Before, when running the "coccicheck" target, only the source files
which were being compiled would have been checked by Coccinelle.
However, just because we aren't compiling a source file doesn't mean we
have to exclude it from analysis. This will allow us to catch more
mistakes, in particular ones that affect Windows-only sources since
Coccinelle currently runs only on Linux.
Make the "coccicheck" target run on all C sources except for those that
are taken from some third-party source. We don't want to patch these
files since we want them to be as close to upstream as possible so that
it'll be easier to pull in upstream updates.
When running a build on Arch Linux with no additional flags provided,
after applying this patch, the following sources are now checked:
* block-sha1/sha1.c
* compat/access.c
* compat/basename.c
* compat/fileno.c
* compat/gmtime.c
* compat/hstrerror.c
* compat/memmem.c
* compat/mingw.c
* compat/mkdir.c
* compat/mkdtemp.c
* compat/mmap.c
* compat/msvc.c
* compat/pread.c
* compat/precompose_utf8.c
* compat/qsort.c
* compat/setenv.c
* compat/sha1-chunked.c
* compat/snprintf.c
* compat/stat.c
* compat/strcasestr.c
* compat/strdup.c
* compat/strtoimax.c
* compat/strtoumax.c
* compat/unsetenv.c
* compat/win32/dirent.c
* compat/win32/path-utils.c
* compat/win32/pthread.c
* compat/win32/syslog.c
* compat/win32/trace2_win32_process_info.c
* compat/win32mmap.c
* compat/winansi.c
* ppc/sha1.c
This also results in the following source now being excluded:
* compat/obstack.c
Instead of generating $(FOUND_C_SOURCES) from a
`$(shell $(FIND_SOURCE_FILES))` invocation, an alternative design was
considered which involved converting $(FIND_SOURCE_FILES) into
$(SOURCE_FILES) which would hold a list of filenames from the
$(FIND_SOURCE_FILES) invocation. We would simply filter `%.c` files into
$(ALL_C_SOURCES). $(SOURCE_FILES) would then be passed directly to the
etags, ctags and cscope commands. We can see from the following
invocation
$ git ls-files '*.[hcS]' '*.sh' ':!*[tp][0-9][0-9][0-9][0-9]*' ':!contrib' | wc -c
12779
that the number of characters in this list would pose a problem on
platforms with short command-line length limits (such as CMD which has a
max of 8191 characters). As a result, we don't perform this change.
However, we can see that the same issue may apply when running
Coccinelle since $(COCCI_SOURCES) is also a list of filenames:
if ! echo $(COCCI_SOURCES) | xargs $$limit \
$(SPATCH) --sp-file $< $(SPATCH_FLAGS) \
>$@+ 2>$@.log; \
This is justified since platforms that support Coccinelle generally have
reasonably long command-line length limits and so we are safe for the
foreseeable future.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, $(FIND_SOURCE_FILES) has two modes: if `git ls-files` is
present, it will use that to enumerate the files in the repository; else
it will use `$(FIND) .` to enumerate the files in the directory.
There is a subtle difference between these two methods, however. With
ls-files, filenames don't have a leading `./` while with $(FIND), they
do. This does not currently pose a problem but in a future patch, we
will be using `filter-out` to process the list of files with the
assumption that there is no prefix.
Unify the two possible invocations in $(FIND_SOURCE_FILES) by using sed
to remove the `./` prefix in the $(FIND) case.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some files in our codebase are borrowed from other projects, and
minimally updated to suit our own needs. We'd sometimes need to tell
our own sources and these third-party sources apart for management
purposes (e.g. we may want to be less strict about coding style and
other issues on third-party files).
Define the $(MAKE) variable THIRD_PARTY_SOURCES that can be used to
match names of third-party sources.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
- Replace "SHA-1" by "object name", the modern name for hashes.
- Correct a few grammar weaknesses.
