parameter
that is given by the caller to the push_exclude_per_directory()
function.
While read_directory_recursive() runs, calls to excluded() makes use
the data in the exclusion elements, including this base string. The
caller of read_directory_recursive() is not supposed to free the
buffer it gave to push_exclude_per_directory() earlier, until it
returns.
The test case Bruce Stephens gave in the mailing list discussion
was simplified and added to the t3700 test.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are inconsistencies in the way commands currently handle
the core.excludesfile configuration variable. The problem is
the variable is too new to be noticed by anything other than
git-add and git-status.
* git-ls-files does not notice any of the "ignore" files by
default, as it predates the standardized set of ignore files.
The calling scripts established the convention to use
.git/info/exclude, .gitignore, and later core.excludesfile.
* git-add and git-status know about it because they call
add_excludes_from_file() directly with their own notion of
which standard set of ignore files to use. This is just a
stupid duplication of code that need to be updated every time
the definition of the standard set of ignore files is
changed.
* git-read-tree takes --exclude-per-directory=<gitignore>,
not because the flexibility was needed. Again, this was
because the option predates the standardization of the ignore
files.
* git-merge-recursive uses hardcoded per-directory .gitignore
and nothing else. git-clean (scripted version) does not
honor core.* because its call to underlying ls-files does not
know about it. git-clean in C (parked in 'pu') doesn't either.
We probably could change git-ls-files to use the standard set
when no excludes are specified on the command line and ignore
processing was asked, or something like that, but that will be a
change in semantics and might break people's scripts in a subtle
way. I am somewhat reluctant to make such a change.
On the other hand, I think it makes perfect sense to fix
git-read-tree, git-merge-recursive and git-clean to follow the
same rule as other commands. I do not think of a valid use case
to give an exclude-per-directory that is nonstandard to
read-tree command, outside a "negative" test in the t1004 test
script.
This patch is the first step to untangle this mess.
The next step would be to teach read-tree, merge-recursive and
clean (in C) to use setup_standard_excludes().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a user has an "auto-prop" in his/her ~/.subversion/config file for
automatically setting the svn:keyword Id property on all ".c" files
(a reasonably common configuration in the Subversion world) then one
of the "svn propset" operations in the very first test would become a
no-op, which in turn would make the next commit a no-op.
This then caused the 25th test ('test propget') to fail because it
expects a certain number of commits to have taken place but the actual
number of commits was off by one.
Björn Steinbrink identified the "auto-prop" feature as the cause
of the failure. This patch avoids it by passing the "--no-auto-prop"
flag to "svn import" when setting up the test repository, thus ensuring
that the "svn propset" operation is no longer a no-op, regardless of the
users' settings in their config.
Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We sometimes pick out the original rfc822 'From' header and
include it in the body of the message. If the original
author's name needs encoding, then we should specify that in
the content-type header.
If we already had a content-type header in the mail, then we
may need to re-encode. The logic is there to detect
this case, but it doesn't actually do the re-encoding.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Uwe Kleine-König <Uwe.Kleine-Koenig@digi.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When switching to a new rev, we first made "new-bisect" branch to
point at the chosen commit, attempt to switch to it, and then
finally renamed the new-bisect branch to bisect by hand when
successful. This is so that we can catch checkout failure (your
local modification may interfere with switching to the chosen
version) without losing information on which commit the next
attempt should be made.
Rewrite it using a more modern form but without breaking the
safety.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This removes the last instance of making a ref by hand with
"echo SHA1 >.git/refs/$refname" from the script and replaces it
with update-ref.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
