Junio C Hamano [Wed, 15 Sep 2021 20:15:26 +0000 (13:15 -0700)]
Merge branch 'pb/test-use-user-env'
Teach "test_pause" and "debug" helpers to allow using the HOME and
TERM environment variables the user usually uses.
* pb/test-use-user-env:
test-lib-functions: keep user's debugger config files and TERM in 'debug'
test-lib-functions: optionally keep HOME, TERM and SHELL in 'test_pause'
test-lib-functions: use 'TEST_SHELL_PATH' in 'test_pause'
Junio C Hamano [Wed, 15 Sep 2021 20:15:26 +0000 (13:15 -0700)]
Merge branch 'jc/trivial-threeway-binary-merge'
The "git apply -3" code path learned not to bother the lower level
merge machinery when the three-way merge can be trivially resolved
without the content level merge.
* jc/trivial-threeway-binary-merge:
apply: resolve trivial merge without hitting ll-merge with "--3way"
Jeff King [Wed, 15 Sep 2021 17:24:52 +0000 (13:24 -0400)]
t1400: avoid SIGPIPE race condition on fifo
t1400.190 sometimes fails or even hangs because of the way it uses
fifos. Our goal is to interactively read and write lines from
update-ref, so we have two fifos, in and out. We open a descriptor
connected to "in" and redirect output to that, so that update-ref does
not see EOF as it would if we opened and closed it for each "echo" call.
But we don't do the same for the output. This leads to a race where our
"read response <out" has not yet opened the fifo, but update-ref tries
to write to it and gets SIGPIPE. This can result in the test failing, or
worse, hanging as we wait forever for somebody to write to the pipe.
This is the same proble we fixed in 4783e7ea83 (t0008: avoid SIGPIPE
race condition on fifo, 2013-07-12), and we can fix it the same way, by
opening a second long-running descriptor.
Before this patch, running:
./t1400-update-ref.sh --run=1,190 --stress
failed or hung within a few dozen iterations. After it, I ran it for
several hundred without problems.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jonathan Tan [Wed, 15 Sep 2021 18:59:19 +0000 (11:59 -0700)]
submodule: extract path to submodule gitdir func
We currently store each submodule gitdir in ".git/modules/<name>", but
this has problems with some submodule naming schemes, as described in a
comment in submodule_name_to_gitdir() in this patch.
Extract the determination of the location of a submodule's gitdir into
its own function submodule_name_to_gitdir(). For now, the problem
remains unsolved, but this puts us in a better position for finding a
solution.
This was motivated, at $DAYJOB, by a part of Android's repo hierarchy
[1]. In particular, there is a repo "build", and several repos of the
form "build/<name>".
This is based on earlier work by Brandon Williams [2].
Jeff King [Wed, 15 Sep 2021 18:36:38 +0000 (14:36 -0400)]
ls-refs: reject unknown arguments
The v2 ls-refs command may receive extra arguments from the client, one
per pkt-line. The spec is pretty clear that the arguments must come from
a specified set, but we silently ignore any unknown entries. For a
well-behaved client this doesn't matter, but it makes testing and
debugging more confusing. Let's tighten this up to match the spec.
In theory this liberal behavior _could_ be useful for extending the
protocol. But:
- every other part of the protocol requires that the server first
indicate that it supports the argument; this includes the fetch and
object-info commands, plus the "unborn" capability added to ls-refs
itself
- it's not a very good extension mechanism anyway; without the server
advertising support, clients would have no idea if the argument was
silently ignored, or accepted and simply had no effect
So we're not really losing anything by tightening this.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 15 Sep 2021 18:36:36 +0000 (14:36 -0400)]
serve: reject commands used as capabilities
Our table of v2 "capabilities" contains everything we might tell the
client we support. But there are differences in how we expect the client
to respond. Some of the entries are true capabilities (i.e., we expect
the client to say "yes, I support this"), and some are ones we expect
them to send as commands (with "command=ls-refs" or similar).
When we receive a capability used as a command, we complain about that.
But when we receive a command used as a capability (e.g., just "ls-refs"
in a pkt-line by itself), we silently ignore it.
This isn't really hurting anything (clients shouldn't send it, and we'll
ignore it), but we can tighten up the protocol to match what we expect
to happen.
There are two new tests here. The first one checks a capability used as
a command, which already passes. The second tests a command as a
capability, which this patch fixes.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 15 Sep 2021 18:36:33 +0000 (14:36 -0400)]
serve: reject bogus v2 "command=ls-refs=foo"
When we see a line from the client like "command=ls-refs", we parse
everything after the equals sign as a capability, which we check against
our capabilities table. If we don't recognize the command (e.g.,
"command=foo"), we'll reject it.
But in parse_command(), we use the same get_capability() parser for
parsing non-command lines. So if we see "command=ls-refs=foo", we will
feed "ls-refs=foo" to get_capability(), which will say "OK, that's
ls-refs, with value 'foo'". But then we simply ignore the value
entirely.
The client is violating the spec here, which says:
I.e., the key is not even allowed to have an equals sign in it. Whereas
a real non-command capability does allow a value:
capability = PKT-LINE(key[=value] LF)
So by reusing the same get_capability() parser, we are mixing up the
"key" and "capability" tokens. However, since that parser tells us
whether it saw an "=", we can still use it; we just need to reject any
input that produces a non-NULL value field.
The current behavior isn't really hurting anything (the client should
never send such a request, and if it does, we just ignore the "value"
part). But since it does violate the spec, let's tighten it up to
prevent any surprising behavior.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 15 Sep 2021 18:35:34 +0000 (14:35 -0400)]
docs/protocol-v2: clarify some ls-refs ref-prefix details
We've never documented the fact that a client can provide multiple
ref-prefix capabilities. Let's describe the behavior.
