t5570: fix forwarding of git-daemon messages via cat
The shell function that starts git-daemon wants to read the first line of
the daemon's stderr to ensure that it started correctly. Subsequent daemon
errors should be redirected to fd 4 (which is the terminal in verbose mode
or /dev/null in quiet mode). To that end the shell script used 'read' to
get the first line of output, and then 'cat &' to forward everything else
in a background process.
The problem is, that 'cat >&4 &' does not produce any output because the
shell redirects a background process's stdin to /dev/null. To have this
command invocation do anything useful, we have to redirect its stdin
explicitly (which overrides the /dev/null redirection).
The shell function connects the daemon's stderr to its consumers via a
FIFO. We cannot just do this:
read line <git_daemon_output
cat <git_daemon_output >&4 &
because after the first redirection the pipe is closed and the daemon
could receive SIGPIPE if it writes at the wrong moment. Therefore, we open
the readable end of the FIFO only once on fd 7 in the shell and dup from
there to the stdin of the two consumers.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In asciidoc 7, backticks like `foo` produced a typographic
effect, but did not otherwise affect the syntax. In asciidoc
8, backticks introduce an "inline literal" inside which markup
is not interpreted. To keep compatibility with existing
documents, asciidoc 8 has a "no-inline-literal" attribute to
keep the old behavior. We enabled this so that the
documentation could be built on either version.
It has been several years now, and asciidoc 7 is no longer
in wide use. We can now decide whether or not we want
inline literals on their own merits, which are:
1. The source is much easier to read when the literal
contains punctuation. You can use `master~1` instead
of `master{tilde}1`.
2. They are less error-prone. Because of point (1), we
tend to make mistakes and forget the extra layer of
quoting.
This patch removes the no-inline-literal attribute from the
Makefile and converts every use of backticks in the
documentation to an inline literal (they must be cleaned up,
or the example above would literally show "{tilde}" in the
output).
Problematic sites were found by grepping for '`.*[{\\]' and
examined and fixed manually. The results were then verified
by comparing the output of "html2text" on the set of
generated html pages. Doing so revealed that in addition to
making the source more readable, this patch fixes several
formatting bugs:
- HTML rendering used the ellipsis character instead of
literal "..." in code examples (like "git log A...B")
- some code examples used the right-arrow character
instead of '->' because they failed to quote
- api-config.txt did not quote tilde, and the resulting
HTML contained a bogus snippet like:
<tt><sub></tt> foo <tt></sub>bar</tt>
which caused some parsers to choke and omit whole
sections of the page.
- git-commit.txt confused ``foo`` (backticks inside a
literal) with ``foo'' (matched double-quotes)
- mentions of `A U Thor <author@example.com>` used to
erroneously auto-generate a mailto footnote for
author@example.com
- the description of --word-diff=plain incorrectly showed
the output as "[-removed-] and {added}", not "{+added+}".
- using "prime" notation like:
commit `C` and its replacement `C'`
confused asciidoc into thinking that everything between
the first backtick and the final apostrophe were meant
to be inside matched quotes
- asciidoc got confused by the escaping of some of our
asterisks. In particular,
`credential.\*` and `credential.<url>.\*`
properly escaped the asterisk in the first case, but
literally passed through the backslash in the second
case.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When PATH contains an unreadable directory, alias expansion code did
not kick in, and failed with an error that said "git-subcmd" was not
found.
By Jeff King (1) and Ramsay Jones (1)
* jk/run-command-eacces:
run-command: treat inaccessible directories as ENOENT
compat/mingw.[ch]: Change return type of exec functions to int
The 'push to upstream' implementation was broken in some corner
cases. "git push $there" without refspec, when the current branch is
set to push to a remote different from $there, used to push to $there
using the upstream information to a remote unreleated to $there.
* jc/push-upstream-sanity:
push: error out when the "upstream" semantics does not make sense
Merge branch 'jc/maint-clean-nested-worktree-in-subdir' into maint
"git clean -d -f" (not "-d -f -f") is supposed to protect nested
working trees of independent git repositories that exist in the
current project working tree from getting removed, but the protection
applied only to such working trees that are at the top-level of the
current project by mistake.