- Do not accidentally format a phrase in teletype font where quotes are
intended.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff, log doc: say "patch text" instead of "patches"
A poster on Stackoverflow was confused that the documentation of git-log
promised to generate "patches" or "patch files" with -p, but there were
none to be found. Rewrite the corresponding paragraph to talk about
"patch text" to avoid the confusion.
Shorten the language to say "X does Y" in place of "X does not Z, but Y".
Cross-reference the referred-to commands like the rest of the file does.
Enumerate git-show because it includes the description as well.
Mention porcelain commands before plumbing commands because I guess that
the paragraph is read more frequently in their context.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'progress.c' has seen a few fixes recently [1], and, unfortunately,
some of those fixes required further fixes [2]. It seems it's time to
have a few tests focusing on the subtleties of the progress display.
Add the 'test-tool progress' subcommand to help testing the progress
display, reading instructions from standard input and turning them
into calls to the display_progress() and display_throughput()
functions with the given parameters.
The progress display is, however, critically dependent on timing,
because it's only updated once every second or, if the toal is known
in advance, every 1%, and there is the throughput rate as well. These
make the progress display far too undeterministic for testing as-is.
To address this, add a few testing-specific variables and functions to
'progress.c', allowing the the new test helper to:
- Disable the triggered-every-second SIGALRM and set the
'progress_update' flag explicitly based in the input instructions.
This way the progress line will be updated deterministically when
the test wants it to be updated.
- Specify the time elapsed since start_progress() to make the
throughput rate calculations deterministic.
Add the new test script 't0500-progress-display.sh' to check a few
simple cases with and without throughput, and that a shorter progress
line properly covers up the previously displayed line in different
situations.
[1] See commits 545dc345eb (progress: break too long progress bar
lines, 2019-04-12) and 9f1fd84e15 (progress: clear previous
progress update dynamically, 2019-04-12).
[2] 1aed1a5f25 (progress: avoid empty line when breaking the progress
line, 2019-05-19)
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This reverts commit 5b12e3123b (progress: use term_clear_line(),
2019-06-24), because covering up the entire last line while refreshing
the progress line caused unexpected problems during 'git
clone/fetch/push':
$ git clone ssh://localhost/home/szeder/src/tmp/linux.git/
Cloning into 'linux'...
remote:
remote:
remote:
remote: Enumerating objects: 999295
The length of the progress bar line can shorten when it includes
throughput and the unit changes, or when its length exceeds the width
of the terminal and is broken into two lines. In these cases the
previously displayed longer progress line should be covered up,
because otherwise the leftover characters from the previous progress
line make the output look weird [1]. term_clear_line() makes this
quite simple, as it covers up the entire last line either by using an
ANSI control sequence or by printing a terminal width worth of space
characters, depending on whether the terminal is smart or dumb.
Unfortunately, when accessing a remote repository via any non-local
protocol the remote 'git receive-pack/upload-pack' processes can't
possibly have any idea about the local terminal (smart of dumb? how
wide?) their progress will end up on. Consequently, they assume the
worst, i.e. standard-width dumb terminal, and print 80 spaces to cover
up the previously displayed progress line. The local 'git
clone/fetch/push' processes then display the remote's progress,
including these coverup spaces, with the 'remote: ' prefix, resulting
in a total line length of 88 characters. If the local terminal is
narrower than that, then the coverup gets line-wrapped, and after that
the CR at the end doesn't return to the beginning of the progress
line, but to the first column of its last line, resulting in those
repeated 'remote: <many-spaces>' lines.
By reverting 5b12e3123b (progress: use term_clear_line(),
2019-06-24) we won't cover up the entire last line, but go back to
comparing the length of the current progress bar line with the
previous one, and cover up as many characters as needed.
[1] See commits 545dc345eb (progress: break too long progress bar
lines, 2019-04-12) and 9f1fd84e15 (progress: clear previous
progress update dynamically, 2019-04-12).