We also never discussed the "best effort" nature of the prefixes. The
client side of git.git has always treated them this way, filtering the
result with local patterns. And indeed any client must do this, because
the prefix patterns are not sufficient to express the usual refspecs
(and so for "foo" we ask for "refs/heads/foo", "refs/tags/foo", and so
on).
So this may be considered a change in the spec with respect to client
expectations / requirements, but it's mostly codifying existing
behavior.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 15 Sep 2021 18:35:31 +0000 (14:35 -0400)]
ls-refs: ignore very long ref-prefix counts
Because each "ref-prefix" capability from the client comes in its own
pkt-line, there's no limit to the number of them that a misbehaving
client may send. We read them all into a strvec, which means the client
can waste arbitrary amounts of our memory by just sending us "ref-prefix
foo" over and over.
One possible solution is to just drop the connection when the limit is
reached. If we set it high enough, then only misbehaving or malicious
clients would hit it. But "high enough" is vague, and it's unfriendly if
we guess wrong and a legitimate client hits this.
But we can do better. Since supporting the ref-prefix capability is
optional anyway, the client has to further cull the response based on
their own patterns. So we can simply ignore the patterns once we cross a
certain threshold. Note that we have to ignore _all_ patterns, not just
the ones past our limit (since otherwise we'd send too little data).
The limit here is fairly arbitrary, and probably much higher than anyone
would need in practice. It might be worth limiting it further, if only
because we check it linearly (so with "m" local refs and "n" patterns,
we do "m * n" string comparisons). But if we care about optimizing this,
an even better solution may be a more advanced data structure anyway.
I didn't bother making the limit configurable, since it's so high and
since Git should behave correctly in either case. It wouldn't be too
hard to do, but it makes both the code and documentation more complex.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 15 Sep 2021 18:35:29 +0000 (14:35 -0400)]
serve: drop "keys" strvec
We collect the set of capabilities the client sends us in a strvec.
While this is usually small, there's no limit to the number of
capabilities the client can send us (e.g., they could just send us
"agent" pkt-lines over and over, and we'd keep adding them to the list).
Since all code has been converted away from using this list, let's get
rid of it. This avoids a potential attack where clients waste our
memory.
Note that we do have to replace it with a flag, because some of the
flush-packet logic checks whether we've seen any valid commands or keys.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 14 Sep 2021 23:51:30 +0000 (19:51 -0400)]
serve: provide "receive" function for session-id capability
Rather than pulling the session-id string from the list of collected
capabilities, we can handle it as soon as we receive it. This gets us
closer to dropping the collected list entirely.
The behavior should be the same, with one exception. Previously if the
client sent us multiple session-id lines, we'd report only the first.
Now we'll pass each one along to trace2. This shouldn't matter in
practice, since clients shouldn't do that (and if they do, it's probably
sensible to log them all).
As this removes the last caller of the static has_capability(), we can
remove it, as well (and in fact we must to avoid -Wunused-function
complaining).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 14 Sep 2021 23:51:27 +0000 (19:51 -0400)]
serve: provide "receive" function for object-format capability
We get any "object-format" specified by the client by searching for it
in the collected list of capabilities the client sent. We can instead
just handle it as soon as they send it. This is slightly more efficient,
and gets us one step closer to dropping that collected list.
Note that we do still have to do our final hash check after receiving
all capabilities (because they might not have sent an object-format line
at all, and we still have to check that the default matches our
repository algorithm). Since the check_algorithm() function would now be
down to a single if() statement, I've just inlined it in its only
caller.
There should be no change of behavior here, except for two
broken-protocol cases:
- if the client sends multiple conflicting object-format capabilities
(which they should not), we'll now choose the last one rather than
the first. We could also detect and complain about the duplicates
quite easily now, which we could not before, but I didn't do so
here.
- if the client sends a bogus "object-format" with no equals sign,
we'll now say so, rather than "unknown object format: ''"
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Taylor Blau [Tue, 14 Sep 2021 22:06:11 +0000 (18:06 -0400)]
p5326: don't set core.multiPackIndex unnecessarily
When this performance test was originally written, `core.multiPackIndex`
was not the default and thus had to be enabled. But now that we have 18e449f86b (midx: enable core.multiPackIndex by default, 2020-09-25), we
no longer need this.
Drop the unnecessary setup (even though it's not hurting anything, it is
unnecessary at best and confusing at worst).
Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Taylor Blau [Tue, 14 Sep 2021 22:06:09 +0000 (18:06 -0400)]
p5326: create missing 'perf-tag' tag
Some of the tests in test_full_bitmap rely on having a tag named
perf-tag in place. We could create it in test_full_bitmap(), but we want
to have it in place before the repack starts.
Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Taylor Blau [Tue, 14 Sep 2021 22:06:06 +0000 (18:06 -0400)]
midx.c: respect 'pack.writeBitmapHashcache' when writing bitmaps
In the previous commit, the bitmap writing code learned to propagate
values from an existing hash-cache extension into the bitmap that it is
writing.
Now that that functionality exists, let's expose it by teaching the 'git
multi-pack-index' builtin to respect the `pack.writeBitmapHashCache`
option so that the hash-cache may be written at all.
Two minor points worth noting here:
- The 'git multi-pack-index write' sub-command didn't previously read
any configuration (instead this is handled in the base command). A
separate handler is added here to respect this write-specific
config option.