* jc/maint-clean-nested-worktree-in-subdir:
clean: preserve nested git worktree in subdirectories
Rename detection logic used to match two empty files as renames during
merge-recursive, leading unnatural mismerges.
By Jeff King
* jk/diff-no-rename-empty:
merge-recursive: don't detect renames of empty files
teach diffcore-rename to optionally ignore empty content
make is_empty_blob_sha1 available everywhere
drop casts from users EMPTY_TREE_SHA1_BIN
Merge branch 'jc/commit-unedited-template' into maint
When "git commit --template F" errors out because the user did not
touch the message, it claimed that it aborts due to "empty message",
which was utterly wrong.
By Junio C Hamano (4) and Adam Monsen (1)
* jc/commit-unedited-template:
Documentation/git-commit: rephrase the "initial-ness" of templates
git-commit.txt: clarify -t requires editing message
commit: rephrase the error when user did not touch templated log message
commit: do not trigger bogus "has templated message edited" check
t7501: test the right kind of breakage
Merge branch 'jc/commit-hook-authorship' into maint
"git commit --author=$name" did not tell the name that was being
recorded in the resulting commit to hooks, even though it does do so
when the end user overrode the authorship via the "GIT_AUTHOR_NAME"
environment variable.
* jc/commit-hook-authorship:
commit: pass author/committer info to hooks
t7503: does pre-commit-hook learn authorship?
ident.c: add split_ident_line() to parse formatted ident line
Merge branch 'tr/maint-word-diff-regex-sticky' into maint
The regexp configured with diff.wordregex was incorrectly reused
across files.
By Thomas Rast (2) and Johannes Sixt (1)
* tr/maint-word-diff-regex-sticky:
diff: tweak a _copy_ of diff_options with word-diff
diff: refactor the word-diff setup from builtin_diff_cmd
t4034: diff.*.wordregex should not be "sticky" in --word-diff
Merge branch 'jh/notes-merge-in-git-dir-worktree' into maint
Running "notes merge --commit" failed to perform correctly when run
from any directory inside $GIT_DIR/. When "notes merge" stops with
conflicts, $GIT_DIR/NOTES_MERGE_WORKTREE is the place a user edits
to resolve it.
By Johan Herland (3) and Junio C Hamano (1)
* jh/notes-merge-in-git-dir-worktree:
notes-merge: Don't remove .git/NOTES_MERGE_WORKTREE; it may be the user's cwd
notes-merge: use opendir/readdir instead of using read_directory()
t3310: illustrate failure to "notes merge --commit" inside $GIT_DIR/
remove_dir_recursively(): Add flag for skipping removal of toplevel dir
bundle: remove stray single-quote from error message
After running rev-list --boundary to retrieve the list of boundary
commits, "git bundle create" runs its own revision walk. If in this
stage git encounters an unfamiliar option, it writes a message with an
unbalanced quotation mark:
error: unrecognized argument: --foo'
Drop the stray quote to match the "unrecognized argument: %s" message
used elsewhere and save translators some work.
This is mostly a futureproofing measure: for now, the "rev-list
--boundary" command catches most strange arguments on its own and the
above message is not seen unless you try something esoteric like "git
bundle create test.bundle --header HEAD".
Reported-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
config: reject bogus section names for --rename-section
You can feed junk to "git config --rename-section", which
will result in a config file that git will not even parse
(so you cannot fix it with git-config). We already have
syntactic sanity checks when setting a variable; let's do
the same for section names.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-remote-testgit: fix race when spawning fast-import
Test "pushing to local repo" in t5800-remote-helpers can hang
due to a race condition in git-remote-testgit. Fix it by
setting stdin to unbuffered.
On the writer side, "git push" invokes push_refs_with_export(),
which sends to stdout the command "export\n" and immediately
starts up "git fast-export". The latter writes its output stream
to the same stdout.
On the reader side, remote helper "git-remote-testgit" reads from
stdin to get its next command. It uses getc() to read characters
from libc up until \n. Libc has buffered a potentially much
larger chunk of stdin. When it sees the "export\n" command, it
forks "git fast-import" to read the stream.
If fast-export finishes before git fast-import starts, the
fast-export output can end up in libc's buffer in
git-remote-testgit, rather than in git fast-import. The latter
hangs indefinitely on a now-empty stdin.