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, $(LIB_H) is generated from two modes: if `git ls-files` is
present, it will use that to enumerate the files in the repository; else
it will use `$(FIND) .` to enumerate the files in the directory.
There is a subtle difference between these two methods, however. With
ls-files, filenames don't have a leading `./` while with $(FIND), they
do. This results in $(CHK_HDRS) having to substitute out the leading
`./` before it uses $(LIB_H).
Unify the two possible values in $(LIB_H) by using patsubst to remove the
`./` prefix at its definition.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
During git-fetch, the client checks if the advertised tags' OIDs are
already in the fetch request's want OID set. This check is done in a
linear scan. For a repository that has a lot of refs, repeating this
scan takes 15+ minutes. In order to speed this up, create a oid_set for
other refs' OIDs.
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Let commit_list_count() count the number of parents instead of
duplicating it. Also store the result in an unsigned int, as that's
what the function returns, and the count is never negative.
Signed-off-by: René Scharfe <l.s.r@web.de>
Acked-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reject values that don't fit into an int, as get_parent() and
get_nth_ancestor() cannot handle them. That's better than potentially
returning a random object.
If this restriction turns out to be too tight then we can switch to a
wider data type, but we'd still have to check for overflow.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the number gets too high for an int, weird things may happen, as
signed overflows are undefined. Add a test to show this; rev-parse
"sucessfully" interprets 100000000000000000000000000000000 to be the
same as 0, at least on x64 with GCC 9.2.1 and Clang 8.0.1, which is
obviously bogus.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We use add_excludes_from_blob_to_list() to parse a sparse blob. Since
we don't have a base path, we pass NULL and 0 for the base and baselen,
respectively. But the rest of the exclude code passes a literal empty
string instead of NULL for this case. And indeed, we eventually end up
with match_pathname() calling fspathncmp(), which then calls the system
strncmp(path, base, baselen).
This works on many platforms, which notice that baselen is 0 and do not
look at the bytes of "base" at all. But it does violate the C standard,
and building with SANITIZE=undefined will complain. You can also see it
by instrumenting fspathncmp like this:
diff --git a/dir.c b/dir.c
index d021c908e5..4bb3d3ec96 100644
--- a/dir.c
+++ b/dir.c
@@ -71,6 +71,8 @@ int fspathcmp(const char *a, const char *b)
int fspathncmp(const char *a, const char *b, size_t count)
{
+ if (!a || !b)
+ BUG("null fspathncmp arguments");
return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
}
We could perhaps be more defensive in match_pathname(), but even if we
did so, it makes sense for this code to match the rest of the exclude
callers.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The sparse:oid filter has two error modes: we might fail to resolve the
name to an OID, or we might fail to parse the contents of that OID. In
the latter case, let's give a less generic error message, and mention
the OID we did find.
While we're here, let's also mark both messages as translatable.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The list-objects-filter code has two steps to its initialization:
1. parse_list_objects_filter() makes sure the spec is a filter we know
about and is syntactically correct. This step is done by "rev-list"
or "upload-pack" that is going to apply a filter, but also by "git
clone" or "git fetch" before they send the spec across the wire.
2. list_objects_filter__init() runs the type-specific initialization
(using function pointers established in step 1). This happens at
the start of traverse_commit_list_filtered(), when we're about to
actually use the filter.
It's a good idea to parse as much as we can in step 1, in order to catch
problems early (e.g., a blob size limit that isn't a number). But one
thing we _shouldn't_ do is resolve any oids at that step (e.g., for
sparse-file contents specified by oid). In the case of a fetch, the oid
has to be resolved on the remote side.
The current code does resolve the oid during the parse phase, but
ignores any error (which we must do, because we might just be sending
the spec across the wire). This leads to two bugs:
- if we're not in a repository (e.g., because it's git-clone parsing
the spec), then we trigger a BUG() trying to resolve the name
- if we did hit the error case, we still have to notice that later and
bail. The code path in rev-list handles this, but the one in
upload-pack does not, leading to a segfault.
