Some of our tests try to make sure Git behaves sensibly in a
read-only directory, by dropping 'w' permission bit before doing a
test and then restoring it after it is done. The latter is needed
for the test framework to clean after itself without leaving a
leftover directory that cannot be removed.
Ancient parts of tests however arrange the above with
chmod a-w . &&
... do the test ...
status=$?
chmod 775 .
(exit $status)
which obviously would not work if the test somehow dies before it
has the chance to do "chmod 775". Rewrite them by following a more
robust pattern recently written tests use, which is
test_when_finished "chmod 775 ." &&
chmod a-w . &&
... do the test ...
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We've recently forbidden .gitmodules to be a symlink in
verify_path(). And it's an easy way to circumvent our fsck
checks for .gitmodules content. So let's complain when we
see it.
Signed-off-by: Jeff King <peff@peff.net>
Now that the internal fsck code has all of the plumbing we
need, we can start checking incoming .gitmodules files.
Naively, it seems like we would just need to add a call to
fsck_finish() after we've processed all of the objects. And
that would be enough to cover the initial test included
here. But there are two extra bits:
1. We currently don't bother calling fsck_object() at all
for blobs, since it has traditionally been a noop. We'd
actually catch these blobs in fsck_finish() at the end,
but it's more efficient to check them when we already
have the object loaded in memory.
2. The second pass done by fsck_finish() needs to access
the objects, but we're actually indexing the pack in
this process. In theory we could give the fsck code a
special callback for accessing the in-pack data, but
it's actually quite tricky:
a. We don't have an internal efficient index mapping
oids to packfile offsets. We only generate it on
the fly as part of writing out the .idx file.
b. We'd still have to reconstruct deltas, which means
we'd basically have to replicate all of the
reading logic in packfile.c.
Instead, let's avoid running fsck_finish() until after
we've written out the .idx file, and then just add it
to our internal packed_git list.
This does mean that the objects are "in the repository"
before we finish our fsck checks. But unpack-objects
already exhibits this same behavior, and it's an
acceptable tradeoff here for the same reason: the
quarantine mechanism means that pushes will be
fully protected.
In addition to a basic push test in t7415, we add a sneaky
pack that reverses the usual object order in the pack,
requiring that index-pack access the tree and blob during
the "finish" step.
This already works for unpack-objects (since it will have
written out loose objects), but we'll check it with this
sneaky pack for good measure.
Signed-off-by: Jeff King <peff@peff.net>
As with the previous commit, we must call fsck's "finish"
function in order to catch any queued objects for
.gitmodules checks.
This second pass will be able to access any incoming
objects, because we will have exploded them to loose objects
by now.
This isn't quite ideal, because it means that bad objects
may have been written to the object database (and a
subsequent operation could then reference them, even if the
other side doesn't send the objects again). However, this is
sufficient when used with receive.fsckObjects, since those
loose objects will all be placed in a temporary quarantine
area that will get wiped if we find any problems.
Signed-off-by: Jeff King <peff@peff.net>
Now that the internal fsck code is capable of checking
.gitmodules files, we just need to teach its callers to use
the "finish" function to check any queued objects.
With this, we can now catch the malicious case in t7415 with
git-fsck.
Signed-off-by: Jeff King <peff@peff.net>
* jk/submodule-name-verify-fix:
verify_path: disallow symlinks in .gitmodules
update-index: stat updated files earlier
verify_path: drop clever fallthrough
skip_prefix: add icase-insensitive variant
is_{hfs,ntfs}_dotgitmodules: add tests
path: match NTFS short names for more .git files
is_hfs_dotgit: match other .git files
is_ntfs_dotgit: use a size_t for traversing string
submodule-config: verify submodule names as paths
Note that this includes two bits of evil-merge:
- there's a new call to verify_path() that doesn't actually
have a mode available. It should be OK to pass "0" here,
since we're just manipulating the untracked cache, not an
actual index entry.