Signed-off-by: Pete Wyckoff <pw@padd.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t9300-fast-import: avoid 'exit' in test_expect_success snippets
Exiting from a for-loop early using '|| break' does not propagate the
failure code, and for this reason, the tests used just 'exit'. But this
ends the test script with 'FATAL: Unexpected exit code 1' in the case of
a failed test.
Fix this by moving the loop into a shell function, from which we can
simply return early.
While at it, modernize the style of the affected test cases.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is used by "git pull" to construct a merge message from list of
remote refs. When pulling redundant set of refs, however, it did not
filter them even though the merge itself discards them as unnecessary.
git-gc executes many sub-commands. The argument list for
some of these is constant, but for others we add more
arguments at runtime. The latter is implemented by allocating
a constant extra number of NULLs, and either using a custom
append function, or just referencing unused slots by number.
As of commit 7e52f56, which added two new arguments, it is
possible to exceed the constant number of slots for "repack"
by running "git gc --aggressive", causing "git gc" to die.
This patch converts all of the static argv lists to use
argv-array. In addition to fixing the overflow caused by 7e52f56, it has a few advantages:
1. We can drop the custom append function (which,
incidentally, had an off-by-one error exacerbating the
static limit).
2. We can drop the ugly magic numbers used when adding
arguments to "prune".
3. Adding further arguments will be easier; you can just
add new "push" calls without worrying about increasing
any static limits.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
It can be convenient to push many strings in a single line
(e.g., if you are initializing an array with defaults). This
patch provides a convenience wrapper to allow this.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
An empty argv-array is initialized to point to a static
empty NULL-terminated array. The original implementation
separates the actual storage of the NULL-terminator from the
pointer to the list. This makes the exposed type a "const
char **", which nicely matches the type stored by the
argv-array.
However, this indirection means that one cannot use
empty_argv to initialize a static variable, since it is
not a constant.
Instead, we can expose empty_argv directly, as an array of
pointers. The only place we use it is in the ARGV_ARRAY_INIT
initializer, and it decays to a pointer appropriately there.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When COLUMNS or --stat-width restricts the diff-stat width to near the
minimum, 26 columns, the graph_width value becomes negative. Consequently, the
graph part of diff-stat is not resized properly.
Signed-off-by: Lucian Poston <lucian.poston@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of waiting until we record the parents of resulting merge, reduce
redundant parents (including our HEAD) immediately after reading them.
The change to t7602 illustrates the essence of the effect of this change.
The octopus merge strategy used to be fed with redundant commits only to
discard them as "up-to-date", but we no longer feed such redundant commits
to it and the affected test degenerates to a regular two-head merge.
And obviously the known-to-be-broken test in t6028 is now fixed.
This happens when git merge is run to merge multiple commits that are
descendants of current HEAD (or are HEAD). We've hit this while updating
master to origin/master but accidentaly we called (while being on master):
$ git merge master origin/master
Here is a minimal testcase:
$ git init a && cd a
$ echo a >a && git add a
$ git commit -minitial
$ echo b >a && git add a
$ git commit -msecond
$ git checkout master^
$ git merge master master
Fast-forwarding to: master
Already up-to-date with master
Merge made by the 'octopus' strategy.
a | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
fetch: describe new refs based on where it came from
update_local_ref() used to say "[new branch]" when we stored a new ref
outside refs/tags/ hierarchy, but the message is more about what we
fetched, so use the refname at the origin to make that decision.
Also, only call a new ref a "branch" if it's under refs/heads/.
Signed-off-by: Marc Branchaud <marcnarc@xiplink.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Adjust stat width calculations to take --graph output into account
The recent change to compute the width of diff --stat did not take into
consideration the output from --graph. The consequence is that when both
options are used, e.g. in 'log --stat --graph', the lines are too long.
Adjust stat width calculations to take --graph output into account.
Signed-off-by: Lucian Poston <lucian.poston@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add tests which show that the width of the --prefix added by --graph
is not taken into consideration when the diff stat output width is
calculated.
