Earlier, if you changed a staged path into a directory in the work tree,
we happily ran lstat(2) on it and found that it exists, and declared that
the user changed it to a gitlink.
This is wrong for two reasons:
(1) It may be a directory, but it may not be a submodule, and in the
latter case, the change we need to report is "the blob at the path
has disappeared". We need to check with resolve_gitlink_ref() to be
consistent with what "git add" and "git update-index --add" does.
(2) lstat(2) may have succeeded only because a leading component of the
path was turned into a symbolic link that points at something that
exists in the work tree. In such a case, the path itself does not
exist anymore, as far as the index is concerned.
This fixes these breakages in diff-index that the previous patch has
exposed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff-index and diff-files can get confused in corner cases when an indexed
blob turns into something else in the work tree. This patch adds tests to
expose such breakages.
The test is classified under t2XXX series instead of t4XXX series, because
the ultimate objective is to fix "add -u" (and "commit -a" that shares the
same issue).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Finally, this resurrects the documented behaviour to protect other
objects listed on the command line from getting pruned.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It seems that git prune changed behaviour with respect to revisions added
from command line, probably when it became a builtin. Currently, it prints
a short usage and exits: instead, it should take those revisions into
account and not prune them. So add a couple of test to point this out.
We'll be fixing this by switching to parse_options(), so add tests to
detect bogus command line parameters as well, to keep ourselves from
introducing regressions.
Signed-off-by: Michele Ballabio <barra_cuda@katamail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
Update draft release notes for 1.5.4.5
Documentation: clarify use of .git{ignore,attributes} versus .git/info/*
t/t3800-mktag.sh: use test_must_fail rather than '!'
Conflicts:
t/t3800-mktag.sh
When a git command is run under test_must_fail to make sure that
the argument parser catches bogus command line, it exits with 129.
We need to catch it as a valid "graceful error exit".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This actually sounds like a bug in cvsps, which requires an existing
home directory when asked for the usage through -h
$ HOME=/nonexistent cvsps -h
Cannot create the cvsps directory '.cvsps': No such file or directory
This made t9600 think that cvsps is not available if HOME did not exist,
causing the tests to be skipped
$ HOME=/nonexistent sh t9600-cvsimport.sh
* skipping cvsimport tests, cvsps not found
* passed all 0 test(s)
Now t9600 sets HOME to the current working directory before checking for
the availability of the cvsps program.
This issue has been discovered by Marco Rodrigues, and fixed by Frank
Lichtenheld through
http://bugs.debian.org/471969
Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jc/maint-fetch-regression-1.5.4:
git-fetch test: test tracking fetch results, not just FETCH_HEAD
Fix branches file configuration
Tighten refspec processing
Fix the wrong output of `git-show v1.3.0~155^2~4` in documentation.
We really should have done this long time ago. Existing t5515 test
was written for the specific purpose of catching regression to the
contents of generated FETCH_HEAD file, but it also is a good place
to make sure various fetch configurations do fetch what they intend
to fetch (and nothing else).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This changes the pattern matching code to not store the required final
/ before the *, and then to require each side to be a valid ref (or
empty). In particular, any refspec that looks like it should be a
pattern but doesn't quite meet the requirements will be found to be
invalid as a fallback non-pattern.
This was cherry picked from commit ef00d15 (Tighten refspec processing,
2008-03-17), and two fix-up commits 46220ca (remote.c: Fix overtight
refspec validation, 2008-03-20) and 7d19da4 (refspec: allow colon-less
wildcard "refs/category/*", 2008-03-25) squashed in.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Earlier, git-init tested for a valid HEAD ref, but if the repository
was empty, there was none. Instead, test for the existence of
the file $GIT_DIR/HEAD.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Without this the output of 'git remote show' does not end with a new-line:
bash> git remote show repo
* remote repo
URL: repo.or.cz:/srv/git/kdbg.git
Tracked remote branches
maint master mob
Local branch pushed with 'git push'
+master:masterbash>
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We tightened the refspec validation code in an earlier commit ef00d15
(Tighten refspec processing, 2008-03-17) per my suggestion, but the
suggestion was misguided to begin with and it broke this usage:
$ git push origin HEAD~12:master
The syntax of push refspecs and fetch refspecs are similar in that they
are both colon separated LHS and RHS (possibly prefixed with a + to
force), but the similarity ends there. For example, LHS in a push refspec
can be anything that evaluates to a valid object name at runtime (except
when colon and RHS is missing, or it is a glob), while it must be a
valid-looking refname in a fetch refspec. To validate them correctly, the
caller needs to be able to say which kind of refspecs they are. It is
unreasonable to keep a single interface that cannot tell which kind it is
dealing with, and ask it to behave sensibly.
This commit separates the parsing of the two into different functions, and
clarifies the code to implement the parsing proper (i.e. splitting into
two parts, making sure both sides are wildcard or neither side is).
This happens to also allow pushing a commit named with the esoteric "look
for that string" syntax:
$ git push ../test.git ':/remote.c: Fix overtight refspec:master'
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Give lib-git-svn.sh a few alternate paths to look for apache2.
