We should reject a patch, whether it renames/copies dir/file to
elsewhere with or without modificiation, or updates dir/file in
place, if "dir/" part is actually a symbolic link to elsewhere,
by making sure that the code to read the preimage does not read
from a path that is beyond a symbolic link.
apply: do not read from the filesystem under --index
We currently read the preimage to apply a patch from the index only
when the --cached option is given. Do so also when the command is
running under the --index option. With --index, the index entry and
the working tree file for a path that is involved in a patch must be
identical, so this should not affect the result, but by reading from
the index, we will get the protection to avoid reading an unintended
path beyond a symbolic link automatically.
apply: reject input that touches outside the working area
By default, a patch that affects outside the working area (either a
Git controlled working tree, or the current working directory when
"git apply" is used as a replacement of GNU patch) is rejected as a
mistake (or a mischief). Git itself does not create such a patch,
unless the user bends over backwards and specifies a non-standard
prefix to "git diff" and friends.
When `git apply` is used as a "better GNU patch", the user can pass
the `--unsafe-paths` option to override this safety check. This
option has no effect when `--index` or `--cached` is in use.
The new test was stolen from Jeff King with slight enhancements.
Note that a few new tests for touching outside the working area by
following a symbolic link are still expected to fail at this step,
but will be fixed in later steps.
is_hfs_dotgit: loosen over-eager match of \u{..47}
Our is_hfs_dotgit function relies on the hackily-implemented
next_hfs_char to give us the next character that an HFS+
filename comparison would look at. It's hacky because it
doesn't implement the full case-folding table of HFS+; it
gives us just enough to see if the path matches ".git".
At the end of next_hfs_char, we use tolower() to convert our
32-bit code point to lowercase. Our tolower() implementation
only takes an 8-bit char, though; it throws away the upper
24 bits. This means we can't have any false negatives for
is_hfs_dotgit. We only care about matching 7-bit ASCII
characters in ".git", and we will correctly process 'G' or
'g'.
However, we _can_ have false positives. Because we throw
away the upper bits, code point \u{0147} (for example) will
look like 'G' and get downcased to 'g'. It's not known
whether a sequence of code points whose truncation ends up
as ".git" is meaningful in any language, but it does not
hurt to be more accurate here. We can just pass out the full
32-bit code point, and compare it manually to the upper and
lowercase characters we care about.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the index can block pathnames that can be mistaken
to mean ".git" on NTFS and FAT32, it would be helpful for
fsck to notice such problematic paths. This lets servers
which use receive.fsckObjects block them before the damage
spreads.
Note that the fsck check is always on, even for systems
without core.protectNTFS set. This is technically more
restrictive than we need to be, as a set of users on ext4
could happily use these odd filenames without caring about
NTFS.
However, on balance, it's helpful for all servers to block
these (because the paths can be used for mischief, and
servers which bother to fsck would want to stop the spread
whether they are on NTFS themselves or not), and hardly
anybody will be affected (because the blocked names are
variants of .git or git~1, meaning mischief is almost
certainly what the tree author had in mind).
Ideally these would be controlled by a separate
"fsck.protectNTFS" flag. However, it would be much nicer to
be able to enable/disable _any_ fsck flag individually, and
any scheme we choose should match such a system. Given the
likelihood of anybody using such a path in practice, it is
not unreasonable to wait until such a system materializes.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of disallowing ".git" in the index is that we
would never want to accidentally overwrite files in the
repository directory. But this means we need to respect the
filesystem's idea of when two paths are equal. The prior
commit added a helper to make such a comparison for NTFS
and FAT32; let's use it in verify_path().
We make this check optional for two reasons:
1. It restricts the set of allowable filenames, which is
unnecessary for people who are not on NTFS nor FAT32.
In practice this probably doesn't matter, though, as
the restricted names are rather obscure and almost
certainly would never come up in practice.
2. It has a minor performance penalty for every path we
insert into the index.
This patch ties the check to the core.protectNTFS config
option. Though this is expected to be most useful on Windows,
we allow it to be set everywhere, as NTFS may be mounted on
other platforms. The variable does default to on for Windows,
though.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow paths with a ".git" component to be added to
the index, as that would mean repository contents could
overwrite our repository files. However, asking "is this
path the same as .git" is not as simple as strcmp() on some
filesystems.
On NTFS (and FAT32), there exist so-called "short names" for
backwards-compatibility: 8.3 compliant names that refer to the same files
as their long names. As ".git" is not an 8.3 compliant name, a short name
is generated automatically, typically "git~1".