- the lstat() in builtin/update-index.c:update_one() needs
to be updated to handle the fsmonitor case (without this
it still behaves correctly, but does an unnecessary
lstat).
This tests primarily for NTFS issues, but also adds one example of an
HFS+ issue.
Thanks go to Congyi Wu for coming up with the list of examples where
NTFS would possibly equate the filename with `.gitmodules`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Jeff King <peff@peff.net>
Submodule "names" come from the untrusted .gitmodules file,
but we blindly append them to $GIT_DIR/modules to create our
on-disk repo paths. This means you can do bad things by
putting "../" into the name (among other things).
Let's sanity-check these names to avoid building a path that
can be exploited. There are two main decisions:
1. What should the allowed syntax be?
It's tempting to reuse verify_path(), since submodule
names typically come from in-repo paths. But there are
two reasons not to:
a. It's technically more strict than what we need, as
we really care only about breaking out of the
$GIT_DIR/modules/ hierarchy. E.g., having a
submodule named "foo/.git" isn't actually
dangerous, and it's possible that somebody has
manually given such a funny name.
b. Since we'll eventually use this checking logic in
fsck to prevent downstream repositories, it should
be consistent across platforms. Because
verify_path() relies on is_dir_sep(), it wouldn't
block "foo\..\bar" on a non-Windows machine.
2. Where should we enforce it? These days most of the
.gitmodules reads go through submodule-config.c, so
I've put it there in the reading step. That should
cover all of the C code.
We also construct the name for "git submodule add"
inside the git-submodule.sh script. This is probably
not a big deal for security since the name is coming
from the user anyway, but it would be polite to remind
them if the name they pick is invalid (and we need to
expose the name-checker to the shell anyway for our
test scripts).
This patch issues a warning when reading .gitmodules
and just ignores the related config entry completely.
This will generally end up producing a sensible error,
as it works the same as a .gitmodules file which is
missing a submodule entry (so "submodule update" will
barf, but "git clone --recurse-submodules" will print
an error but not abort the clone.
There is one minor oddity, which is that we print the
warning once per malformed config key (since that's how
the config subsystem gives us the entries). So in the
new test, for example, the user would see three
warnings. That's OK, since the intent is that this case
should never come up outside of malicious repositories
(and then it might even benefit the user to see the
message multiple times).
Credit for finding this vulnerability and the proof of
concept from which the test script was adapted goes to
Etienne Stalmans.
Signed-off-by: Jeff King <peff@peff.net>
Hotfix for recently graduated topic that give help to completion
scripts from the Git subcommands that are being completed
* nd/parseopt-completion:
t9902: disable test on the list of merge-strategies under GETTEXT_POISON
completion: clear cached --options when sourcing the completion script
The code to learn the list of merge strategies from the output of
"git merge -s help" forces C locale, so that it can notice the
message shown to indicate where the list starts in the output.
However, GETTEXT_POISON build corrupts its output even when run in
the C locale, and we cannot expect this test to succeed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The established way to update the completion script in an already
running shell is to simply source it again: this brings in any new
--options and features, and clears caching variables. E.g. it clears
the variables caching the list of (all|porcelain) git commands, so
when they are later lazy-initialized again, then they will list and
cache any newly installed commmands as well.
Unfortunately, since d401f3debc (git-completion.bash: introduce
__gitcomp_builtin, 2018-02-09) and subsequent patches this doesn't
work for a lot of git commands' options. To eliminate a lot of
hard-to-maintain hard-coded lists of options, those commits changed
the completion script to use a bunch of programmatically created and
lazy-initialized variables to cache the options of those builtin
porcelain commands that use parse-options. These variables are not
cleared upon sourcing the completion script, therefore they continue
caching the old lists of options, even when some commands recently
learned new options or when deprecated options were removed.
Always 'unset' these variables caching the options of builtin commands
when sourcing the completion script.