We can fix both by moving the oid resolution into the sparse-oid init
function. At that point we know we have a repository (because we're
about to traverse), and handling the error there fixes the segfault.
As a bonus, we can drop the NULL sparse_oid_value check in rev-list,
since this is now handled in the sparse-oid-filter init function.
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We test in t5317 that "sparse:oid" filters work with rev-list, but
there's no coverage at all confirming that they work with a fetch or
clone (and in fact, there are several bugs). Let's do a basic test that
a clone fetches the correct objects.
[jk: extracted from Jon's earlier fix patches. I also simplified the
setup down to a single sparse file, and I added checks that we got the
right blobs]
Signed-off-by: Jon Simons <jon@jonsimons.org>
Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After the previous commit, AsciiDoc and Asciidoctor render the manpage
headers identically, so we no longer need the "cut the header" part of
our `--cut-header-footer` option. We do still need the "cut the footer"
part, though. The previous commits improved the rendering of the footer
in Asciidoctor by quite a bit, but the two programs still disagree on
how to format the date in the footer: 01/01/1970 vs 1970-01-01.
We could keep using `--cut-header-footer`, but it would be nice if we
had a slightly smaller hammer `--cut-footer` that would be less likely
to hide regressions. Rather than simply adding such an option, let's
also drop `--cut-header-footer`, i.e., rework it to lose the "header"
part of its name and functionality.
`--cut-header-footer` is just a developer tool and it probably has no
more than a handful of users, so we can afford to be aggressive.
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As can be seen from the previous commit, there are three attributes that
we provide to AsciiDoc through asciidoc.conf. Asciidoctor ignores that
file. After that patch, newer versions of Asciidoctor pick up the
`manmanual` and `mansource` attributes as we invoke `asciidoctor`, but
they don't pick up `manversion`. ([1] says: "Not used by Asciidoctor.")
Older versions (<1.5.7) don't handle these attributes at all. As a
result, we are missing one or three `<refmiscinfo/>` tags in each
xml-file produced when we build with Asciidoctor.
Because of this, xmlto doesn't include the Git version number in the
rendered manpages. And in particular, with versions <1.5.7, the manpage
footers instead contain the fairly ugly "[FIXME: source]".
That Asciidoctor ignores asciidoc.conf is nothing new. This is why we
implement the `linkgit:` macro in asciidoc.conf *and* in
asciidoctor-extensions.rb. Follow suit and provide these tags in
asciidoctor-extensions.rb, using a "postprocessor" extension where we
just search and replace in the XML, treated as text.
We may consider a few alternatives:
* Inject these lines into the xml-files from the *Makefile*, e.g.,
using `sed`. That would reduce repetition, but it feels wrong to
impose another step and another risk on the AsciiDoc-processing only
to benefit the Asciidoctor-one.
* I tried providing a "docinfo processor" to inject these tags, but
could not figure out how to "merge" the two <refmeta/> sections that
resulted. To avoid xmlto barfing on the result, I needed to use
`xmlto --skip-validation ...`, which seems unfortunate.
Let's instead inject the missing tags using a postprocessor. We'll make
it fairly obvious that we aim to inject the exact same three lines of
`<refmiscinfo/>` that asciidoc.conf provides. We inject them in
*post*-processing so we need to do the variable expansion ourselves. We
do introduce the bug that asciidoc.conf already has in that we won't do
any escaping, e.g., of funky versions like "some v <2.25, >2.20".
The postprocessor we add here works on the XML as raw text and doesn't
really use the full potential of XML to do a more structured injection.