Depending on the Windows version, any combination of trailing spaces and
periods are ignored, too, so that both "git~1." and ".git." still refer
to the Git directory. The reason is that 8.3 stores file names shorter
than 8 characters with trailing spaces. So literally, it does not matter
for the short name whether it is padded with spaces or whether it is
shorter than 8 characters, it is considered to be the exact same.
The period is the separator between file name and file extension, and
again, an empty extension consists just of spaces in 8.3 format. So
technically, we would need only take care of the equivalent of this
regex:
(\.git {0,4}|git~1 {0,3})\. {0,3}
However, there are indications that at least some Windows versions might
be more lenient and accept arbitrary combinations of trailing spaces and
periods and strip them out. So we're playing it real safe here. Besides,
there can be little doubt about the intention behind using file names
matching even the more lenient pattern specified above, therefore we
should be fine with disallowing such patterns.
Extra care is taken to catch names such as '.\\.git\\booh' because the
backslash is marked as a directory separator only on Windows, and we want
to use this new helper function also in fsck on other platforms.
A big thank you goes to Ed Thomson and an unnamed Microsoft engineer for
the detailed analysis performed to come up with the corresponding fixes
for libgit2.
This commit adds a function to detect whether a given file name can refer
to the Git directory by mistake.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the index can block pathnames that case-fold to
".git" on HFS+, it would be helpful for fsck to notice such
problematic paths. This lets servers which use
receive.fsckObjects block them before the damage spreads.
Note that the fsck check is always on, even for systems
without core.protectHFS set. This is technically more
restrictive than we need to be, as a set of users on ext4
could happily use these odd filenames without caring about
HFS+.
However, on balance, it's helpful for all servers to block
these (because the paths can be used for mischief, and
servers which bother to fsck would want to stop the spread
whether they are on HFS+ themselves or not), and hardly
anybody will be affected (because the blocked names are
variants of .git with invisible Unicode code-points mixed
in, meaning mischief is almost certainly what the tree
author had in mind).
Ideally these would be controlled by a separate
"fsck.protectHFS" flag. However, it would be much nicer to
be able to enable/disable _any_ fsck flag individually, and
any scheme we choose should match such a system. Given the
likelihood of anybody using such a path in practice, it is
not unreasonable to wait until such a system materializes.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of disallowing ".git" in the index is that we
would never want to accidentally overwrite files in the
repository directory. But this means we need to respect the
filesystem's idea of when two paths are equal. The prior
commit added a helper to make such a comparison for HFS+;
let's use it in verify_path.
We make this check optional for two reasons:
1. It restricts the set of allowable filenames, which is
unnecessary for people who are not on HFS+. In practice
this probably doesn't matter, though, as the restricted
names are rather obscure and almost certainly would
never come up in practice.
2. It has a minor performance penalty for every path we
insert into the index.
This patch ties the check to the core.protectHFS config
option. Though this is expected to be most useful on OS X,
we allow it to be set everywhere, as HFS+ may be mounted on
other platforms. The variable does default to on for OS X,
though.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow paths with a ".git" component to be added to
the index, as that would mean repository contents could
overwrite our repository files. However, asking "is this
path the same as .git" is not as simple as strcmp() on some
filesystems.
HFS+'s case-folding does more than just fold uppercase into
lowercase (which we already handle with strcasecmp). It may
also skip past certain "ignored" Unicode code points, so
that (for example) ".gi\u200ct" is mapped ot ".git".
The full list of folds can be found in the tables at:
Implementing a full "is this path the same as that path"
comparison would require us importing the whole set of
tables. However, what we want to do is much simpler: we
only care about checking ".git". We know that 'G' is the
only thing that folds to 'g', and so on, so we really only
need to deal with the set of ignored code points, which is
much smaller.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We complain about ".git" in a tree because it cannot be
loaded into the index or checked out. Since we now also
reject ".GIT" case-insensitively, fsck should notice the
same, so that errors do not propagate.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not allow ".git" to enter into the index as a path
component, because checking out the result to the working
tree may causes confusion for subsequent git commands.
However, on case-insensitive file systems, ".Git" or ".GIT"
is the same. We should catch and prevent those, too.