Redirect 'unset's stderr to /dev/null, because ZSH's 'unset' complains
if it's invoked without any arguments, i.e. no variables caching
builtin's options are set. This can happen, if someone were to source
the completion script twice without completing any --options in
between. Bash stays silent in this case.
Add tests to ensure that these variables are indeed cleared when the
completion script is sourced; not just the variables caching options,
but all other caching variables, i.e. the variables caching commands,
porcelain commands and merge strategies as well.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The transfer.fsckobjects configuration tells "git fetch" to
validate the data and connected-ness of objects in the received
pack; the code to perform this check has been taught about the
narrow clone's convention that missing objects that are reachable
from objects in a pack that came from a promissor remote is OK.
* jt/transfer-fsck-with-promissor:
fetch-pack: do not check links for partial fetch
index-pack: support checking objects but not links
In a way similar to how "git tag" learned to honor the pager
setting only in the list mode, "git config" learned to ignore the
pager setting when it is used for setting values (i.e. when the
purpose of the operation is not to "show").
* ma/config-page-only-in-list-mode:
config: change default of `pager.config` to "on"
config: respect `pager.config` in list/get-mode only
t7006: add tests for how git config paginates
Allow running a couple of tests with "sh -x".
* sg/cvs-tests-with-x:
t9402-git-cvsserver-refs: don't check the stderr of a subshell
t9400-git-cvsserver-server: don't rely on the output of 'test_cmp'
When doing a partial clone or fetch with transfer.fsckobjects=1, use the
--fsck-objects instead of the --strict flag when invoking index-pack so
that links are not checked, only objects. This is because incomplete
links are expected when doing a partial clone or fetch.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The index-pack command currently supports the
--check-self-contained-and-connected argument, for internal use only,
that instructs it to only check for broken links and not broken objects.
For partial clones, we need the inverse, so add a --fsck-objects
argument that checks for broken objects and not broken links, also for
internal use only.
This will be used by fetch-pack in a subsequent patch.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach parse-options API an option to help the completion script,
and make use of the mechanism in command line completion.
* nd/parseopt-completion: (45 commits)
completion: more subcommands in _git_notes()
completion: complete --{reuse,reedit}-message= for all notes subcmds
completion: simplify _git_notes
completion: don't set PARSE_OPT_NOCOMPLETE on --rerere-autoupdate
completion: use __gitcomp_builtin in _git_worktree
completion: use __gitcomp_builtin in _git_tag
completion: use __gitcomp_builtin in _git_status
completion: use __gitcomp_builtin in _git_show_branch
completion: use __gitcomp_builtin in _git_rm
completion: use __gitcomp_builtin in _git_revert
completion: use __gitcomp_builtin in _git_reset
completion: use __gitcomp_builtin in _git_replace
remote: force completing --mirror= instead of --mirror
completion: use __gitcomp_builtin in _git_remote
completion: use __gitcomp_builtin in _git_push
completion: use __gitcomp_builtin in _git_pull
completion: use __gitcomp_builtin in _git_notes
completion: use __gitcomp_builtin in _git_name_rev
completion: use __gitcomp_builtin in _git_mv
completion: use __gitcomp_builtin in _git_merge_base
...
The "interactive.diffFilter" used by "git add -i" must retain
one-to-one correspondence between its input and output, but it was
not enforced and caused end-user confusion. We now at least make
sure the filtered result has the same number of lines as its input
to detect a broken filter.
* jk/add-i-diff-filter:
add--interactive: detect bogus diffFilter output
t3701: add a test for interactive.diffFilter
"git worktree" learned move and remove subcommands.
* nd/worktree-move:
t2028: fix minor error and issues in newly-added "worktree move" tests
worktree remove: allow it when $GIT_WORK_TREE is already gone
worktree remove: new command
worktree move: refuse to move worktrees with submodules
worktree move: accept destination as directory
worktree move: new command
worktree.c: add update_worktree_location()
worktree.c: add validate_worktree()
"git add -p" has been lazy in coalescing split patches before
passing the result to underlying "git apply", leading to corner
case bugs; the logic to prepare the patch to be applied after hunk
selections has been tightened.