Signed-off-by: Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl> Signed-off-by: Lucian Poston <lucian.poston@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff: avoid stack-buffer-read-overrun for very long name
Due to the use of strncpy without explicit NUL termination,
we could end up passing names n1 or n2 that are not NUL-terminated
to queue_diff, which requires NUL-terminated strings.
Ensure that each is NUL terminated.
Signed-off-by: Jim Meyering <meyering@redhat.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a commit-ish passed to cherry-pick or revert happens to have a file
of the same name, git complains that the argument is ambiguous and
advises to use '--'. To make things worse, the '--' argument is removed
by parse_options, und so passing '--' has no effect.
Instead, always interpret cherry-pick/revert arguments as revisions.
Signed-off-by: Clemens Buchacher <drizzd@aon.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Various failure modes in the repository detection code path currently
quote the wrong directory in their error message. The working directory
is changed iteratively to the parent directory until a git repository is
found. If the working directory cannot be changed to the parent
directory for some reason, the detection gives up and prints an error
message. The error message should report the current working directory.
Instead of continually updating the 'cwd' variable, which is actually
used to remember the original working directory, the 'offset' variable
is used to keep track of the current working directory. At the point
where the affected error handling code is called, 'offset' already
points to the end of the parent of the working directory, rather than
the current working directory.
Fix this by explicitly using a variable 'offset_parent' and update
'offset' concurrently with the call to chdir.
In a similar fashion, the function get_device_or_die() would print the
original working directory in case of a failure, rather than the current
working directory. Fix this as well by making use of the 'offset'
variable.
Lastly, replace the phrase 'mount parent' with 'mount point'. The former
appears to be a typo.
Signed-off-by: Clemens Buchacher <drizzd@aon.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
submodules: recursive fetch also checks new tags for submodule commits
Since 88a21979c (fetch/pull: recurse into submodules when necessary) all
fetched commits are examined if they contain submodule changes (unless
configuration or command line options inhibit that). If a newly recorded
submodule commit is not present in the submodule, a fetch is run inside
it to download that commit.
Checking new refs was done in an else branch where it wasn't executed for
tags. This normally isn't a problem because tags are only fetched with
the branches they live on, then checking the new commits in the fetched
branches for submodule commits will also process all tags. But when a
specific tag is fetched (or the refspec contains refs/tags/) commits only
reachable by tags won't be searched for submodule commits, which is a bug.
Fix that by moving the code outside the if/else construct to handle new
tags just like any other ref. The performance impact of adding tags that
most of the time lie on a branch which is checked anyway for new submodule
commit should be minimal, as since 6859de4 (fetch: avoid quadratic loop
checking for updated submodules) all ref-tips are collected first and then
fed to a single rev-list.
Spotted-by: Jeff King <peff@peff.net> Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http: use newer curl options for setting credentials
We give the username and password to curl by sticking them
in a buffer of the form "user:pass" and handing the result
to CURLOPT_USERPWD. Since curl 7.19.1, there is a split
mechanism, where you can specify each element individually.
This has the advantage that a username can contain a ":"
character. It also is less code for us, since we can hand
our strings over to curl directly. And since curl 7.17.0 and
higher promise to copy the strings for us, we we don't even
have to worry about memory ownership issues.
Unfortunately, we have to keep the ugly code for old curl
around, but as it is now nicely #if'd out, we can easily get
rid of it when we decide that 7.19.1 is "old enough".
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we have a credential to give to curl, we must copy it
into a "user:pass" buffer and then hand the buffer to curl.
Old versions of curl did not copy the buffer, and we were
expected to keep it valid. Newer versions of curl will copy
the buffer.
Our solution was to use a strbuf and detach it, giving
ownership of the resulting buffer to curl. However, this
meant that we were leaking the buffer on newer versions of
curl, since curl was just copying it and throwing away the
string we passed. Furthermore, when we replaced a
credential (e.g., because our original one was rejected), we
were also leaking on both old and new versions of curl.
This got even worse in the last patch, which started
replacing the credential (and thus leaking) on every http
request.
Instead, let's use a static buffer to make the ownership
more clear and less leaky. We already keep a static "struct
credential", so we are only handling a single credential at
a time, anyway.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gc: do not explode objects which will be immediately pruned
When we pack everything into one big pack with "git repack
-Ad", any unreferenced objects in to-be-deleted packs are
exploded into loose objects, with the intent that they will
be examined and possibly cleaned up by the next run of "git
prune".