Note that technically we could allow this for repos on
case-sensitive filesystems. But there's not much point. It's
unlikely that anybody cares, and it creates a repository
that is unexpectedly non-portable to other systems.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
read-tree: add tests for confusing paths like ".." and ".git"
We should prevent nonsense paths from entering the index in
the first place, as they can cause confusing results if they
are ever checked out into the working tree. We already do
so, but we never tested it.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
unpack-trees: propagate errors adding entries to the index
When unpack_trees tries to write an entry to the index,
add_index_entry may report an error to stderr, but we ignore
its return value. This leads to us returning a successful
exit code for an operation that partially failed. Let's make
sure to propagate this code.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: fix missing text for rev-parse --verify
The caret (^) is used as a markup symbol in AsciiDoc. Due to the
inability of AsciiDoc to parse a line containing an unmatched caret, it
omitted the line from the output, resulting in the man page missing the
end of a sentence. Escape this caret so that the man page ends up with
the complete text.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Simplify the code and get rid of some magic constants by using
argv_array to build the argument list for cmd_blame. Be lazy and let
the OS release our allocated memory, as before.
Signed-off-by: Rene Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the middle step fails but leaves the directory (e.g., the
bug is that clean does not notice the failure), this
pollutes the test repo with an unremovable directory. Not
only does this cause further tests to fail, but it means
that "rm -rf" fails on the whole trash directory, and the
user has to intervene manually to even re-run the test script.
We can bump the "chmod 755" recovery to a test_when_finished
block to be sure that it always runs.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'bm/merge-base-octopus-dedup' into maint
"git merge-base --octopus" used to leave cleaning up suboptimal
result to the caller, but now it does the clean-up itself.
* bm/merge-base-octopus-dedup:
merge-base --octopus: reduce the result from get_octopus_merge_bases()
merge-base: separate "--independent" codepath into its own helper
Merge branch 'jc/revision-range-unpeel' into maint
"git log --left-right A...B" lost the "leftness" of commits
reachable from A when A is a tag as a side effect of a recent
bugfix. This is a regression in 1.8.4.x series.
* jc/revision-range-unpeel:
revision: propagate flag bits from tags to pointees
revision: mark contents of an uninteresting tree uninteresting
Merge branch 'jk/allow-fetch-onelevel-refname' into maint
"git clone" would fail to clone from a repository that has a ref
directly under "refs/", e.g. "refs/stash", because different
validation paths do different things on such a refname. Loosen the
client side's validation to allow such a ref.
* jk/allow-fetch-onelevel-refname:
fetch-pack: do not filter out one-level refs
Merge branch 'jk/interpret-branch-name-fix' into maint
A handful of bugs around interpreting $branch@{upstream} notation
and its lookalike, when $branch part has interesting characters,
e.g. "@", and ":", have been fixed.
* jk/interpret-branch-name-fix:
interpret_branch_name: find all possible @-marks
interpret_branch_name: avoid @{upstream} past colon
interpret_branch_name: always respect "namelen" parameter
interpret_branch_name: rename "cp" variable to "at"
interpret_branch_name: factor out upstream handling
Merge branch 'as/tree-walk-fix-aggressive-short-cut' into maint
The pathspec matching code, while comparing two trees (e.g. "git
diff A B -- path1 path2") was too aggressive and failed to match
some paths when multiple pathspecs were involved.
* as/tree-walk-fix-aggressive-short-cut:
tree_entry_interesting: match against all pathspecs
Merge branch 'jh/rlimit-nofile-fallback' into maint
When we figure out how many file descriptors to allocate for
keeping packfiles open, a system with non-working getrlimit() could
cause us to die(), but because we make this call only to get a
rough estimate of how many is available and we do not even attempt
to use up all file descriptors available ourselves, it is nicer to
fall back to a reasonable low value rather than dying.
* jh/rlimit-nofile-fallback:
get_max_fd_limit(): fall back to OPEN_MAX upon getrlimit/sysconf failure
Merge branch 'jl/commit-v-strip-marker' into maint
"git commit -v" appends the patch to the log message before
editing, and then removes the patch when the editor returned
control. However, the patch was not stripped correctly when the
first modified path was a submodule.
* jl/commit-v-strip-marker:
commit -v: strip diffs and submodule shortlogs from the commit message
SSL-related options were not passed correctly to underlying socket
layer in "git send-email".
* tr/send-email-ssl:
send-email: set SSL options through IO::Socket::SSL::set_client_defaults
send-email: --smtp-ssl-cert-path takes an argument
send-email: pass Debug to Net::SMTP::SSL::new
Merge branch 'tb/clone-ssh-with-colon-for-port' into maint
Remote repository URL expressed in scp-style host:path notation are
parsed more carefully (e.g. "foo/bar:baz" is local, "[::1]:/~user" asks
to connect to user's home directory on host at address ::1.
