The credential helper protocol was designed to be very flexible: the
fields it takes as input are treated as a pattern, and any missing
fields are taken as wildcards. This allows unusual things like:
echo protocol=https | git credential reject
to delete all stored https credentials (assuming the helpers themselves
treat the input that way). But when helpers are invoked automatically by
Git, this flexibility works against us. If for whatever reason we don't
have a "host" field, then we'd match _any_ host. When you're filling a
credential to send to a remote server, this is almost certainly not what
you want.
Prevent this at the layer that writes to the credential helper. Add a
check to the credential API that the host and protocol are always passed
in, and add an assertion to the credential_write function that speaks
credential helper protocol to be doubly sure.
There are a few ways this can be triggered in practice:
- the "git credential" command passes along arbitrary credential
parameters it reads from stdin.
- until the previous patch, when the host field of a URL is empty, we
would leave it unset (rather than setting it to the empty string)
- a URL like "example.com/foo.git" is treated by curl as if "http://"
was present, but our parser sees it as a non-URL and leaves all
fields unset
- the recent fix for URLs with embedded newlines blanks the URL but
otherwise continues. Rather than having the desired effect of
looking up no credential at all, many helpers will return _any_
credential
Our earlier test for an embedded newline didn't catch this because it
only checked that the credential was cleared, but didn't configure an
actual helper. Configuring the "verbatim" helper in the test would show
that it is invoked (it's obviously a silly helper which doesn't look at
its input, but the point is that it shouldn't be run at all). Since
we're switching this case to die(), we don't need to bother with a
helper. We can see the new behavior just by checking that the operation
fails.
We'll add new tests covering partial input as well (these can be
triggered through various means with url-parsing, but it's simpler to
just check them directly, as we know we are covered even if the url
parser changes behavior in the future).
[jn: changed to die() instead of logging and showing a manual
username/password prompt]
Reported-by: Carlo Arenas <carenas@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
We may feed a URL like "cert:///path/to/cert.pem" into the credential
machinery to get the key for a client-side certificate. That
credential has no hostname field, which is about to be disallowed (to
avoid confusion with protocols where a helper _would_ expect a
hostname).
This means as of the next patch, credential helpers won't work for
unlocking certs. Let's fix that by doing two things:
- when we parse a url with an empty host, set the host field to the
empty string (asking only to match stored entries with an empty
host) rather than NULL (asking to match _any_ host).
- when we build a cert:// credential by hand, similarly assign an
empty string
It's the latter that is more likely to impact real users in practice,
since it's what's used for http connections. But we don't have good
infrastructure to test it.
The url-parsing version will help anybody using git-credential in a
script, and is easy to test.
Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Many of the tests in t0300 give partial inputs to git-credential,
omitting a protocol or hostname. We're checking only high-level things
like whether and how helpers are invoked at all, and we don't care about
specific hosts. However, in preparation for tightening up the rules
about when we're willing to run a helper, let's start using input that's
a bit more realistic: pretend as if http://example.com is being
examined.
This shouldn't change the point of any of the tests, but do note we have
to adjust the expected output to accommodate this (filling a credential
will repeat back the protocol/host fields to stdout, and the helper
debug messages and askpass prompt will change on stderr).
Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
We test a toy credential helper that writes "quit=1" and confirms that
we stop running other helpers. However, that helper is unrealistic in
that it does not bother to read its stdin at all.
For now we don't send any input to it, because we feed git-credential a
blank credential. But that will change in the next patch, which will
cause this test to racily fail, as git-credential will get SIGPIPE
writing to the helper rather than exiting because it was asked to.
Let's make this one-off helper more like our other sample helpers, and
have it source the "dump" script. That will read stdin, fixing the
SIGPIPE problem. But it will also write what it sees to stderr. We can
make the test more robust by checking that output, which confirms that
we do run the quit helper, don't run any other helpers, and exit for the
reason we expected.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Add new method in transport-helper to reject all references if any
reference is failed for atomic push.
This method is reused in "send-pack.c" and "transport-helper.c", one for
SSH, git and file protocols, and the other for HTTP protocol.
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit v2.22.0-1-g3bca1e7f9f (transport-helper: enforce atomic in
push_refs_with_push, 2019-07-11) noticed the incomplete report of
failure of an atomic push for HTTP protocol. But the implementation
has a flaw that mark all remote references as failure.
Only mark necessary references as failure in `push_refs_with_push()` of
transport-helper.
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When pushing with SSH or other smart protocol, references are validated
by function `check_to_send_update()` before they are sent in commands
to `send_pack()` of "receve-pack". For atomic push, if a reference is
rejected after the validation, only references pushed by user should be
marked as failure, instead of report failure on all remote references.
