This is just a basic sanity test to see whether
core.gitproxy works at all. Until now, we were not testing
anywhere.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 09c9957c fixes a deadlock in which pack-objects
fails, the remote end is still waiting for pack data, and we
are still waiting for the remote end to say something (see
that commit for a much more in-depth explanation).
We solved the problem there by making sure the output pipe
is closed on error; thus the remote sees EOF, and proceeds
to complain and close its end of the connection.
However, in the special case of push over git://, we don't
have a pipe, but rather a full-duplex socket, with another
dup()-ed descriptor in place of the second half of the pipe.
In this case, closing the second descriptor signals nothing
to the remote end, and we still deadlock.
This patch calls shutdown() explicitly to signal EOF to the
other side.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
They might care because they want to do a half-duplex close.
With pipes, that means simply closing the output descriptor;
with a socket, you must actually call shutdown.
Instead of exposing the magic no_fork child_process struct,
let's encapsulate the test in a function.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The git_connect function returns two ends of a pipe for
talking with a remote, plus a struct child_process
representing the other end of the pipe. If we have a direct
socket connection, then this points to a special "no_fork"
child process.
The code path for doing git-over-pipes or git-over-ssh sets
up this child process to point to the child git command or
the ssh process. When we call finish_connect eventually, we
check wait() on the command and report its return value.
The code path for git://, on the other hand, always sets it
to no_fork. In the case of a direct TCP connection, this
makes sense; we have no child process. But in the case of a
proxy command (configured by core.gitproxy), we do have a
child process, but we throw away its pid, and therefore
ignore its return code.
Instead, let's keep that information in the proxy case, and
respect its return code, which can help catch some errors
(though depending on your proxy command, it will be errors
reported by the proxy command itself, and not propagated
from git commands. Still, it is probably better to propagate
such errors than to ignore them).
It also means that the child_process field can reliably be
used to determine whether the returned descriptors are
actually a full-duplex socket, which means we should be
using shutdown() instead of a simple close.
Signed-off-by: Jeff King <peff@peff.net>
Helped-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 09c9957 (send-pack: avoid deadlock when pack-object
dies early, 2011-04-25) attempted to fix a hang in the
stateless rpc case by closing a file descriptor early, but
we still need that descriptor.
Basically the deadlock can happen when pack-objects fails,
and the descriptor to upstream is left open. We never send
the pack, so the upstream is left waiting for us to say
something, and we are left waiting for upstream to close the
connection.
In the non-rpc case, our descriptor points straight to the
upstream. We hand it off to run-command, which takes
ownership and closes the descriptor after pack-objects
finishes (whether it succeeds or not).
Commit 09c9957 tried to emulate that in the rpc case. That
isn't right, though. We actually have a descriptor going
back to the remote-helper, and we need to keep using it
after pack-objects is finished. Closing it early completely
breaks pushing via smart-http.
We still need to do something on error to signal the
remote-helper that we won't be sending any pack data
(otherwise we get the deadlock). In an ideal world, we
would send a special packet back that says "Sorry, there was
an error". But the remote-helper doesn't understand any such
packet, so the best we can do is close the descriptor and
let it report that we hung up unexpectedly.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Send-pack deadlocks in two ways when pack-object dies early (for example,
because there is some repo corruption).
The first deadlock happens with the smart push protocol (--stateless-rpc).
After the initial rev-exchange, the remote is waiting for the pack data
to arrive, and the sideband demuxer at the local side continues trying to
stream data from the remote repository until it gets EOF. Meanwhile,
send-pack (in function pack_objects()) has noticed that pack-objects did
not produce output and died. Back in send_pack(), it now tries to clean
up the sideband demuxer using finish_async(). The demuxer, however, waits
for the remote end to close down, the remote waits for pack data, and
the reason that it still waits is that send-pack forgot to close the
outgoing channel. Add the missing close() in pack_objects().
The second deadlock happens in a similar constellation when the sideband
demuxer runs in a forked process (rather than in a thread). Again, the
remote end waits for pack data to arrive, the sideband demuxer waits for
the remote to shut down, and send-pack (in the regular clean-up) waits for
the demuxer to terminate. This time, the send-pack parent process closes
the writable end of the outgoing channel (in start_command() that spawned
pack-objects) so that after the death of the pack-objects process all
writable ends should have been closed and the remote repo should see EOF.
This does not happen, however, because when the sideband demuxer was forked
earlier, it also inherited a writable end; it remains open and keeps the
remote repo from seeing EOF. To break this deadlock, close the writable end
in the demuxer.
Analyzed-by: Jeff King <peff@peff.net>
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the client requests the side-band-64k protocol capability we
now wrap the status report data inside of packets sent to band #1.
This permits us to later send additional progress or informational
messages down band #2.
