An earlier commit cbd64af added a check that prevents "git-am"
to run without its standard input connected to a terminal while
resuming operation. This was to catch a user error to try
feeding a new patch from its standard input while recovery.
The assumption of the check was that it is an indication that a
new patch is being fed if the standard input is not connected to
a terminal. It is however not quite correct (the standard input
can be /dev/null if the user knows the operation does not need
any input, for example). This broke t3403 when the test was run
with its standard input connected to /dev/null.
When git-am is given an explicit command such as --skip, there
is no reason to insist that the standard input is a terminal; we
are not going to read a new patch anyway.
Credit goes to Gerrit Pape for noticing and reporting the
problem with t3403-rebase-skip test.
Fetch over http from a repository that uses alternates to borrow
from neighbouring repositories were quite broken, apparently for
some time now.
We parse input and count bytes to allocate the new buffer, and
when we copy into that buffer we know exactly how many bytes we
want to copy from where. Using strlcpy for it was simply
stupid, and the code forgot to take it into account that strlcpy
terminated the string with NUL.
All but one callers have ignore the return value from this
function, but the only caller, builtin-tar-tree.c::remote_tar(),
assumed it returns non-zero on failure and zero on success. The
implementation however was returning either the waited pid
(which must be the same as its input) or -1 (an error).
Fix this thinko, while getting rid of an assignment of return
value from waitpid() into a variable of type int.
sample commit-msg hook: no silent exit on duplicate Signed-off-by lines
git-commit would silently exit if duplicate Signed-off-by
lines were found. Users of git-commit would not know it,
unless they checked '$?'. This patch makes git-commit
actually print out a message that nothing was commited
since duplicate Signed-off-lines were found.
Signed-off-by: Luben Tuikov <ltuikov@yahoo.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
Better error message when we are unable to lock the index file
Most of the callers except the one in refs.c use the function to
update the index file. Among the index writers, everybody
except write-tree dies if they cannot open it for writing.
This gives the function an extra argument, to tell it to die
when it cannot create a new file as the lockfile.
The only caller that does not have to die is write-tree, because
updating the index for the cache-tree part is optional and not
being able to do so does not affect the correctness. I think we
do not have to be so careful and make the failure into die() the
same way as other callers, but that would be a different patch.
git-am: give better diagnostics when the patch does not apply during --3way
If the user tries to apply a patch that was hand-edited in such
a way that it does not apply to the original file recorded on
its "index" line anymore, we did detect the situation but did
not issue an error message that is specific enough.
git-svn: split the path from the url correctly with limited perms
This version of the splitter (that only affects SVN:: library
users) works when one only has limited read-permissions to
the repository they're fetching from.
Updated from the original patch to workaround some SVN bug
somewhere, which only seems to happen against file://
repositories... Here's the diff against the original patch I
submitted:
@@ -1159,8 +1159,8 @@ sub repo_path_split {
}
if ($_use_lib) {
- $SVN = libsvn_connect($full_url);
- my $url = $SVN->get_repos_root;
+ my $tmp = libsvn_connect($full_url);
+ my $url = $tmp->get_repos_root;
$full_url =~ s#^\Q$url\E/*##;
push @repo_path_split_cache, qr/^(\Q$url\E)/;
return ($url, $full_url);
Somehow connecting to a repository with the full url makes the
returned SVN::Ra object act strangely and break things, so now
we just drop the SVN::Ra object that we made our initial
connection with.
Signed-off-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-sh-setup: do not use repo-config to test the git directory
Since repo-config does not fail in non-git directory, it is not
a good command to use to test the git-ness nor validate the
repository revision of $GIT_DIR.
Original patch by Robert Shearman but with minor fixes.
git-svn: bugfix: allow SVN:: lib users to track the root of the repository
I'm not sure if anybody has hit this (besides me), but this
fixes the problem where I ran into while attempting to import a
small repo at the root level: I ended up with all the commits, but
with no file/tree changes at all throughout the entire history.
Also, fix a warning if the commit message is not defined for revision 0.
Signed-off-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-svn: correctly kill keyword expansion without munging EOLs
This bugfix applies to users of the svn command-line client only.
We no longer muck with newlines when killing keyword expansion.
This tended to generate unintended diffs in commits because svn
revert -R would destroy the manual EOL changes we were doing. Of
course, we didn't need the EOL munging in the first place, as
svn seems to do it for us even in the text-base files.
Now we set the mtime and atime the files changed by keyword
expansion killing to avoid triggering a change on svn revert,
which svn still seems to want to do.
Thanks to Seth Falcon for reporting this bug.
Signed-off-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <junkio@cox.net>
As Fredrik points out the current interface of has_extension() is
potentially confusing. Its parameters include both a nul-terminated
string and a length-limited string.