* tb/clone-ssh-with-colon-for-port:
git_connect(): use common return point
connect.c: refactor url parsing
git_connect(): refactor the port handling for ssh
git fetch: support host:/~repo
t5500: add test cases for diag-url
git fetch-pack: add --diag-url
git_connect: factor out discovery of the protocol and its parts
git_connect: remove artificial limit of a remote command
t5601: add tests for ssh
t5601: remove clear_ssh, refactor setup_ssh_wrapper
tree_entry_interesting: match against all pathspecs
The current basedir compare aborts early in order to avoid futile
recursive searches. However, a match may still be found by another
pathspec. This can cause an error while checking out files from a branch
when using multiple pathspecs:
$ git checkout master -- 'a/*.txt' 'b/*.txt'
error: pathspec 'a/*.txt' did not match any file(s) known to git.
Signed-off-by: Andy Spencer <andy753421@gmail.com> Acked-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the original shell version of git-repack, any options
destined for pack-objects were left as strings, and passed
as a whole. Since the C rewrite in commit a1bbc6c (repack:
rewrite the shell script in C, 2013-09-15), we now parse
these values to integers internally, then reformat the
integers when passing the option to pack-objects.
This has the advantage that we catch format errors earlier
(i.e., when repack is invoked, rather than when pack-objects
is invoked).
It has three disadvantages, though:
1. Our internal data types may not be the right size. In
the case of "--window-memory" and "--max-pack-size",
these are "unsigned long" in pack-objects, but we can
only represent a regular "int".
2. Our parsing routines might not be the same as those of
pack-objects. For the two options above, pack-objects
understands "100m" to mean "100 megabytes", but repack
does not.
3. We have to keep a sentinel value to know whether it is
worth passing the option along. In the case of
"--window-memory", we currently do not pass it if the
value is "0". But that is a meaningful value to
pack-objects, where it overrides any configured value.
We can fix all of these by simply passing the strings from
the user along to pack-objects verbatim. This does not
actually fix anything for "--depth" or "--window", but these
are converted, too, for consistency.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we use OPT_STRING to parse an option, we get back a
pointer into the argv array, which should be "const char *".
The compiler doesn't notice because it gets passed through a
"void *" in the option struct.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Subversion serf backend in versions 1.8.5 and below has a bug(*) that the
function creating the descriptor of a file change -- add_file() --
doesn't make a copy of its third argument when storing it on the
returned descriptor. As a result, by the time this field is used (in
transactions of file copying or renaming) it may well be released, and
the memory reused.
One of its possible manifestations is the svn assertion triggering on an
invalid path, with a message
This patch works around this bug, by storing the value to be passed as
the third argument to add_file() in a local variable with the same scope
as the file change descriptor, making sure their lifetime is the same.
* [ew: fixed in Subversion r1553376 as noted by Jonathan Nieder]
Cc: Benjamin Pabst <benjamin.pabst85@gmail.com> Signed-off-by: Eric Wong <normalperson@yhbt.net> Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Roman Kagan <rkagan@mail.ru>
STARTTLS failed! SSL connect attempt failed with unknown error
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate
verify failed at /usr/libexec/git-core/git-send-email line 1236.
The current code detects the presence of /etc/ssl/certs directory
(it actually is a symlink to another directory, but that does not
matter) and uses SSL_ca_path to point at it when initializing the
connection with IO::Socket::SSL or Net::SMTP::SSL. However, on the
said platform, it seems that this directory is not designed to be
used as SSL_ca_path. Using a single file inside that directory
(cert.pem, which is a Mozilla CA bundle) with SSL_ca_file does work,
and also not specifying any SSL_ca_file/SSL_ca_path (and letting the
library use its own default) and asking for peer verification does
work.
By removing the code that blindly defaults $smtp_ssl_cert_path to
"/etc/ssl/certs", we can prevent the codepath that treats any
directory specified with that variable as usable for SSL_ca_path
from incorrectly triggering.
This change could introduce a regression for people on a platform
whose certificate directory is /etc/ssl/certs but its IO::Socket:SSL
somehow fails to use it as SSL_ca_path without being told. Using
/etc/ssl/certs directory as SSL_ca_path by default like the current
code does would have been hiding such a broken installation without
its user needing to do anything. These users can still work around
such a platform bug by setting the configuration variable explicitly
to point at /etc/ssl/certs.