* pw/add-p-recount:
add -p: don't rely on apply's '--recount' option
add -p: fix counting when splitting and coalescing
add -p: calculate offset delta for edited patches
add -p: adjust offsets of subsequent hunks when one is skipped
t3701: add failing test for pathological context lines
t3701: don't hard code sha1 hash values
t3701: use test_write_lines and write_script
t3701: indent here documents
add -i: add function to format hunk header
Running test scripts under -x option of the shell is often not a
useful way to debug them, because the error messages from the
commands tests try to capture and inspect are contaminated by the
tracing output by the shell. An earlier work done to make it more
pleasant to run tests under -x with recent versions of bash is
extended to cover posix shells that do not support BASH_XTRACEFD.
* sg/test-x:
travis-ci: run tests with '-x' tracing
t/README: add a note about don't saving stderr of compound commands
t1510-repo-setup: mark as untraceable with '-x'
t9903-bash-prompt: don't check the stderr of __git_ps1()
t5570-git-daemon: don't check the stderr of a subshell
t5526: use $TRASH_DIRECTORY to specify the path of GIT_TRACE log file
t5500-fetch-pack: don't check the stderr of a subshell
t3030-merge-recursive: don't check the stderr of a subshell
t1507-rev-parse-upstream: don't check the stderr of a shell function
t: add means to disable '-x' tracing for individual test scripts
t: prevent '-x' tracing from interfering with test helpers' stderr
"git diff" and friends learned "--compact-summary" that shows the
information usually given with the "--summary" option on the same
line as the diffstat output of the "--stat" option (which saves
vertical space and keeps info on a single path at the same place).
* nd/diff-stat-with-summary:
diff: add --compact-summary
diff.c: refactor pprint_rename() to use strbuf
Four 'cvs diff' related tests in 't9402-git-cvsserver-refs.sh' fail
when the test script is run with '-x' tracing (and using a shell other
than a Bash version supporting BASH_XTRACEFD). The reason for those
failures is that the tests check the emptiness of a subshell's stderr,
which includes the trace of commands executed in that subshell as
well, throwing off the emptiness check.
Save the stdout and stderr of the invoked 'cvs' command instead of the
whole subshell, so the latter remains free from tracing output. (Note
that changing how stdout is saved is only done for the sake of
consistency, it's not necessary for correctness.)
After this change t9402 passes with '-x', even when running with
/bin/sh.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The test 'cvs update (-p)' redirects and checks 'test_cmp's stdout and
even its stderr. The commit introducing this test in 6e8937a084
(cvsserver: Add test for update -p, 2008-03-27) doesn't discuss why,
in fact its log message only consists of that subject line. Anyway,
weird as it is, it kind of made sense due to the way that test was
structured:
After a bit of preparation, this test updates four files via CVS and
checks their contents using 'test_cmp', but it does so in a for loop
iterating over the names of those four files. Now, the exit status of
a for loop is the exit status of the last command executed in the
loop, meaning that the test can't simply rely on the exit code of
'test_cmp' in the loop's body. Instead, the test works it around by
relying on the stdout of 'test_cmp' being silent on success and
showing the diff on failure, as it appends the stdout of all four
'test_cmp' invocations to a single file and checks that file's
emptiness after the loop (with 'test -z "$(cat ...)"', no less; there
was no 'test_must_be_empty' back then). Furthermore, the test
redirects the stderr of those 'test_cmp' invocations to this file,
too: while 'test_cmp' itself doesn't output anything to stderr, the
invoked 'diff' or 'cmp' commands do send their error messages there,
e.g. if they can't open a file because its name was misspelled.