Commit v2.22.0-1-g3bca1e7f9f (transport-helper: enforce atomic in
push_refs_with_push, 2019-07-11) wanted to fix report issue of HTTP
protocol, but marked all remote references failure for atomic push.
In order to fix the issue of status report for SSH or other built-in
smart protocol, revert part of that commit and add additional status
for function `atomic_push_failure()`. The additional status for it
except the "REF_STATUS_EXPECTING_REPORT" status are:
- REF_STATUS_NONE : Not marked as "REF_STATUS_EXPECTING_REPORT" yet.
- REF_STATUS_OK : Assume OK for dryrun or status_report is disabled.
This fix won't resolve the issue of status report in transport-helper
for HTTP or other protocols, and breaks test case in t5541. Will fix
it in additional commit.
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we push some references to the git server, we expect git to report
the status of the references we are pushing; no more, no less. But when
pusing with atomic mode, if some references cannot be pushed, Git reports
the reject message on all references in the remote repository.
Add new test cases in t5543, and fix them in latter commit.
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The porcelain output of a failed `git-push` command is inconsistent for
different protocols. For example, the following `git-push` command
may fail due to the failure of the `pre-receive` hook.
git push --porcelain origin HEAD:refs/heads/master
For SSH protocol, the porcelain output does not end with a "Done"
message:
To <URL/of/upstream.git>
! HEAD:refs/heads/master [remote rejected] (pre-receive hook declined)
While for HTTP protocol, the porcelain output does end with a "Done"
message:
To <URL/of/upstream.git>
! HEAD:refs/heads/master [remote rejected] (pre-receive hook declined)
Done
The following code at the end of function `send_pack()` indicates that
`send_pack()` should not return an error if some references are rejected
in porcelain mode.
int send_pack(...)
... ...
if (args->porcelain)
return 0;
for (ref = remote_refs; ref; ref = ref->next) {
switch (ref->status) {
case REF_STATUS_NONE:
case REF_STATUS_UPTODATE:
case REF_STATUS_OK:
break;
default:
return -1;
}
}
return 0;
}
So if atomic push failed, must check the porcelain mode before return
an error. And `receive_status()` should not return an error for a
failed updated reference, because `send_pack()` will check them instead.
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add missing 'restore' and 'switch' sub commands to zsh completion
candidate output. E.g.
$ git re<tab>
rebase -- forward-port local commits to the updated upstream head
reset -- reset current HEAD to the specified state
restore -- restore working tree files
$ git s<tab>
show -- show various types of objects
status -- show the working tree status
switch -- switch branches
Signed-off-by: Terry Moschou <tmoschou@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The changed-path Bloom filters help reduce the amount of tree
parsing required during history queries. Before calculating a
diff, we can ask the filter if a path changed between a commit
and its first parent. If the filter says "no" then we can move
on without parsing trees. If the filter says "maybe" then we
parse trees to discover if the answer is actually "yes" or "no".
When computing a blame, there is a section in find_origin() that
computes a diff between a commit and one of its parents. When this
is the first parent, we can check the Bloom filters before calling
diff_tree_oid().
In order to make this work with the blame machinery, we need to
initialize a struct bloom_key with the initial path. But also, we
need to add more keys to a list if a rename is detected. We then
check to see if _any_ of these keys answer "maybe" in the diff.
During development, I purposefully left out this "add a new key
when a rename is detected" to see if the test suite would catch
my error. That is how I discovered the issues with
GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS from the previous change.
With that change, we can feel some confidence in the coverage of
this change.
If a user requests copy detection using "git blame -C", then there
are more places where the set of "important" files can expand. I
do not know enough about how this happens in the blame machinery.
Thus, the Bloom filter integration is explicitly disabled in this
mode. A later change could expand the bloom_key data with an
appropriate call (or calls) to add_bloom_key().
If we did not disable this mode, then the following tests would
fail:
t8003-blame-corner-cases.sh
t8011-blame-split-file.sh
Generally, this is a performance enhancement and should not
change the behavior of 'git blame' in any way. If a repo has a
commit-graph file with computed changed-path Bloom filters, then
they should notice improved performance for their 'git blame'
commands.