This change should not negate what 35035bbf (send-email: be explicit
with SSL certificate verification, 2013-07-18), which was the
original change that introduced the defaulting to /etc/ssl/certs/,
attempted to do, which is to make sure we do not communicate over
insecure connection by default, triggering warning from the library.
Tested-by: Igor Gnatenko <i.gnatenko.brain@gmail.com> Signed-off-by: Ruben Kerkhof <ruben@rubenkerkhof.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
revision: propagate flag bits from tags to pointees
With the previous fix 895c5ba3 (revision: do not peel tags used in
range notation, 2013-09-19), handle_revision_arg() that processes
command line arguments for the "git log" family of commands no
longer directly places the object pointed by the tag in the pending
object array when it sees a tag object. We used to place pointee
there after copying the flag bits like UNINTERESTING and
SYMMETRIC_LEFT.
This change meant that any flag that is relevant to later history
traversal must now be propagated to the pointed objects (most often
these are commits) while starting the traversal, which is partly
done by handle_commit() that is called from prepare_revision_walk().
We did propagate UNINTERESTING, but did not do so for others, most
notably SYMMETRIC_LEFT. This caused "git log --left-right v1.0..."
(where "v1.0" is a tag) to start losing the "leftness" from the
commit the tag points at.
revision: mark contents of an uninteresting tree uninteresting
"git rev-list --objects ^A^{tree} B^{tree}" ought to mean "I want a
list of objects inside B's tree, but please exclude the objects that
appear inside A's tree".
we see the top-level tree marked as uninteresting (i.e. ^A^{tree} in
the above example) and call mark_tree_uninteresting() on it; this
unfortunately prevents us from recursing into the tree and marking
the objects in the tree as uninteresting.
The reason why "git log ^A A" yields an empty set of commits,
i.e. we do not have a similar issue for commits, is because we call
mark_parents_uninteresting() after seeing an uninteresting commit.
The uninteresting-ness of the commit itself does not prevent its
parents from being marked as uninteresting.
Introduce mark_tree_contents_uninteresting() and structure the code
in handle_commit() in such a way that it makes it the responsibility
of the callchain leading to this function to mark commits, trees and
blobs as uninteresting, and also make it the responsibility of the
helpers called from this function to mark objects that are reachable
from them.
Note that this is a very old bug that probably dates back to the day
when "rev-list --objects" was introduced. The line to clear
tree->object.parsed at the end of mark_tree_contents_uninteresting()
can be removed when this fix is merged to the codebase after 6e454b9a (clear parsed flag when we free tree buffers, 2013-06-05).
When we parse a string like "foo@{upstream}", we look for
the first "@"-sign, and check to see if it is an upstream
mark. However, since branch names can contain an @, we may
also see "@foo@{upstream}". In this case, we check only the
first @, and ignore the second. As a result, we do not find
the upstream.
We can solve this by iterating through all @-marks in the
string, and seeing if any is a legitimate upstream or
empty-at mark.
Another strategy would be to parse from the right-hand side
of the string. However, that does not work for the
"empty_at" case, which allows "@@{upstream}". We need to
find the left-most one in this case (and we then recurse as
"HEAD@{upstream}").
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
interpret_branch_name: avoid @{upstream} past colon
get_sha1() cannot currently parse a valid object name like
"HEAD:@{upstream}" (assuming that such an oddly named file
exists in the HEAD commit). It takes two passes to parse the
string:
1. It first considers the whole thing as a ref, which
results in looking for the upstream of "HEAD:".
2. It finds the colon, parses "HEAD" as a tree-ish, and then
finds the path "@{upstream}" in the tree.
For a path that looks like a normal reflog (e.g.,
"HEAD:@{yesterday}"), the first pass is a no-op. We try to
dwim_ref("HEAD:"), that returns zero refs, and we proceed
with colon-parsing.
For "HEAD:@{upstream}", though, the first pass ends up in
interpret_upstream_mark, which tries to find the branch
"HEAD:". When it sees that the branch does not exist, it
actually dies rather than returning an error to the caller.
As a result, we never make it to the second pass.
One obvious way of fixing this would be to teach
interpret_upstream_mark to simply report "no, this isn't an
upstream" in such a case. However, that would make the
error-reporting for legitimate upstream cases significantly
worse. Something like "bogus@{upstream}" would simply report
"unknown revision: bogus@{upstream}", while the current code
diagnoses a wide variety of possible misconfigurations (no
such branch, branch exists but does not have upstream, etc).