Since the exploded objects will receive the mtime of the
pack from which they come, if the source pack is old, those
loose objects will end up pruned immediately. In that case,
it is much more efficient to skip the exploding step
entirely for these objects.
This patch teaches pack-objects to receive the expiration
information and avoid writing these objects out. It also
teaches "git gc" to pass the value of gc.pruneexpire to
repack (which in turn learns to pass it along to
pack-objects) so that this optimization happens
automatically during "git gc" and "git gc --auto".
Signed-off-by: Jeff King <peff@peff.net> Acked-by: Nicolas Pitre <nico@fluxnic.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The allocations made by unpack_nondirectories() using create_ce_entry()
are never freed.
In the non-merge case, we duplicate them using add_entry() and later
only look at the first allocated element (src[0]), perhaps even only
by mistake. Split out the actual addition from add_entry() into the
new helper do_add_entry() and call this non-duplicating function
instead of add_entry() to avoid the leak.
Valgrind reports this for the command "git archive v1.7.9" without
the patch:
==13372== LEAK SUMMARY:
==13372== definitely lost: 230,986 bytes in 2,325 blocks
==13372== indirectly lost: 0 bytes in 0 blocks
==13372== possibly lost: 98 bytes in 1 blocks
==13372== still reachable: 2,259,198 bytes in 3,243 blocks
==13372== suppressed: 0 bytes in 0 blocks
And with the patch applied:
==13375== LEAK SUMMARY:
==13375== definitely lost: 65 bytes in 1 blocks
==13375== indirectly lost: 0 bytes in 0 blocks
==13375== possibly lost: 0 bytes in 0 blocks
==13375== still reachable: 2,364,417 bytes in 3,245 blocks
==13375== suppressed: 0 bytes in 0 blocks
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
unpack-trees: don't perform any index operation if we're not merging
src[0] points to the index entry in the merge case and to the first
tree to unpack in the non-merge case. We only want to mark the index
entry, so check first if we're merging.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
remote-curl: main test case for the OS command line overflow
This is main test case for the original problem that triggered this
patch series. We create a repo with 50k tags and then test whether
git-clone over the smart HTTP protocol succeeds.
Note that we construct the repo in a slightly different way than the
original script used to reproduce the problem. This is because the
original script just created 50k tags all pointing to the same commit,
so if there was a bug where remote-curl.c was not passing all the refs
to fetch-pack we wouldn't know. The clone would succeed even if only one
tag was passed, because all the other tags were pointing at the same SHA
and would be considered present.
Instead we create a repo with 50k independent (dangling) commits and
then tag each of those commits with a unique tag. This way if one of the
tags is not given to fetch-pack, later stages of the clone would
complain about it.
This allows us to test both that the command line overflow was fixed, as
well as that it was fixed in a way that doesn't leave out any of the
refs.
Signed-off-by: Ivan Todoroski <grnch@gmx.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
These test cases focus only on testing the parsing of refs on stdin,
without bothering with the rest of the fetch-pack machinery. We pass in
the refs using different combinations of command line and stdin and then
we watch fetch-pack's stdout to see whether it prints all the refs we
specified (but we ignore their order).
Signed-off-by: Ivan Todoroski <grnch@gmx.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we can throw an arbitrary number of refs at fetch-pack using
its --stdin option, we use it in the remote-curl helper to bypass the
OS command line length limit.
Signed-off-by: Ivan Todoroski <grnch@gmx.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The syntax for the use of mark references in fast-import
demands either a SP (space) or LF (end-of-line) after
a mark reference. Fast-import does not complain when garbage
appears after a mark reference in some cases.
Factor out parsing of mark references and complain if
errant characters are found. Also be a little more careful
when parsing "inline" and SHA1s, complaining if extra
characters appear or if the form of the dataref is unrecognized.
Buggy input can cause fast-import to produce the wrong output,
silently, without error. This makes it difficult to track
down buggy generators of fast-import streams. An example is
seen in the last line of this commit command:
commit refs/heads/S2
committer Name <name@example.com> 1112912893 -0400
data <<COMMIT
commit message
COMMIT
from :1M 100644 :103 hello.c
It is missing a newline and should be:
[...]
from :1
M 100644 :103 hello.c
What fast-import does is to produce a commit with the same
contents for hello.c as in refs/heads/S2^. What the buggy
program was expecting was the contents of blob :103. While
the resulting commit graph looked correct, the contents in
some commits were wrong.