- I briefly considered adding a 'bitmap_flags' field to the static
options struct, but decided against it since it would require
plumbing through a new parameter to the write_midx_file() function.
Instead, a new MIDX-specific flag is added, which is translated to
the corresponding bitmap one.
Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Taylor Blau [Tue, 14 Sep 2021 22:06:04 +0000 (18:06 -0400)]
pack-bitmap.c: propagate namehash values from existing bitmaps
When an old bitmap exists while writing a new one, we load it and build
a "reposition" table which maps bit positions of objects from the old
bitmap to their respective positions in the new bitmap. This can help
when we encounter a commit which was selected in both the old and new
bitmap, since we only need to permute its bit (not recompute it from
scratch).
We do not, however, repurpose existing namehash values in the case of
the hash-cache extension. There has been thus far no good reason to do
so, since all of the namehash values for objects in the new bitmap would
be populated during the traversal that was just performed by
pack-objects when generating single-pack reachability bitmaps.
But this isn't the case for multi-pack bitmaps, which are written via
`git multi-pack-index write --bitmap` and do not perform any traversal.
In this case all namehash values are set to zero, but we don't even
bother to check the `pack.writeBitmapHashcache` option anyway, so it
fails to matter.
There are two approaches we could take to fill in non-zero hash-cache
values:
- have either the multi-pack-index builtin run its own
traversal to attempt to fill in some values, or let a hypothetical
caller (like `pack-objects` when `repack` eventually drives the
`multi-pack-index` builtin) fill in the values they found during
their traversal
- or copy any existing namehash values that were stored in an
existing bitmap to their corresponding positions in the new bitmap
In a system where a repository is generally repacked with `git repack
--geometric=<d>` and occasionally repacked with `git repack -a`, the
hash-cache coverage will tend towards all objects.
Since populating the hash-cache is additive (i.e., doing so only helps
our delta search), any intermediate lack of full coverage is just fine.
So let's start by just propagating any values from the existing
hash-cache if we see one.
The next patch will respect the `pack.writeBitmapHashcache` option while
writing MIDX bitmaps, and then test this new behavior.
Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Taylor Blau [Tue, 14 Sep 2021 22:06:02 +0000 (18:06 -0400)]
t/helper/test-bitmap.c: add 'dump-hashes' mode
The pack-bitmap writer code is about to learn how to propagate values
from an existing hash-cache. To prepare, teach the test-bitmap helper to
dump the values from a bitmap's hash-cache extension in order to test
those changes.
Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 14 Sep 2021 15:31:07 +0000 (11:31 -0400)]
serve: add "receive" method for v2 capabilities table
We have a capabilities table that tells us what we should tell the
client we are capable of, and what to do when a client gives us a
particular command (e.g., "command=ls-refs"). But it doesn't tell us
what to do when the client sends us back a capability (e.g.,
"object-format=sha256"). We just collect them all in a strvec and hope
somebody can use them later.
Instead, let's provide a function pointer in the table to act on these.
This will eventually help us avoid collecting the strings, which will be
more efficient and less prone to mischief.
Using the new method is optional, which helps in two ways:
- we can move existing capabilities over to this new system gradually
in individual commits
- some capabilities we don't actually do anything with anyway. For
example, the client is free to say "agent=git/1.2.3" to us, but we
do not act on the information in any way.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 14 Sep 2021 15:30:50 +0000 (11:30 -0400)]
serve: return capability "value" from get_capability()
When the client sends v2 capabilities, they may be simple, like:
foo
or have a value like:
foo=bar
(all of the current capabilities actually expect a value, but the
protocol allows for boolean ones).
We use get_capability() to make sure the client's pktline matches a
capability. In doing so, we parse enough to see the "=" and the value
(if any), but we immediately forget it. Nobody cares for now, because they end
up parsing the values out later using has_capability(). But in
preparation for changing that, let's pass back a pointer so the callers
know what we found.
Note that unlike has_capability(), we'll return NULL for a "simple"
capability. Distinguishing these will be useful for some future patches.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 14 Sep 2021 15:30:20 +0000 (11:30 -0400)]
serve: rename is_command() to parse_command()
The is_command() function not only tells us whether the pktline is a
valid command string, but it also parses out the command (and complains
if we see a duplicate). Let's rename it to make those extra functions
a bit more obvious.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Tue, 14 Sep 2021 17:56:05 +0000 (10:56 -0700)]
Merge branch 'ab/serve-cleanup' into jk/reduce-malloc-in-v2-servers
* ab/serve-cleanup:
upload-pack: document and rename --advertise-refs
serve.[ch]: remove "serve_options", split up --advertise-refs code
{upload,receive}-pack tests: add --advertise-refs tests
serve.c: move version line to advertise_capabilities()
serve: move transfer.advertiseSID check into session_id_advertise()
serve.[ch]: don't pass "struct strvec *keys" to commands
serve: use designated initializers
transport: use designated initializers
transport: rename "fetch" in transport_vtable to "fetch_refs"
serve: mark has_capability() as static
documentation: add documentation for 'git version'
While 'git version' is probably the least complex git command,
it is a non-experimental user-facing builtin command. As such
it should have a help page.
Both `git help` and `git version` can be called as options
(`--help`/`--version`) that internally get converted to the
corresponding command. Add a small paragraph to
Documentation/git.txt describing how these two options
interact with each other and link to this help page for the
sub-options that `--version` can take. Well, currently there
is only one sub-option, but that could potentially increase
in future versions of Git.
Signed-off-by: Matthias Aßhauer <mha1993@live.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
help: make sure local html page exists before calling external processes
We check that git.html exists, regardless of the page the user wants to open.