However, we can take advantage of the fact that a branch
name cannot contain a colon. Therefore even if we find an
upstream mark, any prefix with a colon must mean that
the upstream mark we found is actually a pathname, and
should be disregarded completely. This patch implements that
logic.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
interpret_branch_name gets passed a "name" buffer to parse,
along with a "namelen" parameter representing its length. If
"namelen" is zero, we fallback to the NUL-terminated
string-length of "name".
However, it does not necessarily follow that if we have
gotten a non-zero "namelen", it is the NUL-terminated
string-length of "name". E.g., when get_sha1() is parsing
"foo:bar", we will be asked to operate only on the first
three characters.
Yet in interpret_branch_name and its helpers, we use string
functions like strchr() to operate on "name", looking past
the length we were given. This can result in us mis-parsing
object names. We should instead be limiting our search to
"namelen" bytes.
There are three distinct types of object names this patch
addresses:
- The intrepret_empty_at helper uses strchr to find the
next @-expression after our potential empty-at. In an
expression like "@:foo@bar", it erroneously thinks that
the second "@" is relevant, even if we were asked only
to look at the first character. This case is easy to
trigger (and we test it in this patch).
- When finding the initial @-mark for @{upstream}, we use
strchr. This means we might treat "foo:@{upstream}" as
the upstream for "foo:", even though we were asked only
to look at "foo". We cannot test this one in practice,
because it is masked by another bug (which is fixed in
the next patch).
- The interpret_nth_prior_checkout helper did not receive
the name length at all. This turns out not to be a
problem in practice, though, because its parsing is so
limited: it always starts from the far-left of the
string, and will not tolerate a colon (which is
currently the only way to get a smaller-than-strlen
"namelen"). However, it's still worth fixing to make the
code more obviously correct, and to future-proof us
against callers with more exotic buffers.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
interpret_branch_name: rename "cp" variable to "at"
In the original version of this function, "cp" acted as a
pointer to many different things. Since the refactoring in
the last patch, it only marks the at-sign in the string.
Let's use a more descriptive variable name.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
interpret_branch_name: factor out upstream handling
This function checks a few different @{}-constructs. The
early part checks for and dispatches us to helpers for each
construct, but the code for handling @{upstream} is inline.
Let's factor this out into its own function. This makes
interpret_branch_name more readable, and will make it much
simpler to further refactor the function in future patches.
While we're at it, let's also break apart the refactored
code into a few helper functions. These will be useful if we
eventually implement similar @{upstream}-like constructs.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently fetching a one-level ref like "refs/foo" does not
work consistently. The outer "git fetch" program filters the
list of refs, checking each against check_refname_format.
Then it feeds the result to do_fetch_pack to actually
negotiate the haves/wants and get the pack. The fetch-pack
code does its own filter, and it behaves differently.
The fetch-pack filter looks for refs in "refs/", and then
feeds everything _after_ the slash (i.e., just "foo") into
check_refname_format. But check_refname_format is not
designed to look at a partial refname. It complains that the
ref has only one component, thinking it is at the root
(i.e., alongside "HEAD"), when in reality we just fed it a
partial refname.
As a result, we omit a ref like "refs/foo" from the pack
request, even though "git fetch" then tries to store the
resulting ref. If we happen to get the object anyway (e.g.,
because the ref is contained in another ref we are
fetching), then the fetch succeeds. But if it is a unique
object, we fail when trying to update "refs/foo".
We can fix this by just passing the whole refname into
check_refname_format; we know the part we were omitting is
"refs/", which is acceptable in a refname. This at least
makes the checks consistent with each other.
This problem happens most commonly with "refs/stash", which
is the only one-level ref in wide use. However, our test
does not use "refs/stash", as we may later want to restrict
it specifically (not because it is one-level, but because
of the semantics of stashes).
We may also want to do away with the multiple levels of
filtering (which can cause problems when they are out of
sync), or even forbid one-level refs entirely. However,
those decisions can come later; this fixes the most
immediate problem, which is the mismatch between the two.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: "git pull" does not have the "-m" option
Even though "--[no-]edit" can be used with "git pull", the
explanation of the interaction between this option and the "-m"
option does not make sense within the context of "git pull". Use
the conditional inclusion mechanism to remove this part from "git
pull" documentation, while keeping it for "git merge".