This also makes this test fail when the test script is run with '-x'
tracing (and using a shell other than a Bash version supporting
BASH_XTRACEFD), because 'test_cmp's stderr contains the trace of the
'diff' command executed inside the helper function, throwing off the
subsequent emptiness check.
Stop relying on 'test_cmp's output and instead run 'test_cmp a b ||
return 1' in the for loop in order to make 'test_cmp's error code fail
the test. Furthermore, add the missing && after the cvs command to
create a && chain in the loop's body.
After this change t9400 passes with '-x', even when running with
/bin/sh.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Writing out the index file when the only thing that changed in it
is the untracked cache information is often wasteful, and this has
been optimized out.
* bp/untracked-cache-noflush:
untracked cache: use git_env_bool() not getenv() for customization
dir.c: don't flag the index as dirty for changes to the untracked cache
"git status" can spend a lot of cycles to compute the relation
between the current branch and its upstream, which can now be
disabled with "--no-ahead-behind" option.
* jh/status-no-ahead-behind:
status: support --no-ahead-behind in long format
status: update short status to respect --no-ahead-behind
status: add --[no-]ahead-behind to status and commit for V2 format.
stat_tracking_info: return +1 when branches not equal
Since Git 1.7.9, "git merge" defaulted to --no-ff (i.e. even when
the side branch being merged is a descendant of the current commit,
create a merge commit instead of fast-forwarding) when merging a
tag object. This was appropriate default for integrators who pull
signed tags from their downstream contributors, but caused an
unnecessary merges when used by downstream contributors who
habitually "catch up" their topic branches with tagged releases
from the upstream. Update "git merge" to default to --no-ff only
when merging a tag object that does *not* sit at its usual place in
refs/tags/ hierarchy, and allow fast-forwarding otherwise, to
mitigate the problem.
* jc/allow-ff-merging-kept-tags:
merge: allow fast-forward when merging a tracked tag
The new "--show-current-patch" option gives an end-user facing way
to get the diff being applied when "git rebase" (and "git am")
stops with a conflict.
* nd/rebase-show-current-patch:
rebase: introduce and use pseudo-ref REBASE_HEAD
rebase: add --show-current-patch
am: add --show-current-patch
Clarify how configured fetch refspecs interact with the "--prune"
option of "git fetch", and also add a handy short-hand for getting
rid of stale tags that are locally held.
* ab/fetch-prune:
fetch: make the --prune-tags work with <url>
fetch: add a --prune-tags option and fetch.pruneTags config
fetch tests: add scaffolding for the new fetch.pruneTags
git-fetch & config doc: link to the new PRUNING section
git remote doc: correct dangerous lies about what prune does
git fetch doc: add a new section to explain the ins & outs of pruning
fetch tests: fetch <url> <spec> as well as fetch [<remote>]
fetch tests: expand case/esac for later change
fetch tests: double quote a variable for interpolation
fetch tests: test --prune and refspec interaction
fetch tests: add a tag to be deleted to the pruning tests
fetch tests: re-arrange arguments for future readability
fetch tests: refactor in preparation for testing tag pruning
remote: add a macro for "refs/tags/*:refs/tags/*"
fetch: stop accessing "remote" variable indirectly
fetch: trivially refactor assignment to ref_nr
fetch: don't redundantly NULL something calloc() gave us
"git tag" learned an explicit "--edit" option that allows the
message given via "-m" and "-F" to be further edited.
* nm/tag-edit:
tag: add --edit option
Recently-added "git worktree move" tests include a minor error and a few
small issues. Specifically:
* checking non-existence of wrong file ("source" instead of
"destination")
* unneeded redirect (">empty")
* unused variable ("toplevel")
* restoring a worktree location by means of a separate test somewhat
distant from the test which moved it rather than using
test_when_finished() to restore it in a self-contained fashion
* having git command on the left-hand-side of a pipe ("git foo | grep")
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Acked-by: Duy Nguyen <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In some projects contributions from groups are only accepted from a
common group email address. But every individual may want to receive
replies to her own personal address. That's what we have 'Reply-To'
headers for in SMTP. So introduce an optional '--reply-to' command
line option.