Checking whether the requested page exists instead gives us a smoother user
experience in two use cases:
1) The requested page doesn't exist
When calling a git command and there is an error, most users reasonably expect
git to produce an error message on the standard error stream, but in this case
we pass the filepath to git web--browse which passes it on to a browser (or a
helper program like xdg-open or start that should in turn open a browser)
without any error and many GUI based browsers or helpers won't output such a
message onto the standard error stream.
Especially the helper programs tend to show the corresponding error message in
a message box and wait for user input before exiting. This leaves users in
interactive console sessions without an error message in their console,
without a console prompt and without the help page they expected.
2) git.html is missing for some reason, but the user asked for some other page
We currently refuse to show any local html help page when we can't find
git.html. Even if the requested help page exists. If we check for the requested
page instead, we can show the user all available pages and only error out on
those that don't exist.
Signed-off-by: Matthias Aßhauer <mha1993@live.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-compat-util: include declaration for unix sockets in windows
Available since Windows 10 release 1803 and Windows Server 2019.
NO_UNIX_SOCKETS is still the default for Windows builds, as they need
to keep backward compatibility with releases up to Windows 7, but allow
including the header otherwise.
Signed-off-by: Carlo Marcelo Arenas Belón <carenas@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
credential-cache: check for windows specific errors
Connect and reset errors aren't what will be expected by POSIX but
are instead compatible with the ones used by WinSock.
To avoid any possibility of confusion with other systems, checks
for disconnection and availability had been abstracted into helper
functions that are platform specific.
Signed-off-by: Carlo Marcelo Arenas Belón <carenas@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In preparation for a future patch that will allow building with
Unix Sockets in Windows, workaround a couple of issues from the
Mingw-W64 compatibility layer.
test -S is not able to detect that a file is a socket, so use
test -e instead (through a library function).
`mkdir -m` can't represent a valid ACL directly and fails with
permission problems, so instead call mkdir followed by chmod, which
has been enhanced to do so.
The last invocation of mkdir would likely need the same treatment
but SYMLINK is unlikely to be enabled on Windows so it has been
punted for now.
Signed-off-by: Carlo Marcelo Arenas Belón <carenas@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Philip Oakley [Mon, 13 Sep 2021 21:23:05 +0000 (22:23 +0100)]
doc: config, tell readers of `git help --config`
The `git help` command gained the ability to list config variables in 3ac68a93fd (help: add --config to list all available config, 2018-05-26)
but failed to tell readers of the config documenation itself.
Provide that cross reference.
Signed-off-by: Philip Oakley <philipoakley@iee.email> Signed-off-by: Junio C Hamano <gitster@pobox.com>
After reimplementation of `git bisect run` in C,
`--bisect-next-check` subcommand is not needed anymore.
Let's remove it from options list and code.
Mentored by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Miriam Rubio <mirucam@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
bisect--helper: reimplement `bisect_visualize()` shell function in C
Reimplement the `bisect_visualize()` shell function
in C and also add `--bisect-visualize` subcommand to
`git bisect--helper` to call it from git-bisect.sh.
Mentored-by: Christian Couder <chriscool@tuxfamily.org> Mentored-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Tanushree Tumane <tanushreetumane@gmail.com> Signed-off-by: Miriam Rubio <mirucam@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Remove the `static` keyword from `exists_in_PATH()` function
and declare the function in `run-command.h` file.
The function will be used in bisect_visualize() in a later
commit.
Mentored by: Christian Couder <chriscool@tuxfamily.org>
Mentored by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Tanushree Tumane <tanushreetumane@gmail.com> Signed-off-by: Miriam Rubio <mirucam@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
9512177b68 ("rebase: add --quit to cleanup rebase, leave everything
else untouched", 2016-11-12) seems to have copied the --abort tests
but added two separate tests for the two rebase backends rather than
adding a single test into the existing testrebase() function.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
$dotest holds the name of the rebase state directory and was named
after the state directory used at the time the test was added. As we
no longer use that name rename the variable to reflect its purpose.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Simplify the setup by using test_commit. Note that this replaces the
branch "pre-rebase" with a tag of the same name. "pre-rebase" is only
used as a fixed reference point so it makes sense to use a tag rather
than a branch.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 97b88dd58c ("git-rebase.sh: Fix --merge --abort failures when
path contains whitespace", 2008-05-04) started running these tests in
a subdirectory with a space in its name. At that time $TEST_DIRECTORY
did not contain a space but shortly after in 4a7aaccd83 ("Rename the
test trash directory to contain spaces.", 2008-05-04) $TEST_DIRECTORY
was changed to contain a space so we no longer need to run these tests
in a subdirectory.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use the new git-curl-compat.h header to define CURL_SOCKOPT_OK to its
known value if we're on an older curl version that doesn't have it. It
was hardcoded in http.c in a15d069a198 (http: enable keepalive on TCP
sockets, 2013-10-12).
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http: centralize the accounting of libcurl dependencies
As discussed in 644de29e220 (http: drop support for curl < 7.19.4,
2021-07-30) checking against LIBCURL_VERSION_NUM isn't as reliable as
checking specific symbols present in curl, as some distros have been
known to backport features.
However, while some of the curl_easy_setopt() arguments we rely on are
macros, others are enum, and we can't assume that those that are
macros won't change into enums in the future.
So we're still going to have to check LIBCURL_VERSION_NUM, but by
doing that in one central place and using a macro definition of our
own, anyone who's backporting features can define it themselves, and
thus have access to more modern curl features that they backported,
even if they didn't bump the LIBCURL_VERSION_NUM.