Explicitly define the LockFile so httpd will actually start under OS X
Signed-off-by: Kevin Ballard <kevin@sb.org>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-svn project names are percent-escaped ever since f5530b8
(git-svn: support for funky branch and project names over HTTP(S),
2007-11-11).
Unfortunately this breaks the scenario where the user hands git-svn an
already-escaped URI. Fix the regexp to skip over what looks like
existing percent escapes, and test this scenario.
Signed-off-by: Kevin Ballard <kevin@sb.org>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
For symbolic refs, a sane notion of being "stale" is that the ref
they point to no longer exists. Since this is checked already,
"remote show" does not need to show them at all.
Incidentally, this fixes the issue that "HEAD" was shown as a
stale ref by "remote show" in a freshly cloned repository.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The revision limiter uses the commit date to decide when it has seen
enough commits to finalize the revision list, but that can get confused
if there are incorrect dates far in the past on some commits.
This makes the logic a bit more robust by
- we always walk an extra SLOP commits from the source list even if we
decide that the source list is probably all done (unless the source is
entirely empty, of course, because then we really can't do anything at
all)
- we keep track of the date of the last commit we added to the
destination list (this will *generally* be the oldest entry we've seen
so far)
- we compare that with the youngest entry (the first one) of the source
list, and if the destination is older than the source, we know we want
to look at the source.
which causes occasional date mishaps to be handled cleanly.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git-config name = value" doesn't do anything most of the time. The
test meant "git-config name value", but that leaves the configuration
such that later tests will be confused, so move it to the end.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This currently fails not because we refuse to check out, but because we
detect error but incorrectly discard it in the callchain.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When your test creates an unwritable directory that test framework cannot
clean out by "rm -fr trash", later tests cannot start in a fresh state
they expect to. Detect this and error out early.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a merge result creates a new file, and when our side already has a
file in the path, taking the merge result may clobber the untracked file.
However, the logic to detect this situation was totally the wrong way. We
should complain when the file exists, not when the file does not exist.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In commit 34110cd4e3 ("Make 'unpack_trees()'
have a separate source and destination index") I introduced a really
stupid bug in that it would always add merged entries with the CE_UPDATE
flag set. That caused us to always re-write the file, even when it was
already up-to-date in the source index.
Not only is that really stupid from a performance angle, but more
importantly it's actively wrong: if we have dirty state in the tree when
we merge, overwriting it with the result of the merge will incorrectly
overwrite that dirty state.
This trivially fixes the problem - simply don't set the CE_UPDATE flag
when the merge result matches the old state.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We had a handful test updates since we accepted 82ebb0b (add test_cmp
function for test scripts). This fixes them up.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jk/portable:
t6000lib: re-fix tr portability
t7505: use SHELL_PATH in hook
t9112: add missing #!/bin/sh header
filter-branch: use $SHELL_PATH instead of 'sh'
filter-branch: don't use xargs -0
add NO_EXTERNAL_GREP build option
t6000lib: tr portability fix
t4020: don't use grep -a
add test_cmp function for test scripts
remove use of "tail -n 1" and "tail -1"
grep portability fix: don't use "-e" or "-q"
more tr portability test script fixes
t0050: perl portability fix
tr portability fixes
Once upon a time shortlog could be run from a non-git directory
and still do its job. Fix this regression and add a small test
for it.
Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Earlier, the callchain from pretty_print_commit() down to pp_title_line()
had an unwarranted assumption that the presense of "after_subject"
parameter, means the caller has already output MIME headers for
attachments. The parameter's primary purpose is to give extra header
lines the caller wants to place after pp_title_line() generates the
"Subject: " line.
This assumption does not hold when the user used the format.header
configuration variable to pass extra headers, and caused a message with
non-ASCII character to lack proper MIME headers (e.g. 8-bit CTE header).
The earlier logic also failed to suppress duplicated MIME headers when
"format-patch -s --attach" is asked for and the signer's name demanded
8-bit clean transport.
This patch fixes the logic by introducing a separate need_8bit_cte
parameter passed down the callchain. This can have one of these values:
-1 : we've already done MIME crap and we do not want to add extra header
to say this is 8bit in pp_title_line();
0 : we haven't done MIME and we have not seen anything that is 8bit yet;
1 : we haven't done MIME and we have seen something that is 8bit;
pp_title_line() must add MIME header.
It adds two tests by Jeff King who independently diagnosed this issue.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It seems that some implementations of tr don't like a
replacement string of '-----...'; they try to find the
double-dash option "---...".
Instead of this pipeline of tr and sed invocations, just use a
single perl invocation.
Signed-off-by: Jeff King <peff@peff.net>
* maint:
merge-file: handle empty files gracefully
merge-recursive: handle file mode changes
Minor wording changes in the keyboard descriptions in git-add --interactive.
git fetch: Take '-n' to mean '--no-tags'
quiltimport: fix misquoting of parsed -p<num> parameter
git-quiltimport: better parser to grok "enhanced" series files.