If side-band-64k was enabled, we always send a final flush packet
to let the client know we are done transmitting.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Moving capability advertisement into the packet_write call itself
makes it easier to add additional capabilities to the list, be
it optional by configuration, or always present in the protocol.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the server advertises side-band-64k capability, we request
it and pull the status report data out of side band #1, and let
side band #2 go to our stderr. The latter channel be used by the
remote side to send our user messages. This basically mirrors the
side-band-64k capability in upload-pack.
Servers may choose to use side band #2 to send error messages from
hook scripts that are meant for the push end user.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch adds the possibility to supply a set of non-0 file
descriptors for async process communication instead of the
default-created pipe.
Additionally, we now support bi-directional communiction with the
async procedure, by giving the async function both read and write
file descriptors.
To retain compatiblity and similar "API feel" with start_command,
we require start_async callers to set .out = -1 to get a readable
file descriptor. If either of .in or .out is 0, we supply no file
descriptor to the async process.
[sp: Note: Erik started this patch, and a huge bulk of it is
his work. All bugs were introduced later by Shawn.]
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Like .out, .err may now be set to a file descriptor > 0, which
is a writable pipe/socket/file that the child's stderr will be
redirected into.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The '--full' option to git fsck does two things:
1) Check objects in packs
2) Check alternate objects
This is documented in the git fsck manual; this patch reflects that in
the short git fsck option help message as well.
Signed-off-by: Wesley J. Landaker <wjl@icecavern.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Caught by valgrind in t5500, but it is pretty obvious from
reading the code that this is shifting elements of an array
to the left, which needs memmove.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commits a2d725b7 (Use an external program to implement fetching
with curl, 2009-08-05) and c9e388bb (Make the
"traditionally-supported" URLs a special case, 2009-09-03) remote
transport helpers like 'remote-ftp' and 'remote-curl' are offered by the
completion script as available subcommands. Not good, since they are
helpers, therefore should not be offered, so filter them out.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* 'maint' of git://git.spearce.org/git-gui:
git-gui: work from the .git dir
git-gui: Fix applying a line when all following lines are deletions
git-gui: Correct file_states when unstaging partly staged entry
git-gui: Fix gitk for branch whose name matches local file
git-gui: Keep repo_config(gui.recentrepos) and .gitconfig in sync
git-gui: handle really long error messages in updateindex.
git-gui: Add hotkeys for "Unstage from commit" and "Revert changes"
git-gui: Makefile: consolidate .FORCE-* targets
When git-gui is run from a .git dir, _gitdir would be set to "." by
rev-parse, something that confuses the worktree detection.
Fix by expanding the value of _gitdir to pwd in this special case.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If a diff looked like:
@@
context
-del1
-del2
and you wanted to stage the deletion 'del1', the generated patch
wouldn't apply because it was missing the line 'del2' converted to
context, but this line was counted in the @@-line
Signed-off-by: Jeff Epler <jepler@unpythonic.net>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When unstaging a partly staged file or submodule, the file_states
list was not updated properly (unless unstaged linewise). Its
index_info part did not contain the former head_info as it should
have but kept its old value.
This seems not to have had any bad effects but diminishes the value
of the file_states list for future enhancements.
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When trying to run gitk on a branch name whose name matches a local
file, it will toss an error saying that the name is ambiguous. Adding
a pair of dashes will make gitk parse the options to the left of
it as branch names. Since wish eats the first pair of dashes we
throw at it, we need to add a second one to ensure they get through.
Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When the number of recent repo's gets to ten there can be a
situation where an item is removed from the .gitconfig file via
a call to git config --unset, but the internal representation of
that file (repo_config(gui.recentrepo)) is not updated. Then a
subsequent attempt to remove an item from the list fails because
git-gui attempts to call --unset on a value that has already
been removed. This leads to duplicates in the .gitconfig file,
which then also cause errors if the git-gui tries to --unset them
(rather than using --unset-all. --unset-all is not used because it
is not expected that duplicates should ever be allowed to exist.)
When loading the list of recent repositories (proc _get_recentrepos)
if a repo in the list is not considered a valid git reposoitory
then we should go ahead and remove it so it doesn't take up a slot
in the list (since we limit to 10 items). This will prevent a bunch
of invalid entries in the list (which are not shown) from making
valid entries dissapear off the list even when there are less than
ten valid entries.
See: http://code.google.com/p/msysgit/issues/detail?id=362
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
As reported to msysGit (bug #340) it is possible to get some very
long error messages when updating the index. The use of a label to
display this prevents scrolling the output. This patch replaces the
label with a scrollable text widget configured to look like a label.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The function takes two paths, an early part of abs is supposed to match
base; otherwise abs is not a path under base and the function returns the
full path of abs. The caller can easily confuse the implementation by
giving duplicated and needless slashes in these path arguments.