Here are some example timings that I found by blaming some paths
in the Linux kernel repository:
git blame arch/x86/kernel/topology.c >/dev/null
Before: 0.83s
After: 0.24s
git blame kernel/time/time.c >/dev/null
Before: 0.72s
After: 0.24s
git blame tools/perf/ui/stdio/hist.c >/dev/null
Before: 0.27s
After: 0.11s
I specifically looked for "deep" paths that were also edited many
times. As a counterpoint, the MAINTAINERS file was edited many
times but is located in the root tree. This means that the cost of
computing a diff relative to the pathspec is very small. Here are
the timings for that command:
git blame MAINTAINERS >/dev/null
Before: 20.1s
After: 18.0s
These timings are the best of five. The worst-case runs were on the
order of 2.5 minutes for both cases. Note that the MAINTAINERS file
has 18,740 lines across 17,000+ commits. This happens to be one of
the cases where this change provides the least improvement.
The lack of improvement for the MAINTAINERS file and the relatively
modest improvement for the other examples can be easily explained.
The blame machinery needs to compute line-level diffs to determine
which lines were changed by each commit. That makes up a large
proportion of the computation time, and this change does not
attempt to improve on that section of the algorithm. The
MAINTAINERS file is large and changed often, so it takes time to
determine which lines were updated by which commit. In contrast,
the code files are much smaller, and it takes longer to comute
the line-by-line diff for a single patch on the Linux mailing
lists.
Outside of the "-C" integration, I believe there is little more to
gain from the changed-path Bloom filters for 'git blame' after this
patch.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The GIT_TEST_COMMIT_GRAPH environment variable updates the commit-
graph file whenever "git commit" is run, ensuring that we always
have an updated commit-graph throughout the test suite. The
GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS environment variable was
introduced to write the changed-path Bloom filters whenever "git
commit-graph write" is run. However, the GIT_TEST_COMMIT_GRAPH
trick doesn't launch a separate process and instead writes it
directly.
To expand the number of tests that have commits in the commit-graph
file, add a helper method that computes the commit-graph and place
that helper inside "git commit" and "git merge".
In the helper method, check GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS
to ensure we are writing changed-path Bloom filters whenever
possible.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The changed-path Bloom filters work only when we can compute an
explicit Bloom filter key in advance. When a pathspec is given
that allows case-insensitive checks or wildcard matching, we
must disable the Bloom filter performance checks.
By checking the pathspec in prepare_to_use_bloom_filters(), we
avoid setting up the Bloom filter data and thus revert to the
usual logic.
Before this change, the following tests would fail*:
t6004-rev-list-path-optim.sh (Tests 6-7)
t6130-pathspec-noglob.sh (Tests 3-6)
t6131-pathspec-icase.sh (Tests 3-5)
*These tests would fail when using GIT_TEST_COMMIT_GRAPH and
GIT_TEST_COMMIT_GRAPH_BLOOM_FILTERS except that the latter
environment variable was not set up correctly to write the changed-
path Bloom filters in the test suite. That will be fixed in the
next change.
Helped-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To help pinpoint the source of a regression, it is useful to know some
info about the compiler which the user's Git client was built with. By
adding a generic get_compiler_info() in 'compat/' we can choose which
relevant information to share per compiler; to get started, let's
demonstrate the version of glibc if the user built with 'gcc'.
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
Helped-by: Đoàn Trần Công Danh <congdanhqx@gmail.com>
Helped-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The contents of uname() can give us some insight into what sort of
system the user is running on, and help us replicate their setup if need
be. The domainname field is not guaranteed to be available, so don't
collect it.
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Knowing which version of Git a user has and how it was built allows us
to more precisely pin down the circumstances when a certain issue
occurs, so teach bugreport how to tell us the same output as 'git
version --build-options'.
It's not ideal to directly call 'git version --build-options' because
that output goes to stdout. Instead, wrap the version string in a helper
within help.[ch] library, and call that helper from within the bugreport
library.
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach Git how to prompt the user for a good bug report: reproduction
steps, expected behavior, and actual behavior. Later, Git can learn how
to collect some diagnostic information from the repository.
If users can send us a well-written bug report which contains diagnostic
information we would otherwise need to ask the user for, we can reduce
the number of question-and-answer round trips between the reporter and
the Git contributor.
Users may also wish to send a report like this to their local "Git
expert" if they have put their repository into a state they are confused
by.
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Starting in 3ac68a93fd, help.o began to depend on builtin/branch.o,
builtin/clean.o, and builtin/config.o. This meant that help.o was
unusable outside of the context of the main Git executable.
To make help.o usable by other commands again, move list_config_help()
into builtin/help.c (where it makes sense to assume other builtin libraries
are present).
When command-list.h is included but a member is not used, we start to
hear a compiler warning. Since the config list is generated in a fairly
different way than the command list, and since commands and config
options are semantically different, move the config list into its own
header and move the generator into its own script and build rule.