This is actually precisely what the Asciidoctor User Manual does in its
postprocessor example [2]. I looked into two other approaches:
1. The nokogiri library is apparently the "modern" way of doing XML
in ruby. I got it working fairly easily:
require 'nokogiri'
doc = Nokogiri::XML(output)
doc.search("refmeta").each { |n| n.add_child(new_tags) }
output = doc.to_xml
However, this adds another dependency (e.g., the "ruby-nokogiri"
package on Ubuntu). Using Asciidoctor is not our default, but it
will probably need to become so soon. Let's avoid adding a
dependency just so that we can say "search...add_child" rather than
"sub(regex...)".
2. The older REXML is apparently always(?) bundled with ruby, but I
couldn't even parse the original document:
require 'rexml/document'
doc = REXML::Document.new(output)
...
The error was "no implicit conversion of nil into String" and I
stopped there.
I don't think it's unlikely that doing a plain old search-and-replace
will work just as fine or better compared to parsing XML and worrying
about libraries and library versions.
[1] https://asciidoctor.org/docs/user-manual/#builtin-attributes
[2] https://asciidoctor.org/docs/user-manual/#postprocessor-example
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rather than hardcoding "Git Manual" and "Git" as the manual and source
in asciidoc.conf, provide them as attributes `manmanual` and
`mansource`. Rename the `git_version` attribute to `manversion`.
These new attribute names are not arbitrary, see, e.g., [1].
For AsciiDoc (8.6.10) and Asciidoctor <1.5.7, this is a no-op. Starting
with Asciidoctor 1.5.7, `manmanual` and `mansource` actually end up in
the xml-files and eventually in the rendered manpages. In particular,
the manpage headers now render just as with AsciiDoc.
No versions of Asciidoctor pick up the `manversion` [2], and older
versions don't pick up any of these attributes. -- We'll fix that with a
bit of a hack in the next commit.
[1] https://asciidoctor.org/docs/user-manual/#man-pages
[2] Note how [1] says "Not used by Asciidoctor".
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In b57e8119e6 (submodule: teach set-branch subcommand, 2019-02-08), the
`set-branch` subcommand was added for submodules. When the documentation
was written, the syntax for a "index term" in AsciiDoc was
accidentally used. This caused the documentation to be rendered as
set-branch -d|--default)|(-b|--branch <branch> [--] <path>
instead of
set-branch ((-d|--default)|(-b|--branch <branch>)) [--] <path>
In addition to this, the original documentation was possibly confusing
as it made it seem as if the `-b` option didn't accept a `<branch>`
argument.
Break `--default` and `--branch` into their own separate invocations to
make it obvious that these options are mutually exclusive. Also, this
removes the surrounding parentheses so that the "index term" syntax is
not triggered.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our documentation toolchain has traditionally been built around DocBook
4.5. This version of DocBook is the last DTD-based version of DocBook.
In 2009, DocBook 5 was introduced using namespaces and its syntax is
expressed in RELAX NG, which is more expressive and allows a wider
variety of syntax forms.
Asciidoctor, one of the alternatives for building our documentation,
moved support for DocBook 4.5 out of core in its recent 2.0 release and
now only supports DocBook 5 in the main release. The DocBoook 4.5
converter is still available as a separate component, but this is not
available in most distro packages. This would not be a problem but for
the fact that we use xmlto, which is still stuck in the DocBook 4.5 era.
xmlto performs DTD validation as part of the build process. This is not
problematic for DocBook 4.5, which has a valid DTD, but it clearly
cannot work for DocBook 5, since no DTD can adequately express its full
syntax. In addition, even if xmlto did support RELAX NG validation,
that wouldn't be sufficient because it uses the libxml2-based xmllint to
do so, which has known problems with validating interleaves in RELAX NG.
Fortunately, there's an easy way forward: ask Asciidoctor to use its
DocBook 5 backend and tell xmlto to skip validation. Asciidoctor has
supported DocBook 5 since v0.1.4 in 2013 and xmlto has supported
skipping validation for probably longer than that.
We also need to teach xmlto how to use the namespaced DocBook XSLT
stylesheets instead of the non-namespaced ones it usually uses.