Credit for test script, motivation and initial patch goes to Thomas Rast.
A follow-up fix (squashed) is by Hannes.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
merge-recursive calls write_tree_from_memory() to come up with a virtual
tree, with possible conflict markers inside the blob contents, while
merging multiple common ancestors down. It is a bug to call the function
with unmerged entries in the index, even if the merge to come up with the
common ancestor resulted in conflicts. Otherwise the result won't be
expressible as a tree object.
We _might_ want to suggest the user to set GIT_MERGE_VERBOSITY to 5 and
re-run the merge in the message. At least we will know which part of
process_renames() or process_entry() functions is not correctly handling
the unmerged paths, and it might help us diagnosing the issue.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When "info/refs" is a static file and not behind a CGI handler, some
servers may not handle a GET request for it with a query string
appended (eg. "?foo=bar") properly.
If such a request fails, retry it sans the query string. In addition,
ensure that the "smart" http protocol is not used (a service has to be
specified with "?service=<service name>" to be conformant).
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Reported-and-tested-by: Yaroslav Halchenko <debian@onerussian.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Giving "Notes" information in the default output format of "log" and
"show" is a sensible progress (the user has asked for it by having the
notes), but for some commands (e.g. "format-patch") spewing notes into the
formatted commit log message without being asked is too aggressive.
Enable notes output only for "log", "show", "whatchanged" by default and
only when the user didn't ask any specific --pretty/--format from the
command line; users can explicitly override this default with --show-notes
and --no-notes option.
Parts of tests are taken from Jeff King's fix.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* sb/maint-octopus:
octopus: remove dead code
octopus: reenable fast-forward merges
octopus: make merge process simpler to follow
Conflicts:
git-merge-octopus.sh
* bg/maint-add-all-doc:
git-rm doc: Describe how to sync index & work tree
git-add/rm doc: Consistently back-quote
Documentation: 'git add -A' can remove files
* maint-1.6.5:
Git 1.6.5.8
Fix mis-backport of t7002
bash completion: factor submodules into dirty state
reset: unbreak hard resets with GIT_WORK_TREE
Conflicts:
Documentation/git.txt
GIT-VERSION-GEN
RelNotes
We need to update the index before hooks run when actually making a
commit, but we shouldn't have to write the index when running "status".
If we can, then we have already spent cycles to refresh the index and
it is a waste not to write it out, but it is not a disaster if we cannot
write it out. The main reason the user is running "git status" is to get
the "status", and refreshing the index is a mere side effect that we can
do without.
Discovery and initial attempted fix by Dscho.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Remove the trailing 's' from "revisions" and "steps" when there is
only one.
Signed-off-by: David Ripton <dripton@ripton.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint-1.6.4:
Fix mis-backport of t7002
base85: Make the code more obvious instead of explaining the non-obvious
base85: encode_85() does not use the decode table
base85 debug code: Fix length byte calculation
checkout -m: do not try to fall back to --merge from an unborn branch
branch: die explicitly why when calling "git branch [-a|-r] branchname".
textconv: stop leaking file descriptors
commit: --cleanup is a message option
git count-objects: handle packs bigger than 4G
t7102: make the test fail if one of its check fails
* maint-1.6.3:
base85: Make the code more obvious instead of explaining the non-obvious
base85: encode_85() does not use the decode table
base85 debug code: Fix length byte calculation
checkout -m: do not try to fall back to --merge from an unborn branch
branch: die explicitly why when calling "git branch [-a|-r] branchname".
textconv: stop leaking file descriptors
commit: --cleanup is a message option
git count-objects: handle packs bigger than 4G
t7102: make the test fail if one of its check fails
Conflicts:
builtin-commit.c
* maint-1.6.2:
base85: Make the code more obvious instead of explaining the non-obvious
base85: encode_85() does not use the decode table
base85 debug code: Fix length byte calculation
checkout -m: do not try to fall back to --merge from an unborn branch
branch: die explicitly why when calling "git branch [-a|-r] branchname".
textconv: stop leaking file descriptors
commit: --cleanup is a message option
git count-objects: handle packs bigger than 4G
t7102: make the test fail if one of its check fails
Conflicts:
diff.c
The original patch that became cfe370c (grep: do not segfault when -f is
used, 2009-10-16), was made against "maint" or newer branch back then, but
the fix addressed the issue that was present as far as in 1.6.4 series.
The maintainer backported the patch to the 1.6.4 maintenance branch, but
failed to notice that the new tests assumed the setup done by the script
in "maint", which did quite a lot more than the same test script in 1.6.4
series, and the output didn't match the expected result.
This should fix it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>