File mode changes should be handled similarly to changes of content.
That is, if the file mode changed in only one branch, keep the changed
version, and if both branch changed to different mode, mark it as a
conflict.
Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The hook doesn't run properly under Solaris /bin/sh. Let's
use the SHELL_PATH the user told us about already instead.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some versions of tr complain if the number of characters in
both sets isn't the same. So here we must manually expand
the dashes in set2.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Solaris /usr/bin/grep doesn't understand "-a". In this case
we can just include the expected output with the test, which
is a better test anyway.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many scripts compare actual and expected output using
"diff -u". This is nicer than "cmp" because the output shows
how the two differ. However, not all versions of diff
understand -u, leading to unnecessary test failure.
This adds a test_cmp function to the test scripts and
switches all "diff -u" invocations to use it. The function
uses the contents of "$GIT_TEST_CMP" to compare its
arguments; the default is "diff -u".
On systems with a less-capable diff, you can do:
GIT_TEST_CMP=cmp make test
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "-n" syntax is not supported by System V versions of
tail (which prefer "tail -1"). Unfortunately "tail -1" is
not actually POSIX. We had some of both forms in our
scripts.
Since neither form works everywhere, this patch replaces
both with the equivalent sed invocation:
sed -ne '$p'
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
System V versions of grep (such as Solaris /usr/bin/grep)
don't understand either of these options. git's usage of
"grep -e pattern" fell into one of two categories:
1. equivalent to "grep pattern". -e is only useful here if
the pattern begins with a "-", but all of the patterns
are hardcoded and do not begin with a dash.
2. stripping comments and blank lines with
grep -v -e "^$" -e "^#"
We can fortunately do this in the affirmative as
grep '^[^#]'
Uses of "-q" can be replaced with redirection to /dev/null.
In many tests, however, "grep -q" is used as "if this string
is in the expected output, we are OK". In this case, it is
fine to just remove the "-q" entirely; it simply makes the
"verbose" mode of the test slightly more verbose.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Dealing with NULs is not always safe with tr. On Solaris,
incoming NULs are silently deleted by both the System V and
UCB versions of tr. When converting to NULs, the System V
version works fine, but the UCB version silently ignores the
request to convert the character.
This patch changes all instances of tr using NULs to use
"perl -pe 'y///'" instead.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Older versions of perl (such as 5.005) don't understand -CO, nor
do they understand the "U" pack specifier. Instead of using perl,
let's just printf the binary bytes we are interested in.
Signed-off-by: Jeff King <peff@peff.net>
Tested-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The only reason we did not call "prune" in git-gc was that it is an
inherently dangerous operation: if there is a commit going on, you will
prune loose objects that were just created, and are, in fact, needed by the
commit object just about to be created.
Since it is dangerous, we told users so. That led to many users not even
daring to run it when it was actually safe. Besides, they are users, and
should not have to remember such details as when to call git-gc with
--prune, or to call git-prune directly.
Of course, the consequence was that "git gc --auto" gets triggered much
more often than we would like, since unreferenced loose objects (such as
left-overs from a rebase or a reset --hard) were never pruned.
Alas, git-prune recently learnt the option --expire <minimum-age>, which
makes it a much safer operation. This allows us to call prune from git-gc,
with a grace period of 2 weeks for the unreferenced loose objects (this
value was determined in a discussion on the git list as a safe one).
If you want to override this grace period, just set the config variable
gc.pruneExpire to a different value; an example would be
[gc]
pruneExpire = 6.months.ago
or even "never", if you feel really paranoid.
Note that this new behaviour makes "--prune" be a no-op.
While adding a test to t5304-prune.sh (since it really tests the implicit
call to "prune"), also the original test for "prune --expire" was moved
there from t1410-reflog.sh, where it did not belong.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Specifying character ranges in tr differs between System V
and POSIX. In System V, brackets are required (e.g.,
'[A-Z]'), whereas in POSIX they are not.
We can mostly get around this by just using the bracket form
for both sets, as in:
tr '[A-Z] '[a-z]'
in which case POSIX interpets this as "'[' becomes '['",
which is OK.
However, this doesn't work with multiple sequences, like:
# rot13
tr '[A-Z][a-z]' '[N-Z][A-M][n-z][a-m]'
where the POSIX version does not behave the same as the
System V version. In this case, we must simply enumerate the
sequence.
This patch fixes problematic uses of tr in git scripts and
test scripts in one of three ways:
- if a single sequence, make sure it uses brackets
- if multiple sequences, enumerate
- if extra brackets (e.g., tr '[A]' 'a'), eliminate
brackets
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* js/remote:
"remote update": print remote name being fetched from
builtin remote rm: remove symbolic refs, too
remote: fix "update [group...]"
remote show: Clean up connection correctly if object fetch wasn't done
builtin-remote: prune remotes correctly that were added with --mirror
Make git-remote a builtin
Test "git remote show" and "git remote prune"
parseopt: add flag to stop on first non option
path-list: add functions to work with unsorted lists
Conflicts:
parse-options.c