For reasons explained in 976aaedc (msvc: add a Makefile target to
pre-generate the Visual Studio solution, 2019-07-29), some build
artifacts we consider non-source files cannot be generated in the
Visual Studio environment, and we already have some Makefile tweaks
to help Visual Studio to use generated command-list.h header file.
Do the same to a new generated file, config-list.h, introduced by
this change.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Emily Shaffer <emilyshaffer@google.com>
In 'git log', the --decorate-refs-exclude option appends a pattern
to a string_list. This list is used to prevent showing some refs
in the decoration output, or even by --simplify-by-decoration.
Users may want to use their refs space to store utility refs that
should not appear in the decoration output. For example, Scalar [1]
runs a background fetch but places the "new" refs inside the
refs/scalar/hidden/<remote>/* refspace instead of refs/<remote>/*
to avoid updating remote refs when the user is not looking. However,
these "hidden" refs appear during regular 'git log' queries.
A similar idea to use "hidden" refs is under consideration for core
Git [2].
Add the 'log.excludeDecoration' config option so users can exclude
some refs from decorations by default instead of needing to use
--decorate-refs-exclude manually. The config value is multi-valued
much like the command-line option. The documentation is careful to
point out that the config value can be overridden by the
--decorate-refs option, even though --decorate-refs-exclude would
always "win" over --decorate-refs.
Since the 'log.excludeDecoration' takes lower precedence to
--decorate-refs, and --decorate-refs-exclude takes higher
precedence, the struct decoration_filter needed another field.
This led also to new logic in load_ref_decorations() and
ref_filter_match().
There are several tests in t4202-log.sh that test the
--decorate-refs-(include|exclude) options, so these are extended.
Since the expected output is already stored as a file, most tests
could simply replace a "--decorate-refs-exclude" option with an
in-line config setting. Other tests involve the precedence of
the config option compared to command-line options and needed more
modification.
[1] https://github.com/microsoft/scalar
[2] https://lore.kernel.org/git/77b1da5d3063a2404cd750adfe3bb8be9b6c497d.1585946894.git.gitgitgadget@gmail.com/
Helped-by: Junio C Hamano <gister@pobox.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The ref_filter_match() method is defined in refs.h and implemented
in refs.c, but is only used by add_ref_decoration() in log-tree.c.
Move it into that file as a static helper method. The
match_ref_pattern() comes along for the ride.
While moving the code, also make a slight adjustment to have
ref_filter_match() take a struct decoration_filter pointer instead
of multiple string lists. This is non-functional, but will make a
later change be much cleaner.
The diff is easier to parse when using the --color-moved option.
Reported-by: Junio C Hamano <gister@pobox.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "mboxrd" pretty format was introduced in 9f23e04061 (pretty: support
"mboxrd" output format, 2016-06-05) but wasn't mentioned in the
documentation.
Signed-off-by: Emma Brooks <me@pluvano.com>
Acked-by: Eric Wong <e@80x24.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the supplied integer for "precision" is negative in
`"%.*s", len, line` then it is ignored. So the current code is
equivalent to just `"%s", line` because it is executed only if
`len` is negative.
Fix this by saving the value of `len` before overwriting it with the
return value of `parse_git_diff_header()`.
Signed-off-by: Vasil Dimov <vd@FreeBSD.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
`git range-diff` calls `git log` internally and tries to parse its
output. But `git log` output can be customized by the user in their
git config and for certain configurations either an error will be
returned by `git range-diff` or it will crash.
To fix this explicitly set the output format of the internally
executed `git log` with `--pretty=medium`. Because that cancels
`--notes`, add explicitly `--notes` at the end.
Also, make sure we never crash in the same way - trying to dereference
`util` which was never created and has remained NULL. It would happen
if the first line of `git log` output does not begin with 'commit '.
Alternative considered but discarded - somehow disable all git configs
and behave as if no config is present in the internally executed
`git log`, but that does not seem to be possible. GIT_CONFIG_NOSYSTEM
is the closest to it, but even with that we would still read
`.git/config`.
Signed-off-by: Vasil Dimov <vd@FreeBSD.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's unusual to see:
https://example.com?query-parameters
without an intervening slash, like:
https://example.com/some-path?query-parameters
or even:
https://example.com/?query-parameters
but it is a valid end to the hostname (actually "authority component")
according to RFC 3986. Likewise for "#".
And curl will parse the URL according to the standard, meaning it will
contact example.com, but our credential code would ask about a bogus
hostname with a "?" in it. Let's make sure we follow the standard, and
more importantly ask about the same hosts that curl will be talking to.
It would be nice if we could just ask curl to parse the URL for us. But
it didn't grow a URL-parsing API until 7.62, so we'd be stuck with
fallback code either way. Plus we'd need this code in the main Git
binary, where we've tried to avoid having a link dependency on libcurl.