Signed-off-by: Pete Wyckoff <pw@padd.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Check if we even have a parameter before checking its value. Running
this command without any arguments may not make a lot of sense, but
reacting with a segmentation fault is unduly harsh.
While we're at it, avoid casting argv by declaring it const right away.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Print out a trailing newline when --show-prefix is run with cwd
at the top level of the tree which results in an empty prefix.
Behavior is now like --show-cdup.
Fixes an expected failure in t1501.
Signed-off-by: Ross Lagerwall <rosslagerwall@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
HTTP authentication is currently handled by get_refs and fetch_ref, but
not by fetch_object, fetch_pack or fetch_alternates. In the
single-threaded case, this is not an issue, since get_refs is always
called first. It recognigzes the 401 and prompts the user for
credentials, which will then be used subsequently.
If the curl multi interface is used, however, only the multi handle used
by get_refs will have credentials configured. Requests made by other
handles fail with an authentication error.
Fix this by setting CURLOPT_USERPWD whenever a slot is requested.
Signed-off-by: Clemens Buchacher <drizzd@aon.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
rebase -i continue: don't skip commits that only change submodules
When git-rebase--interactive stops due to a conflict and the only change
to be committed is in a submodule, the test for whether there is
anything to be committed ignores the staged submodule change. This
leads rebase to skip creating the commit for the change.
While unstaged submodule changes should be ignored to avoid needing to
update submodules during a rebase, it is safe to remove the
--ignore-submodules option to diff-index because --cached ensures that
it is only checking the index. This was discussed in [1] and a test is
included to ensure that unstaged changes are still ignored correctly.
Avoid bug in Solaris xpg4/sed as used in submodule
The sed provided by Solaris in /usr/xpg4/bin has a bug whereby an
unanchored regex using * for zero or more repetitions sees two
separate matches fed to the substitution engine in some cases.
This is evidenced by:
$ for sed in /usr/xpg4/bin/sed /usr/bin/sed /opt/csw/gnu/sed; do \
echo 'ab' | $sed -e 's|[a]*|X|g'; \
done
XXbX
XbX
XbX
This bug was triggered during a git submodule clone operation as
exercised in the setup stage of t5526-fetch-submodules when using the
default SANE_TOOL_PATH for Solaris. It led to paths such as
..../.. being used in the submodule .git gitdir reference.
Using the expression 's|\([^/]*\(/*\)\)|..\2|g' provides the desired
result with all three three tested sed implementations but is harder
to read. As we do not need to handle fully qualified paths though,
the expression could actually be [^/]+ which isn't properly handled
either. Instead, use [^/][^/]*, as suggested by Andreas Schwab, which
works on all three tested sed implementations.
The new expression is semantically different than the original one.
It will not place a leading '..' on a fully qualified path as the
original expression did. All of the paths being passed through this
regex are relative and did not rely on this behaviour so it's a safe
change.
Signed-off-by: Ben Walton <bwalton@artsci.utoronto.ca> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* tr/cache-tree:
t0090: be prepared that 'wc -l' writes leading blanks
reset: update cache-tree data when appropriate
commit: write cache-tree data when writing index anyway
Refactor cache_tree_update idiom from commit
Test the current state of the cache-tree optimization
Add test-scrap-cache-tree
Merge branch 'cb/maint-t5541-make-server-port-portable' into maint-1.7.8
* cb/maint-t5541-make-server-port-portable:
t5541: check error message against the real port number used
remote-curl: Fix push status report when all branches fail
Merge branch 'tr/maint-bundle-boundary' into maint-1.7.8
* tr/maint-bundle-boundary:
bundle: keep around names passed to add_pending_object()
t5510: ensure we stay in the toplevel test dir
t5510: refactor bundle->pack conversion
Merge branch 'tr/maint-bundle-long-subject' into maint-1.7.8
* tr/maint-bundle-long-subject:
t5704: match tests to modern style
strbuf: improve strbuf_get*line documentation
bundle: use a strbuf to scan the log for boundary commits
bundle: put strbuf_readline_fd in strbuf.c with adjustments
run-command: treat inaccessible directories as ENOENT
When execvp reports EACCES, it can be one of two things:
1. We found a file to execute, but did not have
permissions to do so.