Normally these stylesheets are interchangeable, but the non-namespaced
ones have a bug that causes them not to strip whitespace automatically
from certain elements when namespaces are in use. This results in
additional whitespace at the beginning of list elements, which is
jarring and unsightly.
We can do this by passing a custom stylesheet with the -x option that
simply imports the namespaced stylesheets via a URL. Any system with
support for XML catalogs will automatically look this URL up and
reference a local copy instead without us having to know where this
local copy is located. We know that anyone using xmlto will already
have catalogs set up properly since the DocBook 4.5 DTD used during
validation is also looked up via catalogs. All major Linux
distributions distribute the necessary stylesheets and have built-in
catalog support, and Homebrew does as well, albeit with a requirement to
set an environment variable to enable catalog support.
On the off chance that someone lacks support for catalogs, it is
possible for xmlto (via xmllint) to download the stylesheets from the
URLs in question, although this will likely perform poorly enough to
attract attention. People still have the option of using the prebuilt
documentation that we ship, so happily this should not be an impediment.
Finally, we need to filter out some messages from other stylesheets that
occur when invoking dblatex in the CI job. This tool strips namespaces
much like the unnamespaced DocBook stylesheets and prints similar
messages. If we permit these messages to be printed to standard error,
our documentation CI job will fail because we check standard error for
unexpected output. Due to dblatex's reliance on Python 2, we may need
to revisit its use in the future, in which case this problem may go
away, but this can be delayed until a future patch.
The final message we filter is due to libxslt on modern Debian and
Ubuntu. The patch which they use to implement reproducible ID
generation also prints messages about the ID generation. While this
doesn't affect our current CI images since they use Ubuntu 16.04 which
lacks this patch, if we upgrade to Ubuntu 18.04 or a modern Debian,
these messages will appear and, like the above messages, cause a CI
failure.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* 'master' of git://ozlabs.org/~paulus/gitk:
gitk: Do not mistake unchanged lines for submodule changes
gitk: Use right colour for remote refs in the "Tags and heads" dialog
gitk: Add Chinese (zh_CN) translation
gitk: Make web links clickable
Selecting whether to "Amend Last Commit" or not does not have a hotkey.
With this patch, the user may toggle between the two options with
CTRL/CMD+e.
Signed-off-by: Birger Skogeng Pedersen <birger.sp@gmail.com>
Rebased-by: Bert Wesarg <bert.wesarg@googlemail.com>
Signed-off-by: Pratyush Yadav <me@yadavpratyush.com>
Unchanged lines are prefixed with a white-space, thus unchanged lines
starting with either " <" or " >" are mistaken for submodule changes.
Check if a line starts with either " <" or " >" only if we are listing
the changes of a submodule.
Signed-off-by: Gabriele Mazzotta <gabriele.mzt@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Makes it easier to see which refs are local and which refs are remote.
Adds consistency with the remote background colour in the graph display.
Signed-off-by: Paul Wise <pabs3@bonedaddy.net>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Change the amend setting from two radio buttons ("New commit" and "Amend
commit") to a single checkbutton. The two radio buttons can never be
selected together because they are exactly the opposite of each other,
so it makes sense to change it to a single checkbutton.
* bw/amend-checkbutton:
git-gui: convert new/amend commit radiobutton to checkbutton
While the commit message widget has a configurable fixed width, it
nevertheless allowed to write commit messages which exceeded this limit.
Though there is no visual clue, that there is scrolling going on. Now
there is a horizontal scrollbar.
There seems to be a bug in at least Tcl/Tk up to version 8.6.8, which
does not update the horizontal scrollbar if one removes the whole
content at once.
Suggested-by: Birger Skogeng Pedersen <birger.sp@gmail.com>
Signed-off-by: Bert Wesarg <bert.wesarg@googlemail.com>
Signed-off-by: Pratyush Yadav <me@yadavpratyush.com>
Its a bi-state anyway and also saves one line in the menu.