This patch drops the length argument, requiring two nul-terminated
strings; all callsites are updated. I checked that all of them indeed
provide nul-terminated strings. Filenames need to be nul-terminated
anyway if they are to be passed to open() etc. The performance penalty
of the additional strlen() is negligible compared to the system calls
which inevitably surround has_extension() calls.
Additionally, change has_extension() to use size_t inside instead of
int, as that is the exact type strlen() returns and memcmp() expects.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
A small howto on how to setup GIT over HTTP transport protocol by
setting up WebDAV access on apache2.
[jc: minimum ispell fixes applied]
Signed-off-by: Rutger Nijlunsing <git@tux.tmfweb.nl> Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-verify-pack: free pack after use and a cleanup
Plug memory leak in verify_one_pack() by freeing the struct packed_git
we got from add_packed_git(). Also rename g to pack and pull an
assignment out of an if statement while we're at it.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
Get rid of that while loop which was apparently used as a way to avoid
goto's (why?). It's easy now because there is only one break left at
the end of it. Also make the comment clearer.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-verify-pack can be called with a filename without .idx extension.
add_packed_git() on the other hand depends on its presence. So
instead of trying to call it with whatever the user gave us check for
that extension and add it if it's missing.
That means that you can't name your index file "blah" and your pack
file ".pack" anymore ("git-verify-pack blah" currently works in that
case). I think this regression is a good change. ;-)
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
Use strlcpy() to copy the filename into a buffer and complain if it
doesn't fit. Also move the path buffer into verify_one_pack(); it is
used only there. Now we can const'ify the first argument of this
function.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
Commit 9919f41 meant to make git-diff A...B to (usually) mean
"git-diff `git-merge-base A B` B", but it got the parameters wrong
and ended up showing "git-diff `git-merge-base A B` A" by mistake.
http-push: Make WebDAV work with (broken?) default apache2 WebDAV module
WebDAV on Debian unstable cannot handle renames on WebDAV from
file.ext to newfile (without ext) when newfile* already
exists. Normally, git creates a file like 'objects/xx/sha1.token',
which is renamed to 'objects/xx/sha1' when transferred completely.
Just use '_' instead of '.' so WebDAV doesn't see it as an extension
change.
Signed-off-by: Rutger Nijlunsing <git@tux.tmfweb.nl> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-apply: applying a patch to make a symlink shorter.
The internal representation of the result is counted string
(i.e. char *buf and ulong size), which is fine for writing out
to regular file, but throwing the buf at symlink(2) was a
no-no.
When an ancient "git" that does not understand "describe"
command is on the $PATH, "git describe" emitted a Usage message
without exiting non-zero status (which is a mistake we cannot
fix retroactively). Catch this case to make sure we do not try
using phoney multi-line string as a version number.
We used to find the first match of the pattern and then if the
match is not for the entire word, declared that the whole line
does not match.
But that is wrong. The command "git grep -w -e mmap" should
find that a line "foo_mmap bar mmap baz" matches, by tring the
second instance of pattern "mmap" on the same line.
Problems an earlier round of "fix" had were pointed out by Morten
Welinder, which have been incorporated in the t7002 tests.
commit walkers: setup_ident() to record correct committer in ref-log.
The function pull() in fetch.c calls write_ref_sha1(), which may
need committer identity to update the ref-log, so they need to
call setup_ident() before calling git_config() function.
Acked-by: Shawn Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Allow config file to specify Signed-off-by identity in format-patch.
Unlike git-commit, git-format-patch was not picking up and using the
user.email config variable for the email part of the committer info.
I was forced to use the GIT_COMMITTER_EMAIL environment variable to
override the default <user@localhost.localdomain>. The fix was to
simply move the call to setup_ident() to come before the git_config()
call.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
Free the root tree object buffer when we're done, plugging a minor leak
in generate_tar(). Note: we cannot simply free(tree.buf) because this
pointer is modified by tree_entry() calls in traverse_tree().
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
Most usage strings, such as for command xxx, start with "git-xxx".
This updates the rebels to conform to the general pattern.
(The git wrapper is an exception to this, of course ...)
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
We ended up merging too many stuff after -rc2, so here is
another round of release candidate. Non bugfixes will be
queued to "next" from now on until a real 1.4.2 happens.
Earlier we deliberately showed only blob1's name because showing
both names labeled the output as a renaming patch. Now the output
routine (namely, diff.c::resolve_rename_copy()) is taught not to
use pathname comparison to tell if a filepair is a rename, we can
safely do this change.
diff.c: do not use pathname comparison to tell renames
The final output from diff used to compare pathnames between
preimage and postimage to tell if the filepair is a rename/copy.