2. We did not have permissions to look in some directory
in the $PATH.
In the former case, we want to consider this a
permissions problem and report it to the user as such (since
getting this for something like "git foo" is likely a
configuration error).
In the latter case, there is a good chance that the
inaccessible directory does not contain anything of
interest. Reporting "permission denied" is confusing to the
user (and prevents our usual "did you mean...?" lookup). It
also prevents git from trying alias lookup, since we do so
only when an external command does not exist (not when it
exists but has an error).
This patch detects EACCES from execvp, checks whether we are
in case (2), and if so converts errno to ENOENT. This
behavior matches that of "bash" (but not of simpler shells
that use execvp more directly, like "dash").
Test stolen from Junio.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
compat/mingw.[ch]: Change return type of exec functions to int
The POSIX standard specifies a return type of int for all six exec
functions. In addition, all exec functions return -1 on error, and
simply do not return on success. However, the current emulation of
the exec functions on mingw are declared with a void return type.
This would cause a problem should any code attempt to call the
exec function in a non-void context. In particular, if an exec
function were used in a conditional it would fail to compile.
In order to improve the fidelity of the emulation, we change the
return type of the mingw_execv[p] functions to int and return -1
on error.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
push: error out when the "upstream" semantics does not make sense
The user can say "git push" without specifying any refspec. When using
the "upstream" semantics via the push.default configuration, the user
wants to update the "upstream" branch of the current branch, which is the
branch at a remote repository the current branch is set to integrate with,
with this command.
However, there are cases that such a "git push" that uses the "upstream"
semantics does not make sense:
- The current branch does not have branch.$name.remote configured. By
definition, "git push" that does not name where to push to will not
know where to push to. The user may explicitly say "git push $there",
but again, by definition, no branch at repository $there is set to
integrate with the current branch in this case and we wouldn't know
which remote branch to update.
- The current branch does have branch.$name.remote configured, but it
does not specify branch.$name.merge that names what branch at the
remote this branch integrates with. "git push" knows where to push in
this case (or the user may explicitly say "git push $remote" to tell us
where to push), but we do not know which remote branch to update.
- The current branch does have its remote and upstream branch configured,
but the user said "git push $there", where $there is not the remote
named by "branch.$name.remote". By definition, no branch at repository
$there is set to integrate with the current branch in this case, and
this push is not meant to update any branch at the remote repository
$there.
The first two cases were already checked correctly, but the third case was
not checked and we ended up updating the branch named branch.$name.merge
at repository $there, which was totally bogus.
add--interactive: ignore unmerged entries in patch mode
When "add -p" sees an unmerged entry, it shows the combined
diff and then immediately skips the hunk. This can be
confusing in a variety of ways, depending on whether there
are other changes to stage (in which case you get the
superfluous combined diff output in between other hunks) or
not (in which case you get the combined diff and the program
exits immediately, rather than seeing "No changes").
The current behavior was not planned, and is just what the
implementation happens to do. Instead, let's explicitly
remove unmerged entries from our list of modified files, and
print a warning that we are ignoring them.
We can cheaply find which entries are unmerged by adding
"--raw" output to the "diff-files --numstat" we already run.
There is one non-obvious thing we must change when parsing
this combined output. Before this patch, when we saw a
numstat line for a file that did not have index changes, we
would create a new record with 'unchanged' in the 'INDEX'
field. Because "--raw" comes before "--numstat", we must
move this special-case down to the raw-line case (and it is
sufficient to move it rather than handle it in both places,
since any file which has a --numstat will also have a --raw
entry).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-commit: rephrase the "initial-ness" of templates
The description of "commit -t <file>" said the file is used "as the
initial version" of the commit message, but in the context of an SCM,
"version" is a loaded word that can needlesslyl confuse readers.