This patch re-uses the $reply_to variable. This could break
out-of-tree patches!
Signed-off-by: Christian Ludwig <chrissicool@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's important that the diff-filter only filter the
individual lines, and that there remain a one-to-one mapping
between the input and output lines. Otherwise, things like
hunk-splitting will behave quite unexpectedly (e.g., you
think you are splitting at one point, but it has a different
effect in the text patch we apply).
We can't detect all problematic cases, but we can at least
catch the obvious case where we don't even have the correct
number of lines.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This feature was added in 01143847db (add--interactive:
allow custom diff highlighting programs, 2016-02-27) but
never tested. Let's add a basic test.
Note that we only apply the filter when color is enabled,
so we have to use test_terminal. This is an open limitation
explicitly mentioned in the original commit. So take this
commit as testing the status quo, and not making a statement
on whether we'd want to enhance that in the future.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a file has no trailing new line at the end diff records this by
appending "\ No newline at end of file" below the last line of the
file. This line should not be counted in the hunk header. Fix the
splitting and coalescing code to count files without a trailing new line
properly and change one of the tests to test splitting without a
trailing new line.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Recount the number of preimage and postimage lines in a hunk after it
has been edited so any change in the number of insertions or deletions
can be used to adjust the offsets of subsequent hunks. If an edited
hunk is subsequently split then the offset correction will be lost. It
would be possible to fix this if it is a problem, however the code
here is still an improvement on the status quo for the common case
where an edited hunk is applied without being split.
This is also a necessary step to removing '--recount' and
'--allow-overlap' from the invocation of 'git apply'. Before
'--recount' can be removed the splitting and coalescing counting needs
to be fixed to handle a missing newline at the end of a file. In order
to remove '--allow-overlap' there needs to be i) some way of verifying
the offset data in the edited hunk (probably by correlating the
preimage (or postimage if the patch is going to be applied in reverse)
lines of the edited and unedited versions to see if they are offset or
if any leading/trailing context lines have been removed) and ii) a way of
dealing with edited hunks that change context lines that are shared
with neighbouring hunks.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This adds xfuncname and word_regex patterns for golang, a quite
popular programming language. It also includes test cases for the
xfuncname regex (t4018) and updated documentation.
The xfuncname regex finds functions, structs and interfaces. Although
the Go language prohibits the opening brace from being on its own
line, the regex does not makes it mandatory, to be able to match
`func` statements like this:
func foo(bar int,
baz int) {
}
This is covered by the test case t4018/golang-long-func.
The word_regex pattern finds identifiers, integers, floats, complex
numbers and operators, according to the go specification.
Signed-off-by: Alban Gruin <alban.gruin@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commit 8cbd431082 ("git-add--interactive: replace hunk
recounting with apply --recount", 2008-7-2) if a hunk is skipped then
we rely on the context lines to apply subsequent hunks in the right
place. While this works most of the time it is possible for hunks to
end up being applied in the wrong place. To fix this adjust the offset
of subsequent hunks to correct for any change in the number of
insertions or deletions due to the skipped hunk. The change in offset
due to edited hunks that have the number of insertions or deletions
changed is ignored here, it will be fixed in the next commit.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a hunk is skipped by add -i the offsets of subsequent hunks are
not adjusted to account for any missing insertions due to the skipped
hunk. Most of the time this does not matter as apply uses the context
lines to apply the subsequent hunks in the correct place, however in
pathological cases the context lines will match at the now incorrect
offset and the hunk will be applied in the wrong place. The offsets of
hunks following an edited hunk that has had the number of insertions
or deletions changed also need to be updated in the same way. Add
failing tests to demonstrate this.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