By explicitly marking the filepair created by diffcore_rename(),
the output routine, resolve_rename_copy(), does not have to do
so anymore. This helps feeding a filepair that has different
pathnames in one and two elements to the diff machinery (most
notably, comparing two blobs).
We still have too few of them, but we have to start from somewhere.
The general rule is to make tests easy to debug when run with -v (notice
use of seemingly useless echo everywhere in the new tests).
This was already documented in the options section of the manpage. This
patch implements it, adds it to the usage message, and mentions it at the
top of the manpage.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <junkio@cox.net>
convert-objects.c sets _XOPEN_SOURCE and _XOPEN_SOURCE_EXTENDED before
including <time.h>, in order to get the declaration of strptime().
This leads to breakage in cache.h, due to S_ISLNK and S_IFLNK no longer
being defined by <sys/stat.h>. These definitions are protected by the
__USE_BSD symbol, which is not set when _XOPEN_SOURCE is set. Moving
the #defines and #include <time.h> below all other #includes does not
fix the problem, however, since now _USE_XOPEN, which protects the
declaration of strptime(), is now not defined (don't ask!).
The fix is to #define _GNU_SOURCE, which enables the definition of
practically everything.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
Remove cmd_usage() routine and re-organize the help/usage code.
The cmd_usage() routine was causing warning messages due to a NULL
format parameter being passed in three out of four calls. This is a
problem if you want to compile with -Werror. A simple solution is to
simply remove the GNU __attribute__ format pragma from the cmd_usage()
declaration in the header file. The function interface was somewhat
muddled anyway, so re-write the code to finesse the problem.
[jc: this incidentally revealed that t9100 test assumed that the output
from "git help" to be fixed in stone, but this patch lower-cases
"Usage" to "usage". Update the test not to rely on "git help" output.]
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
The header builtin.h was, incorrectly, redefining PATH_MAX which
causes a header order dependency in builtin-write-tree.c. The fix
is to simply include <limits.h> directly to obtain the correct
definition of PATH_MAX.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
The NO_C99_FORMAT macro allows compilers that lack support for the
ll,hh,j,z,t size specifiers (eg. gcc 2.95.2) to adapt the code to avoid
runtime errors in the formatted IO functions.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
These changes were originally part of the next patch, but have been
split out since they were peripheral to the main purpose of that patch.
- update comment describing the signature format to reflect
the current code.
- remove trailing \n in calls to error(), since a \n is already
provided by error().
- remove redundant call to get_sha1_hex().
- call sha1_to_hex(sha1) to convert to ascii, rather than attempting
to print the raw sha1.
The new tests provide a regression suite to support the modifications
to git-mktag in this and the next patch.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
Fix annotate test script; notice when git-annotate fails.
The t8001-annotate.sh test claimed all tests pass, when in fact
the git-annotate perl script failed to run! (prior to fixing the
script to work with perl 5.5).
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
Ensure git-clone exits with error if perl script fails.
This helps tests 5400,5600,5700,5710 "fail correctly" rather than
give some false positives. Also ensure cleanup actions in exit trap
work correctly even if user has alias rm='rm -i'.
Signed-off-by: Ramsay Allan Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
fetch/clone: mark messages from remote side stand out.
When dealing with a corrupt or out of sync remote repository,
the user often gets error messages like this:
error: refs/heads/devel does not point to a valid commit object!
which leaves the user wondering if the breakage is on the local
end or on the remote end. This is unnecessarily alarming.
This patch changes the way we display messages received from the
remote side over the git protocol sideband (i.e. stderr stream
of the remote process). It shows them with blue background with
white letters, but this presentation is subject to proposals of
better ways from the list.
* rs/rebase:
rebase: Make the fast-fowarding message more user-friendly by using branch names instead of SHA1 IDs.
rebase: Fix the detection of fast-forwarding of the current branch to upstream.
Builtins: control the use of pager from the command table.
This moves the built-in "always-use-pager" logic for log family
to the command dispatch table of git wrapper. This makes it
easier to change the default use of pager, and has an added
benefit that we fork and exec the pager early before packs are
mmapped.
Pointed out by Juergen Ruehle <j.ruehle@bmiag.de>.
Fix up some fallout from "setup_git_directory()" cleanups
git-ls-files was broken by the setup_git_directory() calling changes,
because I had missed the fact that the "prefix" variable in that file was
static to the whole file, and unlike git-ls-tree (where I had fixed it
up), it ended up using two different variables with the same name
depending on what the scoping happened to be.
This fixes it up properly (by just removing the static variable, and
passing the automatic one around properly), and git-ls-files should work
again.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Doing an "strace" on "git diff" shows that we close() a file descriptor
twice (getting EBADFD on the second one) when we end up in ce_compare_data
if the index does not match the checked-out stat information.
The "index_fd()" function will already have closed the fd for us, so we
should not close it again.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>