More importantly, as shown in a preceding commit doing these version
checks makes for hard to read and possibly buggy code, as shown by the
bug fixed there where we were conflating base 10 for base 16 when
comparing the version.
By doing them all in one place we'll hopefully reduce the chances of
such future mistakes, furthermore it now becomes easier to see at a
glance what the oldest supported version is, which makes it easier to
reason about any future deprecation similar to the recent e48a623dea0 (Merge branch 'ab/http-drop-old-curl', 2021-08-24).
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http: correct curl version check for CURLOPT_PINNEDPUBLICKEY
In aeff8a61216 (http: implement public key pinning, 2016-02-15) a
dependency and warning() was added if curl older than 7.44.0 was used,
but the relevant code depended on CURLOPT_PINNEDPUBLICKEY, introduced
in 7.39.0.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http: correct version check for CURL_HTTP_VERSION_2
In d73019feb44 (http: add support selecting http version, 2018-11-08)
a dependency was added on CURL_HTTP_VERSION_2, but this feature was
introduced in curl version 7.43.0, not 7.47.0, as the incorrect
version check led us to believe.
As looking through the history of that commit on the mailing list will
reveal[1], the reason for this is that an earlier version of it
depended on CURL_HTTP_VERSION_2TLS, which was introduced in libcurl
7.47.0.
But the version that made it in in d73019feb44 had dropped the
dependency on CURL_HTTP_VERSION_2TLS, but the corresponding version
check was not corrected.
The newest symbol we depend on is CURL_HTTP_VERSION_2. It was added in
7.33.0, but the CURL_HTTP_VERSION_2 alias we used was added in
7.47.0. So we could support an even older version here, but let's just
correct the checked version.
In 644de29e220 (http: drop support for curl < 7.19.4, 2021-07-30) we
dropped support for curl < 7.19.4, so we can drop support for this
non-obvious dependency on curl < 7.18.0.
It's non-obvious because in curl's hex version notation 0x071800 is
version 7.24.0, *not* 7.18.0, so at a glance this patch looks
incorrect.
But it's correct, because the existing version check being removed
here is wrong. The check guards use of the following curl defines:
I.e. the oldest version that has these is in fact 7.18.0, not
7.24.0. That we were checking 7.24.0 is just an mistake in 6d7afe07f29 (remote-http(s): support SOCKS proxies, 2015-10-26),
i.e. its author confusing base 10 and base 16.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In 1119a15b5c8 (http: drop support for curl < 7.11.1, 2021-07-30)
support for curl versions older than 7.11.1 was removed, and we
currently require at least version 7.19.4, see 644de29e220 (http: drop
support for curl < 7.19.4, 2021-07-30).
In those changes this Makefile-specific check added in 0890098780f (Decide whether to build http-push in the Makefile,
2005-11-18) was missed, now that we're never going to use such an
ancient curl version we don't need to check that we have at least
7.9.8 here. I have no idea what in http-push.c broke on versions older
than that.
This does not impact "NO_CURL" setups, as this is in the "else" branch
after that check.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
INSTALL: mention that we need libcurl 7.19.4 or newer to build
Without NO_CURL=Y we require at least version "7.19.4" of libcurl, see 644de29e220 (http: drop support for curl < 7.19.4, 2021-07-30). Let's
document this in the "INSTALL" document.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
As was noted in 1a85b49b87a (parse-options: make OPT_ARGUMENT() more
useful, 2019-03-14) there's only ever been one user of the
OPT_ARGUMENT(), that user was added in 20de316e334 (difftool: allow
running outside Git worktrees with --no-index, 2019-03-14).
The OPT_ARGUMENT() feature itself was added way back in 580d5bffdea (parse-options: new option type to treat an option-like
parameter as an argument., 2008-03-02), but as discussed in 1a85b49b87a wasn't used until 20de316e334 in 2019.
Now that the preceding commit has migrated this code over to using
"struct strvec" to manage the "args" member of a "struct
child_process", we can just use that directly instead of relying on
OPT_ARGUMENT.
This has a minor change in behavior in that if we'll pass --no-index
we'll now always pass it as the first argument, before we'd pass it in
whatever position the caller did. Preserving this was the real value
of OPT_ARGUMENT(), but as it turns out we didn't need that either. We
can always inject it as the first argument, the other end will parse
it just the same.
Note that we cannot remove the "out" and "cpidx" members of "struct
parse_opt_ctx_t" added in 580d5bffdea, while they were introduced with
OPT_ARGUMENT() we since used them for other things.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
difftool: use run_command() API in run_file_diff()
Change the run_file_diff() function to use the run_command() API
directly, instead of invoking the run_command_v_opt_cd_env() wrapper.
This allows it, like run_dir_diff(), to use the "args" from "struct
strvec", instead of the "const char **argv" passed into
cmd_difftool(). This will be used in the subsequent commit to get rid
of OPT_ARGUMENT() from cmd_difftool().
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Mon, 13 Sep 2021 03:35:38 +0000 (05:35 +0200)]
difftool: prepare "diff" cmdline in cmd_difftool()
We call into either run_dir_diff() or run_file_diff(), each of which
sets up a child argv starting with "diff" and some hard-coded options
(depending on which mode we're using). Let's extract that logic into the
caller, which will make it easier to modify the options for cases which
affect both functions.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
difftool: prepare "struct child_process" in cmd_difftool()
Move the preparation of the "struct child_process" from run_dir_diff()
to its only caller, cmd_difftool(). This is in preparation for
migrating run_file_diff() to using the run_command() API directly, and
to move more of the shared setup of the two to cmd_difftool().