But let's at least fix our parser. Moving to curl's parser would prevent
other potential discrepancies, but this gives us immediate relief for
the known problem, and would help our fallback code if we eventually use
curl.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update freshen_file() to use a NULL `times', semantically equivalent to
the currently setup, with an explicit `actime' and `modtime' set to the
"current time", but with the advantage that it works with other files
not owned by the current user.
Fixes an issue on shared repos with a split index, where eventually a
user's operation creates a shared index, and another user will later do
an operation that will try to update its freshness, but will instead
raise a warning:
$ git status
warning: could not freshen shared index '.git/sharedindex.bd736fa10e0519593fefdb2aec253534470865b2'
Signed-off-by: Luciano Miguel Ferreira Rocha <luciano.rocha@booking.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When operating on a stream of commit OIDs on stdin, 'git commit-graph
write' checks that each OID refers to an object that is indeed a commit.
This is convenient to make sure that the given input is well-formed, but
can sometimes be undesirable.
For example, server operators may wish to feed the refnames that were
updated during a push to 'git commit-graph write --input=stdin-commits',
and silently discard refs that don't point at commits. This can be done
by combing the output of 'git for-each-ref' with '--format
%(*objecttype)', but this requires opening up a potentially large number
of objects. Instead, it is more convenient to feed the updated refs to
the commit-graph machinery, and let it throw out refs that don't point
to commits.
Introduce '--[no-]check-oids' to make such a behavior possible. With
'--check-oids' (the default behavior to retain backwards compatibility),
'git commit-graph write' will barf on a non-commit line in its input.
With 'no-check-oids', such lines will be silently ignored, making the
above possible by specifying this option.
No matter which is supplied, 'git commit-graph write' retains the
behavior from the previous commit of rejecting non-OID inputs like
"HEAD" and "refs/heads/foo" as before.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'write_commit_graph()' function takes in either a string list of
pack indices, or a string list of hexadecimal commit OIDs. These
correspond to the '--stdin-packs' and '--stdin-commits' mode(s) from
'git commit-graph write'.
Using a string_list of hexadecimal commit IDs is not the most efficient
use of memory, since we can instead use the 'struct oidset', which is
more well-suited for this case.
This has another benefit which will become apparent in the following
commit. This is that we are about to disambiguate the kinds of errors we
produce with '--stdin-commits' into "non-hex input" and "hex-input, but
referring to a non-commit object". By having 'write_commit_graph' take
in a 'struct oidset *' of commits, we place the burden on the caller (in
this case, the builtin) to handle the first case, and the commit-graph
machinery can handle the second case.
Suggested-by: Jeff King <peff@peff.net>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Occasionally, it may be useful for callers to know the number of object
IDs in an oidset. Right now, the only way to compute this is to call
'kh_size' on the internal 'kh_set_oid_t'.
Similar to how we wrap other 'kh_*' functions over the 'oidset' type,
let's allow callers to compute this value by introducing 'oidset_size'.
We will add its first caller in the subsequent commit.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using split commit-graphs, it is sometimes useful to completely
replace the commit-graph chain with a new base.
For example, consider a scenario in which a repository builds a new
commit-graph incremental for each push. Occasionally (say, after some
fixed number of pushes), they may wish to rebuild the commit-graph chain
with all reachable commits.
They can do so with
$ git commit-graph write --reachable
but this removes the chain entirely and replaces it with a single
commit-graph in 'objects/info/commit-graph'. Unfortunately, this means
that the next push will have to move this commit-graph into the first
layer of a new chain, and then write its new commits on top.
Avoid such copying entirely by allowing the caller to specify that they
wish to replace the entirety of their commit-graph chain, while also
specifying that the new commit-graph should become the basis of a fresh,
length-one chain.
This addresses the above situation by making it possible for the caller
to instead write:
$ git commit-graph write --reachable --split=replace
which writes a new length-one chain to 'objects/info/commit-graphs',
making the commit-graph incremental generated by the subsequent push
relatively cheap by avoiding the aforementioned copy.
In order to do this, remove an assumption in 'write_commit_graph_file'
that chains are always at least two incrementals long.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the previous commit, we laid the groundwork for supporting different
splitting strategies. In this commit, we introduce the first splitting
strategy: 'no-merge'.
Passing '--split=no-merge' is useful for callers which wish to write a
new incremental commit-graph, but do not want to spend effort condensing
the incremental chain [1]. Previously, this was possible by passing
'--size-multiple=0', but this no longer the case following 63020f175f
(commit-graph: prefer default size_mult when given zero, 2020-01-02).