Reported-by: Ivan Zakharyaschev Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: exclude irrelevant options from "git pull"
10eb64f5 (git pull manpage: don't include -n from fetch-options.txt,
2008-01-25) introduced a way to exclude some parts of included
source when building git-pull documentation, and later 409b8d82
(Documentation/git-pull: put verbosity options before merge/fetch
ones, 2010-02-24) attempted to use the mechanism to exclude some
parts of merge-options.txt when used from git-pull.txt.
However, the latter did not have an intended effect, because the
macro "git-pull" used to decide if the source is included in
git-pull documentation were defined a bit too late.
mv: let 'git mv file no-such-dir/' error out on Windows, too
The previous commit c57f628 (mv: let 'git mv file no-such-dir/' error out)
relies on that rename("file", "no-such-dir/") fails if the directory does not
exist (note the trailing slash). This does not work as expected on Windows:
This rename() call does not fail, but renames "file" to "no-such-dir" (not to
"no-such-dir/file"). Insert an explicit check for this case to force an error.
This changes the error message from
$ git mv file no-such-dir/
fatal: renaming 'file' failed: Not a directory
to
$ git mv file no-such-dir/
fatal: destination directory does not exist, source=file, destination=no-such-dir/
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When trying to pop/apply a stash specified with an argument
containing IFS whitespace, git-stash will throw an error:
$ git stash pop 'stash@{two hours ago}'
Too many revisions specified: stash@{two hours ago}
This happens because word splitting is used to count non-option
arguments. Make use of rev-parse's --sq option to quote the arguments
for us to ensure a correct count. Add quotes where necessary.
Also add a test that verifies correct behaviour.
Helped-by: Thomas Rast <tr@thomasrast.ch> Signed-off-by: Øystein Walle <oystwa@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/gitmodules: Only 'update' and 'url' are required
Descriptions for all the settings fell under the initial "Each
submodule section also contains the following required keys:". The
example shows sections with just 'path' and 'url' entries, which are
indeed required, but we should still make the required/optional
distinction explicit to clarify that the rest of them are optional.
Signed-off-by: W. Trevor King <wking@tremily.us> Reviewed-by: Heiko Voigt <hvoigt@hvoigt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 64a99eb4 git gc refuses to run without the --force option if
another gc process on the same repository is already running.
However, if the repository is shared and user A runs git gc on the
repository and while that gc is still running user B runs git gc on
the same repository the gc process run by user A will not be noticed
and the gc run by user B will go ahead and run.
The problem is that the kill(pid, 0) test fails with an EPERM error
since user B is not allowed to signal processes owned by user A
(unless user B is root).
Update the test to recognize an EPERM error as meaning the process
exists and another gc should not be run (unless --force is given).
Signed-off-by: Kyle J. McKay <mackyle@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
That commit taught do_askpass to hand ownership of our
buffer back to the caller rather than simply return a
pointer into our internal strbuf. What it failed to notice,
though, was that our internal strbuf is static, because we
are trying to emulate the getpass() interface.
By handing off ownership, we created a memory leak that
cannot be solved. Sometimes git_prompt returns a static
buffer from getpass() (or our smarter git_terminal_prompt
wrapper), and sometimes it returns an allocated string from
do_askpass.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
merge-base --octopus: reduce the result from get_octopus_merge_bases()
Scripts that use "merge-base --octopus" could do the reducing
themselves, but most of them are expected to want to get the reduced
results without having to do any work themselves.
Tests are taken from a message by Василий Макаров
<einmalfel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
We might want to vet the existing callers of the underlying
get_octopus_merge_bases() and find out if _all_ of them are doing
anything extra (like deduping) because the machinery can return
duplicate results. And if that is the case, then we may want to
move the dedupling down the callchain instead of having it here.
merge-base: separate "--independent" codepath into its own helper
It piggybacks on an unrelated handle_octopus() function only because
there are some similarities between the way they need to preprocess
their input and output their result. There is nothing similar in
the true logic between these two operations.
Support for grafts predates Git's strbuf, and hence it is understandable
that there was a hard-coded line length limit of 1023 characters (which
was chosen a bit awkwardly, given that it is *exactly* one byte short of
aligning with the 41 bytes occupied by a commit name and the following
space or new-line character).
While regular commit histories hardly win comprehensibility in general
if they merge more than twenty-two branches in one go, it is not Git's
business to limit grafts in such a way.
In this particular developer's case, the use case that requires
substantially longer graft lines to be supported is the visualization of
the commits' order implied by their changes: commits are considered to
have an implicit relationship iff exchanging them in an interactive
rebase would result in merge conflicts.