Explain the purpose of the mechanism without using "version".
fetch-pack: new --stdin option to read refs from stdin
If a remote repo has too many tags (or branches), cloning it over the
smart HTTP transport can fail because remote-curl.c puts all the refs
from the remote repo on the fetch-pack command line. This can make the
command line longer than the global OS command line limit, causing
fetch-pack to fail.
This is especially a problem on Windows where the command line limit is
orders of magnitude shorter than Linux. There are already real repos out
there that msysGit cannot clone over smart HTTP due to this problem.
Here is an easy way to trigger this problem:
git init too-many-refs
cd too-many-refs
echo bla > bla.txt
git add .
git commit -m test
sha=$(git rev-parse HEAD)
tag=$(perl -e 'print "bla" x 30')
for i in `seq 50000`; do
echo $sha refs/tags/$tag-$i >> .git/packed-refs
done
Then share this repo over the smart HTTP protocol and try cloning it:
$ git clone http://localhost/.../too-many-refs/.git
Cloning into 'too-many-refs'...
fatal: cannot exec 'fetch-pack': Argument list too long
50k tags is obviously an absurd number, but it is required to
demonstrate the problem on Linux because it has a much more generous
command line limit. On Windows the clone fails with as little as 500
tags in the above loop, which is getting uncomfortably close to the
number of tags you might see in real long lived repos.
This is not just theoretical, msysGit is already failing to clone our
company repo due to this. It's a large repo converted from CVS, nearly
10 years of history.
Four possible solutions were discussed on the Git mailing list (in no
particular order):
1) Call fetch-pack multiple times with smaller batches of refs.
This was dismissed as inefficient and inelegant.
2) Add option --refs-fd=$n to pass a an fd from where to read the refs.
This was rejected because inheriting descriptors other than
stdin/stdout/stderr through exec() is apparently problematic on Windows,
plus it would require changes to the run-command API to open extra
pipes.
3) Add option --refs-from=$tmpfile to pass the refs using a temp file.
This was not favored because of the temp file requirement.
4) Add option --stdin to pass the refs on stdin, one per line.
In the end this option was chosen as the most efficient and most
desirable from scripting perspective.
There was however a small complication when using stdin to pass refs to
fetch-pack. The --stateless-rpc option to fetch-pack also uses stdin for
communication with the remote server.
If we are going to sneak refs on stdin line by line, it would have to be
done very carefully in the presence of --stateless-rpc, because when
reading refs line by line we might read ahead too much data into our
buffer and eat some of the remote protocol data which is also coming on
stdin.
One way to solve this would be to refactor get_remote_heads() in
fetch-pack.c to accept a residual buffer from our stdin line parsing
above, but this function is used in several places so other callers
would be burdened by this residual buffer interface even when most of
them don't need it.
In the end we settled on the following solution:
If --stdin is specified without --stateless-rpc, fetch-pack would read
the refs from stdin one per line, in a script friendly format.
However if --stdin is specified together with --stateless-rpc,
fetch-pack would read the refs from stdin in packetized format
(pkt-line) with a flush packet terminating the list of refs. This way we
can read the exact number of bytes that we need from stdin, and then
get_remote_heads() can continue reading from the same fd without losing
a single byte of remote protocol data.
This way the --stdin option only loses generality and scriptability when
used together with --stateless-rpc, which is not easily scriptable
anyway because it also uses pkt-line when talking to the remote server.
Signed-off-by: Ivan Todoroski <grnch@gmx.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: fix tabbed preferences construction when using tcl 8.4
In 8.5 the incr command creates the target variable if it does not exist
but in 8.4 using incr on a non-existing variable raises an error. Ensure
we have created our counter variable when creating the tabbed dialog for
non-themed preferences.
Reported-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Portuguese Portuguese translations from Marco Sousa via Jiang Xin
* 'master' of git://github.com/git-l10n/git-po:
l10n: Add the Dutch translation team and initialize nl.po
l10n: Inital Portuguese Portugal language (pt_PT)
l10n: Improve zh_CN translation for Git 1.7.10-rc3
Make it clear that, when using commit --template, the message *must* be
changed or the commit will be aborted.
Helped-by: Junio C Hamano <gitster@pobox.com> Helped-by: Ivan Heffner <iheffner@gmail.com> Signed-off-by: Adam Monsen <haircut@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>