The motivation for this change is that system call failures are serious
errors that should be reported to the user, but only few callers took the
burden to decode the error codes that the functions returned into error
messages.
If at all, then only an unspecific error message was given. A prominent
example is this:
$ git upload-pack . | :
fatal: unable to run 'git-upload-pack'
In this example, git-upload-pack, the external command invoked through the
git wrapper, dies due to SIGPIPE, but the git wrapper does not bother to
report the real cause. In fact, this very error message is copied to the
syslog if git-daemon's client aborts the connection early.
With this change, system call failures are reported immediately after the
failure and only a generic failure code is returned to the caller. In the
above example the error is now to the point:
$ git upload-pack . | :
error: git-upload-pack died of signal
Note that there is no error report if the invoked program terminated with
a non-zero exit code, because it is reasonable to expect that the invoked
program has already reported an error. (But many run_command call sites
nevertheless write a generic error message.)
There was one special return code that was used to identify the case where
run_command failed because the requested program could not be exec'd. This
special case is now treated like a system call failure with errno set to
ENOENT. No error is reported in this case, because the call site in git.c
expects this as a normal result. Therefore, the callers that carefully
decoded the return value still check for this condition.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As a general guideline, functions in git's code return zero to indicate
success and negative values to indicate failure. The run_command family of
functions followed this guideline. But there are actually two different
kinds of failure:
- failures of system calls;
- non-zero exit code of the program that was run.
Usually, a non-zero exit code of the program is a failure and means a
failure to the caller. Except that sometimes it does not. For example, the
exit code of merge programs (e.g. external merge drivers) conveys
information about how the merge failed, and not all exit calls are
actually failures.
Furthermore, the return value of run_command is sometimes used as exit
code by the caller.
This change arranges that the exit code of the program is returned as a
positive value, which can now be regarded as the "result" of the function.
System call failures continue to be reported as negative values.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Windows does not have signals. At least they cannot be diagnosed by the
parent process; all that the parent process can observe is the exit code.
This also adds a dummy definition of WTERMSIG.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
For some reason, MinGW's bash cannot reliably detect failure of the child
process if a negative value is passed to exit(). This fixes it by
truncating the exit code in all calls of exit().
This issue was worked around in run_builtin() of git.c (2488df84 builtin
run_command: do not exit with -1, 2007-11-15). This workaround is no longer
necessary and is reverted.
Suggested-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Put filenames into the conflict markers only when they are different.
Otherwise they are redundant information clutter.
Print the filename explicitely when warning about a binary conflict.
Signed-off-by: Martin Renold <martinxyz@gmx.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* cc/bisect:
Documentation: remove warning saying that "git bisect skip" may slow bisection
bisect: use a PRNG with a bias when skipping away from untestable commits
* js/daemon-log:
receive-pack: do not send error details to the client
upload-pack: squelch progress indicator if client cannot see it
daemon: send stderr of service programs to the syslog
* sb/quiet-porcelains:
stash: teach quiet option
am, rebase: teach quiet option
submodule, repack: migrate to git-sh-setup's say()
git-sh-setup: introduce say() for quiet options
am: suppress apply errors when using 3-way
t4150: test applying with a newline in subject
Earlier 476cc72 (request-pull: really disable pager, 2009-06-30)
tried to use the correct environment variable to disable paging
from multiple calls to "git log" and friends, but there was one
extra call to "git log" that was not covered by the trick.
Move the setting and exporting of GIT_PAGER much earlier in the
script to cover everybody.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This comment mentions the case where use_terminator is set,
but this case is not handled at all by this chunk of code.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
attr: plug minor memory leak
request-pull: really disable pager
Makes some cleanup/review in gittutorial
Makefile: git.o depends on library headers
git-submodule documentation: fix foreach example
Free the memory allocated for struct strbuf pathbuf when we're done.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When creating a new argv array from a configured alias and the supplied
command line arguments, the new argv was allocated with one element too
many. Since the first element of the original argv array is skipped when
copying it to the new_argv, the number of elements that are allocated
should be reduced by one. 'count' is the number of elements that new_argv
contains, and *argcp is the number of elements in the original argv array.
So the total allocation (including the terminating NULL entry) for the
new_argv array should be:
count + (*argcp - 1) + 1
Also, the explicit assignment of the NULL terminating entry can be avoided
by just copying it over from the original argv array.
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
ff06c74 (Improve request-pull to handle non-rebased branches, 2007-05-01)
attempted to disable pager when running subcommands in this script, but
with a wrong variable. If GIT_PAGER is set, it takes precedence over
PAGER.
Noticed by Michal Marek.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The following invocations did not work as expected on Windows:
git mv foo\bar dest
git mv foo\ dest
The first command was interpreted as
git mv foo/bar dest/foo/bar
because the Windows style directory separator was not obeyed when the
basename of 'foo\bar' was computed.
The second command failed because the Windows style directory separator was
not removed from the source directory, whereupon the lookup of the
directory in the index failed.
This fixes both issues by using is_dir_sep() and basename().
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update to include branch.*.rebase, remote.*.pushurl, and
add.ignore-errors
Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are some different but little cleanup changes to fix some missing
quotes, to fix what seemed to be an unended sentence, to reident a
little paragraph with too large a sentence and fix a branch name that
was referred to twice later by another name.
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@holoscopio.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Backtick and apostrophe are asciidoc markup, so they should be escaped
in order to get the expected result in the rendered manual page.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I do various statistics on git, and one of the things I look at is merges,
because they are often interesting events to count ("how many merges vs
how much 'real development'" kind of statistics). And you can do it with
some fairly straightforward scripting, ie
git rev-list --parents HEAD |
grep ' .* ' |
git diff-tree --always -s --pretty=oneline --stdin |
less -S
will do it.