When '--split=no-merge' is given, the commit-graph machinery will never
condense an existing chain, and it will always write a new incremental.
[1]: This might occur when, for example, a server administrator running
some program after each push may want to ensure that each job runs
proportional in time to the size of the push, and does not "jump" when
the commit-graph machinery decides to trigger a merge.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With '--split', the commit-graph machinery writes new commits in another
incremental commit-graph which is part of the existing chain, and
optionally decides to condense the chain into a single commit-graph.
This is done to ensure that the asymptotic behavior of looking up a
commit in an incremental chain is not dominated by the number of
incrementals in that chain. It can be controlled by the '--max-commits'
and '--size-multiple' options.
In the next two commits, we will introduce additional splitting
strategies that can exert additional control over:
- when a split commit-graph is and isn't written, and
- when the existing commit-graph chain is discarded completely and
replaced with another graph
To prepare for this, make '--split' take an optional strategy (as in
'--split[=<strategy>]'), and add a new enum to describe which strategy
is being used. For now, no strategies are given, and the only enumerated
value is 'COMMIT_GRAPH_SPLIT_UNSPECIFIED', indicating the absence of a
strategy.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In 61df89c8e5 (commit-graph: don't early exit(1) on e.g. "git status",
2019-03-25), the former 'load_commit_graph_one' was refactored into
'open_commit_graph' and 'load_commit_graph_one_fd_st' as a means of
avoiding an early-exit from non-library code.
However, 'load_commit_graph_one' does not support commit-graph chains,
and hence the 'read-graph' test tool does not work with them.
Replace 'load_commit_graph_one' with 'read_commit_graph_one' in order to
support commit-graph chains. In the spirit of 61df89c8e5,
'read_commit_graph_one' does not ever 'die()', making it a suitable
replacement here.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function won't work anywhere else, so let's mark it as an explicit
bug if it is called on a non-Windows platform.
Let's also rename the function to avoid cluttering the global namespace
with an overly-generic function name.
While at it, we also fix the code comment above that function: the
lower-case `windows` refers to something different than `Windows`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function uses Windows' system tool `attrib` to determine the state
of the hidden flag of a file or directory.
We should not actually expect the first `attrib.exe` in the PATH to
be the one we are looking for. Or that it is in the PATH, for that
matter.
Let's use the full path to the tool instead.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `is_hidden` function can be used (only on Windows) to determine
whether a directory or file have their `hidden` flag set.
This function is duplicated between two test scripts. It is better to
move it into `test-lib-functions.sh` so that it is reused.
This patch is best viewed with `--color-moved`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of using `starts_with()`, the magic number 7, `strlen()` and a
fair number of additions to verify the three parts of the config key
"branch.<branch>.mergeoptions", use `skip_prefix()` to jump through them
more explicitly.
We need to introduce a new variable for this (we certainly can't modify
`k` just because we see "branch."!). With `skip_prefix()` we often use
quite bland names like `p` or `str`. Let's do the same. If and when this
function needs to do more prefix-skipping, we'll have a generic variable
ready for this.
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When rebasing against an upstream that has had many commits since the
original branch was created:
O -- O -- ... -- O -- O (upstream)
\
-- O (my-dev-branch)
it must read the contents of every novel upstream commit, in addition to
the tip of the upstream and the merge base, because "git rebase"
attempts to exclude commits that are duplicates of upstream ones. This
can be a significant performance hit, especially in a partial clone,
wherein a read of an object may end up being a fetch.
Add a flag to "git rebase" to allow suppression of this feature. This
flag only works when using the "merge" backend.
This flag changes the behavior of sequencer_make_script(), called from
do_interactive_rebase() <- run_rebase_interactive() <-
run_specific_rebase() <- cmd_rebase(). With this flag, limit_list()
(indirectly called from sequencer_make_script() through
prepare_revision_walk()) will no longer call cherry_pick_list(), and
thus PATCHSAME is no longer set. Refraining from setting PATCHSAME both
means that the intermediate commits in upstream are no longer read (as
shown by the test) and means that no PATCHSAME-caused skipping of
commits is done by sequencer_make_script(), either directly or through
make_script_with_merges().
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the user specifies the apply backend with options that only work
with the merge backend, such as
git rebase --apply --exec /bin/true HEAD~3
the error message has always been
fatal: --exec requires an interactive rebase
This error message is misleading and was one of the reasons we renamed
the interactive backend to the merge backend. Update the error message
to state that these options merely require use of the merge backend.