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git rev-parse --parseopt tests: add more usagestr tests
Add tests for the "usagestr" passed to parse-options.c
usage_with_options_internal() through cmd_parseopt().
These test for edge cases in the existing behavior related to the
"--parseopt" interface doing its own line-splitting with
strbuf_getline(), and the native C interface expecting and potentially
needing to handle newlines within the strings in the array it
accepts. The results are probably something that wasn't anticipated,
but let's make sure we stay backwards compatible with it.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
send-pack: properly use parse_options() API for usage string
When "send-pack" was changed to use the parse_options() API in 068c77a5189 (builtin/send-pack.c: use parse_options API, 2015-08-19)
it was made to use one very long line, instead it should split them up
with newlines.
Furthermore we were including an inline explanation that you couldn't
combine "--all" and "<ref>", but unlike in the "blame" case this was
not preceded by an empty string.
Let's instead show that --all and <ref> can't be combined in the the
usual language of the usage syntax instead. We can make it clear that
one of the two options "--foo" and "--bar" is mandatory, but that the
two are mutually exclusive by referring to them as "( --foo | --bar
)".
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
parse-options API users: align usage output in C-strings
In preparation for having continued usage lines properly aligned in
"git <cmd> -h" output, let's have the "[" on the second such lines
align with the "[" on the first line.
In some cases this makes the output worse, because e.g. the "git
ls-remote -h" output had been aligned to account for the extra
whitespace that the usage_with_options_internal() function in
parse-options.c would add.
In other cases such as builtin/stash.c (not changed here), we were
aligned in the C strings, but since that didn't account for the extra
padding in usage_with_options_internal() it would come out looking
misaligned, e.g. code like this:
Let's change all the usage arrays which use such continued usage
output via "\n"-embedding to be like builtin/stash.c.
This makes the output worse temporarily, but in a subsequent change
I'll improve the usage_with_options_internal() to take this into
account, at which point all of the strings being changed here will
emit prettier output.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the launchctl_boot_plist() function was added in a16eb6b1ff3 (maintenance: skip bootout/bootstrap when plist is
registered, 2021-08-24), an unused call to launchctl_get_uid() was
added along with it. That call appears to have been copy/pasted from
launchctl_boot_plist().
Since we can remove that, we can also get rid of the "result"
variable, whose only purpose was allow for the free() between its
assignment and the return. That pattern also appears to have been
copy/pasted from launchctl_boot_plist().
As the patch shows the returned value from launchctl_get_uid() wasn't
used at all in this function. The launchctl_get_uid() function itself
just calls xstrfmt() and getuid(), neither of which have any subtle
global side-effects, so this removal is safe.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In be5d88e1128 (test-tool run-command: learn to run (parts of) the
testsuite, 2019-10-04) an init pattern was added that would use
TESTSUITE_INIT, but then promptly memset() everything back to 0. We'd
then set the "dup" on the two string lists.
Our setting of "next" to "-1" thus did nothing, we'd reset it to "0"
before using it. Let's set it to "0" instead, and trust the
"STRING_LIST_INIT_DUP" to set "strdup_strings" appropriately for us.
Note that while we compile this code, there's no in-tree user for the
"testsuite" target being modified here anymore, see the discussion at
and around <nycvar.QRO.7.76.6.2109091323150.59@tvgsbejvaqbjf.bet>[1].
Remove untracked files that are unwanted after they are done being used.
While the set of cases in this patch is certainly far from
comprehensive, it was motivated by some work to see what the fallout
would be if we were to make the removal of untracked files as a side
effect of other commands into an error. Some cases were a bit more
involved than the testcase changes included in this patch, but the ones
included here represent the simple cases. While this patch is not that
important since we are not changing the behavior of those other commands
into an error in the near term, I thought these changes were useful
anyway as an explicit documentation of the intent that these untracked
files are no longer useful.
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Acked-by: Derrick Stolee <stolee@gmail.com> Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Sat, 11 Sep 2021 15:01:16 +0000 (11:01 -0400)]
strvec: use size_t to store nr and alloc
We converted argv_array (which later became strvec) to use size_t in 819f0e76b1 (argv-array: use size_t for count and alloc, 2020-07-28) in
order to avoid the possibility of integer overflow. But later, commit d70a9eb611 (strvec: rename struct fields, 2020-07-28) accidentally
converted these back to ints!
Those two commits were part of the same patch series. I'm pretty sure
what happened is that they were originally written in the opposite order
and then cleaned up and re-ordered during an interactive rebase. And
when resolving the inevitable conflict, I mistakenly took the "rename"
patch completely, accidentally dropping the type change.
We can correct it now; better late than never.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
René Scharfe [Sat, 11 Sep 2021 11:45:43 +0000 (13:45 +0200)]
compression: drop write-only core_compression_* variables
Since 8de7eeb54b (compression: unify pack.compression configuration
parsing, 2016-11-15) the variables core_compression_level and
core_compression_seen are only set, but never read. Remove them.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
These two have fallen out of use with the SHA-256 migration.
The last use of $_x40 was removed in fc7e73d7ef (t4013: improve
diff-post-processor logic, 2020-08-21) and
The last use of $_z40 was removed in 7a868c51c2 (t5562: use $ZERO_OID,
2019-12-21), but it was then needlessly refactored to be hash-agnostic
in 192b517589 (t: use hash-specific lookup tables to define test
constants, 2020-02-22). We can just remove it.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-sh-setup: remove unused "pull with rebase" message
Remove the "pull with rebase" message previously used by the
git-pull.sh script, which was removed in 49eb8d39c78 (Remove
contrib/examples/*, 2018-03-25).