But I finally got irritated with the fact that we can skip merges with
'--no-merges', but we can't do the trivial reverse operation.
So this just adds a '--merges' flag that _only_ shows merges. Now you can
do the above with just a
git log --merges --pretty=oneline
which is a lot simpler. It also means that we automatically get a lot of
statistics for free, eg
git shortlog -ns --merges
does exactly what you'd want it to do.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://git.bogomips.org/git-svn:
git svn: Doc update for multiple branch and tag paths
git svn: cleanup t9138-multiple-branches
git-svn: Canonicalize svn urls to prevent libsvn assertion
t9138: remove stray dot in test which broke bash
git-svn: convert globs to regexps for branch destinations
git svn: Support multiple branch and tag paths in the svn repository.
Add 'git svn reset' to unwind 'git svn fetch'
git-svn: speed up find_rev_before
Add 'git svn help [cmd]' which works outside a repo.
git-svn: let 'dcommit $rev' work on $rev instead of HEAD
Using the "svn_cmd" wrapper instead of "svn" alone allows tests
to run consistently for users with customized
~/.subversion/configs. Additionally, using subshells via
"(cd ...)" allow cleaner and less error-prone tests to
be written.
[ew: expanded commit message]
Signed-off-by: Marc Branchaud <marcnarc@xiplink.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
When combining "dumb client" and human-friendly access by using the
'.git' extension to switch between the two, make sure the AliasMatch
covers the entire request. Without a full match, a request for
http://git.example.com/project/shortlog/branch..gitsomething
would result in a 404 because the server would try to access the
the project 'project/shortlog/branch.'
The solution is still not bulletproof, so document the possible failing
case.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a small test case for git archive --remote (and thus
git-upload-archive), which so far went untested.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Cloning/initializing svn repositories with an uncanonicalize url
does not work as libsvn throws an assertion. This patch
canonicalize svn uris for the clone and init command from
git-svn.
[ew: fixed trailing whitespace]
Signed-off-by: Ulrich Dangel <uli@spamt.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
The stray dot broke bash and probably some other shells,
but worked fine with dash in my limited testing.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
For users pulling from machines with self compiled git installs,
in non-PATH locations, they can set the config option
remote.<name>.uploadpack to set the location of git-upload-pack.
When using 'git remote show <name>', the remote HEAD check
did not use the uploadpack configuration setting, and would
not use the configured program.
In builtin-remote.c, the config setting is already loaded
with the call to remote_get(), so this patch passes that remote
along to transport_get().
Signed-off-by: Chris Frey <cdfrey@foursquare.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Marc Branchaud wrote:
> I'm fairly happy with this, except for the way the branch
> subcommand matches refspecs. The patch does a simple string
> comparison, but it'd be better to do an actual glob. I just
> couldn't track down the right function for that, so I left it as
> a strcmp and hope that a gitizen can tell me how to glob here.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
This enables git-svn.perl to read multiple 'branches' and 'tags' entries in
svn-remote config sections. The init and clone subcommands also support
multiple --branches and --tags arguments.
The branch (and tag) subcommand gets a new argument: --destination (or -d).
This argument is required if there are multiple branches (or tags) entries
configured for the remote Subversion repository. The argument's value
specifies which branch (or tag) path to use to create the branch (or tag).
The specified value must match the left side (without wildcards) of one of
the branches (or tags) refspecs in the svn-remote's config.
[ew: avoided explicit loop when combining globs with "push"]
Signed-off-by: Marc Branchaud <marcnarc@xiplink.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Add a command to unwind the effects of fetch by moving the rev_map
and refs/remotes/git-svn back to an old SVN revision. This allows
revisions to be re-fetched. Ideally SVN revs would be immutable,
but permissions changes in the SVN repository or indiscriminate use
of '--ignore-paths' can create situations where fetch cannot make
progress.
Signed-off-by: Ben Jackson <ben@ben.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
By limiting start revision of find_rev_before to max existing
revision. This avoids a long wait if you do
'git svn reset -r 9999999'. The linear search within the
contiguous revisions doesn't seem to be a problem.
[ew: expanded commit message]
Signed-off-by: Ben Jackson <ben@ben.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Previously there was no explicit 'help' command, but 'git svn help'
still printed the usage message (as an invalid command), provided you
got past the initialization steps that required a valid repo.
Signed-off-by: Ben Jackson <ben@ben.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
'git svn dcommit' takes an optional revision argument, but the meaning
of it was rather scary. It completely ignored the current state of
the HEAD, only looking at the revisions between SVN and $rev. If HEAD
was attached to $branch, the branch lost all commits $rev..$branch in
the process.
Considering that 'git svn dcommit HEAD^' has the intuitive meaning
"dcommit all changes on my branch except the last one", we change the
meaning of the revision argument. git-svn temporarily checks out $rev
for its work, meaning that
* if a branch is specified, that branch (_not_ the HEAD) is rebased as
part of the dcommit,
* if some other revision is specified, as in the example, all work
happens on a detached HEAD and no branch is affected.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Acked-by: Eric Wong <normalperson@yhbt.net>
size_t res cannot be less than 0. fread returns 0 on error.
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the shell is not specified using the '#!' notation, then the OS will
use '/bin/sh' to execute the script which may not produce the desired
results. In particular, /bin/sh on Solaris interprets '^' specially which
has an effect on the sed command that this patch touches.
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new test does a 'chmod 0', which does not have the intended
effect on Windows.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