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit d48e5e21da ("rebase (interactive-backend): make --keep-empty the
default", 2020-02-15) turned --keep-empty (for keeping commits which
start empty) into the default. The logic underpinning that commit was:
1) 'git commit' errors out on the creation of empty commits without an
override flag
2) Once someone determines that the override is worthwhile, it's
annoying and/or harmful to required them to take extra steps in
order to keep such commits around (and to repeat such steps with
every rebase).
While the logic on which the decision was made is sound, the result was
a bit of an overcorrection. Instead of jumping to having --keep-empty
being the default, it jumped to making --keep-empty the only available
behavior. There was a simple workaround, though, which was thought to
be good enough at the time. People could still drop commits which
started empty the same way the could drop any commits: by firing up an
interactive rebase and picking out the commits they didn't want from the
list. However, there are cases where external tools might create enough
empty commits that picking all of them out is painful. As such, having
a flag to automatically remove start-empty commits may be beneficial.
Provide users a way to drop commits which start empty using a flag that
existed for years: --no-keep-empty. Interpret --keep-empty as
countermanding any previous --no-keep-empty, but otherwise leaving
--keep-empty as the default.
This might lead to some slight weirdness since commands like
git rebase --empty=drop --keep-empty
git rebase --empty=keep --no-keep-empty
look really weird despite making perfect sense (the first will drop
commits which become empty, but keep commits that started empty; the
second will keep commits which become empty, but drop commits which
started empty). However, --no-keep-empty was named years ago and we are
predominantly keeping it for backward compatibility; also we suspect it
will only be used rarely since folks already have a simple way to drop
commits they don't want with an interactive rebase.
Reported-by: Bryan Turner <bturner@atlassian.com>
Reported-by: Sami Boukortt <sami@boukortt.com>
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While many users who intentionally create empty commits do not want them
thrown away by a rebase, there are third-party tools that generate empty
commits that a user might not want. In the past, users have used rebase
to get rid of such commits (a side-effect of the fact that the --apply
backend is not currently capable of keeping them). While such users
could fire up an interactive rebase and just remove the lines
corresponding to empty commits, that might be difficult if the
third-party tool generates many of them. Simplify this task for users
by marking such lines with a suffix of " # empty" in the todo list.
Suggested-by: Sami Boukortt <sami@boukortt.com>
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While the last few commits have made it possible for the config parser
to handle config files up to the limits of size_t, the rest of the code
isn't really ready for this. In particular, we often feed the keys as
strings into printf "%s" format specifiers. And because the printf
family of functions must return an int to specify the result, they
complain. Here are two concrete examples (using glibc; we're in
uncharted territory here so results may vary):
Generate a gigantic .gitmodules file like this:
git submodule add /some/other/repo foo
{
printf '[submodule "'
perl -e 'print "a" x 2**31'
echo '"]path = foo'
} >.gitmodules
git commit -m 'huge gitmodule'
then try this:
$ git show
BUG: strbuf.c:397: your vsnprintf is broken (returned -1)
The problem is that we end up calling:
strbuf_addf(&sb, "submodule.%s.ignore", submodule_name);
which relies on vsnprintf(), and that function has no way to report back
a size larger than INT_MAX.
Taking that same file, try this:
git config --file=.gitmodules --list --name-only
On my system it produces an output with exactly 4GB of spaces. I
confirmed in a debugger that we reach the config callback with the key
intact: it's 2147483663 bytes and full of a's. But when we print it with
this call:
printf("%s%c", key_, term);
we just get the spaces.
So given the fact that these are insane cases which we have no need to
support, the weird behavior from feeding the results to printf even if
the code is careful, and the possibility of uncareful code introducing
its own integer truncation issues, let's just declare INT_MAX as a limit
for parsing config files.
We'll enforce the limit in get_next_char(), which generalizes over all
sources (blobs, files, etc) and covers any element we're parsing
(whether section, key, value, etc). For simplicity, the limit is over
the length of the _whole_ file, so you couldn't have two 1GB values in
the same file. This should be perfectly fine, as the expected size for
config files is generally kilobytes at most.
With this patch both cases above will yield:
fatal: bad config line 1 in file .gitmodules
That's not an amazing error message, but the parser isn't set up to
provide specific messages (it just breaks out of the parsing loop and
gives that generic error even if see a syntactic issue). And we really
wouldn't expect to see this case outside of somebody maliciously probing
the limits of the config system.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Most of the config parsing infrastructure is limited in what it can
parse only by the size of memory, because it parses character by
character, building up strbufs for keys, values, etc. One exception is
the "baselen" value we keep in git_parse_source(), which is an int.