Signed-off-by: Bert Wesarg <bert.wesarg@googlemail.com>
Signed-off-by: Pratyush Yadav <me@yadavpratyush.com>
Currently, _git_archive() uses a hardcoded list of options for its
completion. However, we can use __gitcomp_builtin() to get a dynamically
generated list of completions instead.
Teach _git_archive() to use __gitcomp_builtin() so that newly
implemented options in archive will be automatically completed without
any mucking around in git-completion.bash. While we're at it, teach it
to complete the missing `--worktree-attributes` option as well.
Unfortunately, since some args are passed through from cmd_archive() to
write_archive() (which calls parse_archive_args()), there's no way that a
`--git-completion-helper` arg can end up reaching parse_archive_args()
since the first call to parse_options() will end up calling exit(0). As
a result, we have to carry the options supported by write_archive() in
the hardcoded string.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, _git_rebase() uses a hardcoded list of options for its
completion. However, we can use __gitcomp_builtin() to get a dynamically
generated list of completions instead.
Teach _git_rebase() to use __gitcomp_builtin() so that newly implemented
options in rebase will be automatically completed without any mucking
around in git-completion.bash.
Signed-off-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The bash completion script knows some options to "git log" and
"git show" only in the positive form, (e.g. "--abbrev-commit"), but not
in their negative form (e.g. "--no-abbrev-commit"). Add them.
Also, the bash completion script is missing some other options to
"git diff", and "git show" (and thus, all other commands that take
"git diff"'s options). Add them. Of note, since "--indent-heuristic" is
no longer experimental, add that too.
Signed-off-by: Max Rothman <max.r.rothman@gmail.com>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the client has asked for certain shallow options like
"deepen-since", we do a custom rev-list walk that pretends to be
shallow. Before doing so, we have to disable the commit-graph, since it
is not compatible with the shallow view of the repository. That's
handled by 829a321569 (commit-graph: close_commit_graph before shallow
walk, 2018-08-20). That commit literally closes and frees our
repo->objects->commit_graph struct.
That creates an interesting problem for commits that have _already_ been
parsed using the commit graph. Their commit->object.parsed flag is set,
their commit->graph_pos is set, but their commit->maybe_tree may still
be NULL. When somebody later calls repo_get_commit_tree(), we see that
we haven't loaded the tree oid yet and try to get it from the commit
graph. But since it has been freed, we segfault!
So the root of the issue is a data dependency between the commit's
lazy-load of the tree oid and the fact that the commit graph can go
away mid-process. How can we resolve it?
There are a couple of general approaches:
1. The obvious answer is to avoid loading the tree from the graph when
we see that it's NULL. But then what do we return for the tree oid?
If we return NULL, our caller in do_traverse() will rightly
complain that we have no tree. We'd have to fallback to loading the
actual commit object and re-parsing it. That requires teaching
parse_commit_buffer() to understand re-parsing (i.e., not starting
from a clean slate and not leaking any allocated bits like parent
list pointers).
2. When we close the commit graph, walk through the set of in-memory
objects and clear any graph_pos pointers. But this means we also
have to "unparse" any such commits so that we know they still need
to open the commit object to fill in their trees. So it's no less
complicated than (1), and is more expensive (since we clear objects
we might not later need).
3. Stop freeing the commit-graph struct. Continue to let it be used
for lazy-loads of tree oids, but let upload-pack specify that it
shouldn't be used for further commit parsing.
4. Push the whole shallow rev-list out to its own sub-process, with
the commit-graph disabled from the start, giving it a clean memory
space to work from.
I've chosen (3) here. Options (1) and (2) would work, but are
non-trivial to implement. Option (4) is more expensive, and I'm not sure
how complicated it is (shelling out for the actual rev-list part is
easy, but we do then parse the resulting commits internally, and I'm not
clear which parts need to be handling shallow-ness).