Thusly implied branches tend to be very shallow in general, and the
resulting thicket of implied branches is usually very wide; It is
actually quite common that *most* of the commits in a topic branch have
not even one implied parent, so that a final merge commit has about as
many implied parents as there are commits in said branch.
[jc: squashed in tests by Jonathan]
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
add: don't complain when adding empty project root
This behavior was added in 07d7bed (add: don't complain when adding
empty project root - 2009-04-28) then broken by 84b8b5d (remove
match_pathspec() in favor of match_pathspec_depth() -
2013-07-14). Reinstate it.
Noticed-by: Thomas Ferris Nicolaisen <tfnico@gmail.com> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
log: properly handle decorations with chained tags
git log did not correctly handle decorations when a tag object referenced
another tag object that was no longer a ref, such as when the second tag was
deleted. The commit would not be decorated correctly because parse_object had
not been called on the second tag and therefore its tagged field had not been
filled in, resulting in none of the tags being associated with the relevant
commit.
Call parse_object to fill in this field if it is absent so that the chain of
tags can be dereferenced and the commit can be properly decorated. Include
tests as well to prevent future regressions.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
get_max_fd_limit(): fall back to OPEN_MAX upon getrlimit/sysconf failure
On broken systems where RLIMIT_NOFILE is visible by the compliers
but underlying getrlimit() system call does not behave, we used to
simply die() when we are trying to decide how many file descriptors
to allocate for keeping packfiles open. Instead, allow the fallback
codepath to take over when we get such a failure from getrlimit().
The same issue exists with _SC_OPEN_MAX and sysconf(); restructure
the code in a similar way to prepare for a broken sysconf() as well.
Noticed-by: Joey Hess <joey@kitenet.net> Helped-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'jh/loose-object-dirs-creation-race' into maint
Two processes creating loose objects at the same time could have
failed unnecessarily when the name of their new objects started
with the same byte value, due to a race condition.
* jh/loose-object-dirs-creation-race:
sha1_file.c:create_tmpfile(): Fix race when creating loose object dirs
Merge branch 'jk/two-way-merge-corner-case-fix' into maint
"git am --abort" sometimes complained about not being able to write
a tree with an 0{40} object in it.
* jk/two-way-merge-corner-case-fix:
t1005: add test for "read-tree --reset -u A B"
t1005: reindent
unpack-trees: fix "read-tree -u --reset A B" with conflicted index
cmd_repack(): remove redundant local variable "nr_packs"
Its value is the same as the number of entries in the "names"
string_list, so just use "names.nr" in its place.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Acked-by: Stefan Beller <stefanbeller@googlemail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cat-file: handle --batch format with missing type/size
Commit 98e2092 taught cat-file to stream blobs with --batch,
which requires that we look up the object type before
loading it into memory. As a result, we now print the
object header from information in sha1_object_info, and the
actual contents from the read_sha1_file. We double-check
that the information we printed in the header matches the
content we are about to show.
Later, commit 93d2a60 allowed custom header lines for
--batch, and commit 5b08640 made type lookups optional. As a
result, specifying a header line without the type or size
means that we will not look up those items at all.
This causes our double-checking to erroneously die with an
error; we think the type or size has changed, when in fact
it was simply left at "0".
For the size, we can fix this by only doing the consistency
double-check when we have retrieved the size via
sha1_object_info. In the case that we have not retrieved the
value, that means we also did not print it, so there is
nothing for us to check that we are consistent with.
We could do the same for the type. However, besides our
consistency check, we also care about the type in deciding
whether to stream or not. So instead of handling the case
where we do not know the type, this patch instead makes sure
that we always trigger a type lookup when we are printing,
so that even a format without the type will stream as we
would in the normal case.
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We currently individually pass the sha1, type, and size
fields calculated by sha1_object_info. However, if we pass
the whole struct, the called function can make more
intelligent decisions about which fields were actually
filled by sha1_object_info.
This patch takes that first refactoring step, passing the
whole struct, so further patches can make those decisions
with less noise in their diffs. There should be no
functional change to this patch (aside from a minor typo fix
in the error message).
As a side effect, we can rename the local variables in the
function to "type" and "size", since the names are no longer
taken.
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use only one return point from git_connect(), doing the
free();
return conn;
only at one place in the code.
There may be a little confusion what the variable "host" is for. At
some places it is only the host part, at other places it may include
the port number, so change host into hostandport here.
Signed-off-by: Torsten Bögershausen <tboegi@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>