Even if some out-of-tree user copy/pasted the old git-pull.sh code,
and relied on passing it a "pull with rebase" argument, we'll fall
back on the "*" case here, they just won't get the "pull with rebase"
part of their message translated.
I don't think it's likely that anyone out-of-tree relied on that, but
I'm being conservative here per the discussion that can be found
upthread of [1].
git-submodule: remove unused is_zero_oid() function
The is_zero_oid() function in git-submodule.sh has not been used since e83e3333b57 (submodule: port submodule subcommand 'summary' from shell
to C, 2020-08-13), so we can remove it.
This was the last user of the sane_egrep() function in
git-sh-setup.sh. I'm not removing it in case some out-of-tree user
relied on it. Per the discussion that can be found upthread of [1].
René Scharfe [Sat, 11 Sep 2021 20:43:26 +0000 (22:43 +0200)]
packfile: use oidset for bad objects
Store the object ID of broken pack entries in an oidset instead of
keeping only their hashes in an unsorted array. The resulting code is
shorter and easier to read. It also handles the (hopefully) very rare
case of having a high number of bad objects better.
Helped-by: Jeff King <peff@peff.net> Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
René Scharfe [Sat, 11 Sep 2021 20:39:31 +0000 (22:39 +0200)]
midx: inline nth_midxed_pack_entry()
fill_midx_entry() finds the position of an object ID and passes it to
nth_midxed_pack_entry(), which uses the position to look up the object
ID for its own purposes. Inline the latter into the former to avoid
that lookup.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
René Scharfe [Sat, 11 Sep 2021 20:36:40 +0000 (22:36 +0200)]
oidset: make oidset_size() an inline function
oidset_size() just reads a single word from memory and returns it.
Avoid the function call overhead for this trivial operation by turning
it into an inline function.
While we're at it, declare its parameter const to allow it to be used
on read-only oidsets.
Suggested-by: Jeff King <peff@peff.net> Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
INSTALL: don't mention the "curl" executable at all
In 1d53f90ed97 (The "curl" executable is no longer required,
2008-06-15) the wording for requiring curl(1) was changed to the
current "you might also want...".
Mentioning the "curl" executable at all is just confusing, someone
building git might want to use it to debug things, but they might also
just use wget(1) or some other http client. The "curl" executable has
the advantage that you might be able to e.g. reproduce a bug in git's
usage of libcurl with it, but anyone going to those extents is
unlikely to be aided by this note in INSTALL.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
help: correct the usage string in -h and documentation
Clarify the usage string in the documentation so we group e.g. -i and
--info, and add the missing short options to the "-h" output.
The alignment of the second line is off now, but will be fixed with
another series of mine[1]. In the meantime let's just assume that fix
will make it in eventually for the purposes of this patch, if it's
misaligned for a bit it doesn't matter much.
Both Johannes and I assumed (perhaps due to familiarity with rebase)
that am --abort would return the user to a clean state. However, since
am, unlike rebase, is intended to be used within a dirty working tree,
--abort will only clean the files involved in the am operation.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
stash: restore untracked files AFTER restoring tracked files
If a user deletes a file and places a directory of untracked files
there, then stashes all these changes, the untracked directory of files
cannot be restored until after the corresponding file in the way is
removed. So, restore changes to tracked files before restoring
untracked files.
There is no counterpart problem to worry about with the user deleting an
untracked file and then add a tracked one in its place. Git does not
track untracked files, and so will not know the untracked file was
deleted, and thus won't be able to stash the removal of that file.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a file is removed from the cache, but there is a file of the same
name present in the working directory, we would normally treat that file
in the working directory as untracked. However, in the case of stash,
doing that would prevent a simple 'git stash push', because the untracked
file would be in the way of restoring the deleted file.
git stash, however, blindly assumes that whatever is in the working
directory for a deleted file is wanted and passes that path along to
update-index. That causes problems when the working directory contains
a directory with the same name as the deleted file. Add some code for
this special case that will avoid passing directory names to
update-index.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are three tests here, because the second bug is documented with
two tests: a file -> directory change and a directory -> file change.
The reason for the two tests is just to verify that both are indeed
broken but that both will be fixed by the same simple change (which will
be provided in a subsequent patch).
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Fri, 10 Sep 2021 14:10:19 +0000 (10:10 -0400)]
docs/protocol-v2: point readers transport config discussion
We recently added tips for server admins to configure various transports
to support v2's GIT_PROTOCOL variable. While the protocol-v2 document is
pretty technical and not of interest to most admins, it may be a
starting point for them to figure out how to turn on v2. Let's put some
pointers from there to the other documentation.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Fri, 10 Sep 2021 14:09:56 +0000 (10:09 -0400)]
docs/git: discuss server-side config for GIT_PROTOCOL
The v2 protocol requires that the GIT_PROTOCOL environment variable gets
passed around, but we don't have any documentation describing how this
is supposed to work. In particular, we need to note what server admins
might need to configure to make things work.
The definition of the GIT_PROTOCOL variable is probably the best place
for this, since:
- we deal with multiple transports (ssh, http, etc).
Transport-specific documentation (like the git-http-backend bits
added in the previous commit) are helpful for those transports, but
this gives a broader overview. Plus we do not have a specific
transport endpoint program for ssh, so this is a reasonable place to
mention it.