That stores the length of the section.subsection base, to which we can
then append individual key names (by truncating back to the baselen with
strbuf_setlen(), and then appending characters for the key name). But
because it's an int, if we see an absurdly long section or subsection,
we may overflow the integer, wrapping negative. That negative value is
then implicitly cast to a size_t when we pass it to strbuf_setlen(),
creating a very large value and triggering a BUG. For example:
$ {
printf '[foo "'
perl -e 'print "a" x 2**31'
echo '"]bar = value'
} >huge
$ git config --file=huge --list
fatal: BUG: strbuf_setlen() beyond buffer
While this is obviously a silly case that we don't care about
supporting, it's worth fixing it by switching to a size_t for a few
reasons:
- we should try to avoid hitting BUG assertions at all
- avoiding integer truncation or overflow sets a good example and
makes it easier to audit the code for more important issues
- the BUG outcome is what happens in _this_ instance, because we wrap
negative. If we used a 2**32 subsection, we'd wrap to a small
positive value and actually generate wrong output (the subsection of
our key would be truncated).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As with the recent change to parse_config_key(), the best type to return
a string length is a size_t, as it won't cause integer truncation for a
gigantic key. And as with that change, this is mostly a clarity /
hygiene issue for now, as our config parser would choke on such a large
key anyway.
There are a few ripple effects within the config code, as callers switch
to using size_t. I also adjusted a few related variables that iterate
over strings. The most unexpected change is that a call to strbuf_addf()
had to switch to strbuf_add(). We can't use a size_t with "%.*s",
because printf precisions must have type "int" (we could cast, of
course, but that would miss the point of using size_t in the first
place).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We compute the length of a subset of a string, but then use that length
only to feed a "%.*s" printf placeholder for the same string. We can
just use "%s" to achieve the same thing.
The variable became useless in cb891a5989 (Use a strbuf for building up
section header and key/value pair strings., 2007-12-14), which swapped
out a write() which _did_ use the length for a strbuf_addf() call.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We return the length to a subset of a string using an "int *"
out-parameter. This is fine most of the time, as we'd expect config keys
to be relatively short, but it could behave oddly if we had a gigantic
config key. A more appropriate type is size_t.
Let's switch over, which lets our callers use size_t as appropriate
(they are bound by our type because they must pass the out-parameter as
a pointer). This is mostly just a cleanup to make it clear this code
handles long strings correctly. In practice, our config parser already
chokes on long key names (because of a similar int/size_t mixup!).
When doing an int/size_t conversion, we have to be careful that nobody
was trying to assign a negative value to the variable. I manually
confirmed that for each case here. They tend to just feed the result to
xmemdupz() or similar; in a few cases I adjusted the parameter types for
helper functions to make sure the size_t is preserved.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The make_branch() and make_rewrite() functions can take a NUL-terminated
string or a ptr/len pair. They use a sentinel value of "0" for the len
to tell the difference between the two. However, when parsing config
like:
[branch ""]
merge = whatever
whose key flattens to:
branch..merge
we might actually have a zero-length branch name. This is obviously
nonsense, but the current code would consider it as a NUL-terminated
string and use the branch name ".merge".
We could use a better sentinel value here (like "-1"), but that gets in
the way of moving to size_t, which is a more appropriate type for a
ptr/len combo.
Let's instead just drop this feature and have the callers (of which
there are only two total) use strlen() themselves. This simplifies the
code, and lets us move to using size_t.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On certain network filesystems (currently encountered with Isilon, but
in theory more network storage solutions could be causing the same
issue), when the directory in question is missing,
`raceproof_create_file()` fails with an `ERROR_INVALID_PARAMETER`
instead of an `ERROR_PATH_NOT_FOUND`.
Since it is highly unlikely that we produce such an error by mistake
(the parameters we pass are fairly benign), we can be relatively certain
that the directory is missing in this instance. So let's just translate
that error automagically.
This fixes https://github.com/git-for-windows/git/issues/1345.
Signed-off-by: Nathan Sanders <spekbukkem@gmail.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Arguably, CI builds' most important task is to not only identify
regressions, but to make it as easy as possible to investigate what went
wrong.
In that light, we will want to provide users with a way to inspect the
tests' output as well as the corresponding directories.
This commit adds build steps that are only executed when tests failed,
uploading the relevant information as build artifacts. These artifacts
can then be downloaded by interested parties to diagnose the failures
more efficiently.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Đoàn Trần Công Danh <congdanhqx@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With this patch, test failures will be annotated with a helpful,
clickable message in GitHub Actions. For details, see
https://github.com/actions/toolkit/blob/master/docs/problem-matchers.md
Note: we need to set `TEST_SHELL_PATH` to Bash so that the problem
matcher is fed a file and line number for each test failure.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Đoàn Trần Công Danh <congdanhqx@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