- the server side of the protocol involves multiple programs. For now,
upload-pack is the only endpoint which uses GIT_PROTOCOL, but that
will likely expand in the future. We're better off with a central
discussion of what the server admin might need to do. However, for
discoverability, this patch adds a pointer from upload-pack's
documentation.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Fri, 10 Sep 2021 14:09:13 +0000 (10:09 -0400)]
docs/http-backend: mention v2 protocol
Historically there was a little bit of configuration needed at the
webserver level in order to get the client's v2 protocol probes to Git.
But when we introduced the v2 protocol, we never documented these.
As of the previous commit, this should mostly work out of the box
without any explicit configuration. But it's worth documenting this to
make it clear how we expect it to work, especially in the face of
webservers which don't provide all headers over the CGI interface. Or
anybody who runs across this documentation but has an older version of
Git (or _used_ to have an older version, and wonders why they still have
a SetEnvIf line in their Apache config and whether it's still
necessary).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a client requests the v2 protocol over HTTP, they set the
Git-Protocol header. Webservers will generally make that available to
our CGI as HTTP_GIT_PROTOCOL in the environment. However, that's not
sufficient for upload-pack, etc, to respect it; they look in
GIT_PROTOCOL (without the HTTP_ prefix).
Either the webserver or the CGI is responsible for relaying that HTTP
header into the GIT_PROTOCOL variable. Traditionally, our tests have
configured the webserver to do so, but that's a burden on the server
admin. We can make this work out of the box by having the http-backend
CGI copy the contents of HTTP_GIT_PROTOCOL to GIT_PROTOCOL.
There are no new tests here. By removing the SetEnvIf line from our
test Apache config, we're now relying on this behavior of http-backend
to trigger the v2 protocol there (and there are numerous tests that fail
if this doesn't work).
There is one subtlety here: we copy HTTP_GIT_PROTOCOL only if there is
no existing GIT_PROTOCOL variable. That leaves the webserver admin free
to override the client's decision if they choose. This is unlikely to be
useful in practice, but is more flexible. And indeed, it allows the
v2-to-v0 fallback test added in the previous commit to continue working.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Fri, 10 Sep 2021 14:04:42 +0000 (10:04 -0400)]
t5551: test v2-to-v0 http protocol fallback
Since we use the v2 protocol by default, the connection of a v2 client
to a v2 server is well covered by the test suite. And with the
GIT_TEST_PROTOCOL_VERSION knob, we can easily test a v0 client
connecting to a v2-aware server (which will then just speak v0). But we
have no regular tests that a v2 client, when encountering a non-v2-aware
server, will correctly fall back to using v0.
In theory this is a job for the cross-version tests in t/interop, but:
- they cover only git:// and file:// clones
- they are not part of the usual test suite, so nobody ever runs them
anyway
Since using v2 over http requires configuring the web server to pass
along the Git-Protocol header, we can easily create a situation where
the server does not respect the v2 probe, and the conversation falls
back to v0.
This works just fine. This new test is not about fixing any particular
bug, but just making sure that the system works (and continues to work)
as expected.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
René Scharfe [Fri, 10 Sep 2021 20:25:50 +0000 (22:25 +0200)]
index-pack: use xopen in init_thread
Support an arbitrary file descriptor expression in the semantic patch
for replacing open+die_errno with xopen, not just an identifier, and
apply it. This makes the error message at the single affected place
more consistent and reduces code duplication.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Test that verify-commit/tag will fail when a gpg key is completely
unknown. To do this we have to generate a key, use it for a signature
and delete it from our keyring aferwards completely.
Signed-off-by: Fabian Stelzer <fs@gigacodes.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
To verify a ssh signature we first call ssh-keygen -Y find-principal to
look up the signing principal by their public key from the
allowedSignersFile. If the key is found then we do a verify. Otherwise
we only validate the signature but can not verify the signers identity.
Verification uses the gpg.ssh.allowedSignersFile (see ssh-keygen(1) "ALLOWED
SIGNERS") which contains valid public keys and a principal (usually
user@domain). Depending on the environment this file can be managed by
the individual developer or for example generated by the central
repository server from known ssh keys with push access. This file is usually
stored outside the repository, but if the repository only allows signed
commits/pushes, the user might choose to store it in the repository.
To revoke a key put the public key without the principal prefix into
gpg.ssh.revocationKeyring or generate a KRL (see ssh-keygen(1)
"KEY REVOCATION LISTS"). The same considerations about who to trust for
verification as with the allowedSignersFile apply.
Using SSH CA Keys with these files is also possible. Add
"cert-authority" as key option between the principal and the key to mark
it as a CA and all keys signed by it as valid for this CA.
See "CERTIFICATES" in ssh-keygen(1).
Signed-off-by: Fabian Stelzer <fs@gigacodes.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
For ssh the user.signingkey can be a filename/path or even a literal ssh pubkey.
In push certs and textual output we prefer the ssh fingerprint instead.
Signed-off-by: Fabian Stelzer <fs@gigacodes.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Implements the actual sign_buffer_ssh operation and move some shared
cleanup code into a strbuf function
Set gpg.format = ssh and user.signingkey to either a ssh public key
string (like from an authorized_keys file), or a ssh key file.
If the key file or the config value itself contains only a public key
then the private key needs to be available via ssh-agent.
gpg.ssh.program can be set to an alternative location of ssh-keygen.
A somewhat recent openssh version (8.2p1+) of ssh-keygen is needed for
this feature. Since only ssh-keygen is needed it can this way be
installed seperately without upgrading your system openssh packages.
Signed-off-by: Fabian Stelzer <fs@gigacodes.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
projects / thirdparty / git.git / log