The new test in t5500 triggers this segfault, but see the comments there
for how horribly intimate it has to be with how both upload-pack and
commit graphs work.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 43d3561805 (commit-graph write: don't die if the existing graph
is corrupt, 2019-03-25) added an environment variable we use only in the
test suite, $GIT_TEST_COMMIT_GRAPH_DIE_ON_LOAD. But it put the check for
this variable at the very top of prepare_commit_graph(), which is called
every time we want to use the commit graph. Most importantly, it comes
_before_ we check the fast-path "did we already try to load?", meaning
we end up calling getenv() for every single use of the commit graph,
rather than just when we load.
getenv() is allowed to have unexpected side effects, but that shouldn't
be a problem here; we're lazy-loading the graph so it's clear that at
least _one_ invocation of this function is going to call it.
But it is inefficient. getenv() typically has to do a linear search
through the environment space.
We could memoize the call, but it's simpler still to just bump the check
down to the actual loading step. That's fine for our sole user in t5318,
and produces this minor real-world speedup:
[before]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 1.460 s ± 0.017 s [User: 1.174 s, System: 0.285 s]
Range (min … max): 1.440 s … 1.491 s 10 runs
[after]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 1.391 s ± 0.005 s [User: 1.118 s, System: 0.273 s]
Range (min … max): 1.385 s … 1.399 s 10 runs
Of course that actual speedup depends on how big your environment is. We
can game it like this:
for i in $(seq 10000); do
export dummy$i=$i
done
in which case I get:
[before]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 6.257 s ± 0.061 s [User: 6.005 s, System: 0.250 s]
Range (min … max): 6.174 s … 6.337 s 10 runs
[after]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 1.403 s ± 0.005 s [User: 1.146 s, System: 0.256 s]
Range (min … max): 1.396 s … 1.412 s 10 runs
So this is really more about avoiding the pathological case than
providing a big real-world speedup.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When traverse_commit_list() processes each commit, it queues the
commit's root tree in the pending array. Then, after all commits are
processed, it calls traverse_trees_and_blobs() to walk over the pending
list, calling process_tree() on each. But if revs->tree_objects is not
set, process_tree() just exists immediately!
We can save ourselves some work by not even bothering to queue these
trees in the first place. There are a few subtle points to make:
- we also detect commits with a NULL tree pointer here. But this isn't
an interesting check for broken commits, since the lookup_tree()
we'd have done during commit parsing doesn't actually check that we
have the tree on disk. So we're not losing any robustness.
- besides queueing, we also set the NOT_USER_GIVEN flag on the tree
object. This is used by the traverse_commit_list_filtered() variant.
But if we're not exploring trees, then we won't actually care about
this flag, which is used only inside process_tree() code-paths.
- queueing trees eventually leads to us queueing blobs, too. But we
don't need to check revs->blob_objects here. Even in the current
code, we still wouldn't find those blobs, because we'd never open up
the tree objects to list their contents.
- the user-visible impact to the caller is minimal. The pending trees
are all cleared by the time the function returns anyway, by
traverse_trees_and_blobs(). We do call a show_commit() callback,
which technically could be looking at revs->pending during the
callback. But it seems like a rather unlikely thing to do (if you
want the tree of the current commit, then accessing the tree struct
member is a lot simpler).
So this should be safe to do. Let's look at the benefits:
[before]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 7.651 s ± 0.021 s [User: 7.399 s, System: 0.252 s]
Range (min … max): 7.607 s … 7.683 s 10 runs
[after]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 7.593 s ± 0.023 s [User: 7.329 s, System: 0.264 s]
Range (min … max): 7.565 s … 7.634 s 10 runs
Not too impressive, but then we're really just avoiding sticking a
pointer into a growable array. But still, I'll take a free 0.75%
speedup.
Let's try it after running "git commit-graph write":
[before]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 1.458 s ± 0.011 s [User: 1.199 s, System: 0.259 s]
Range (min … max): 1.447 s … 1.481 s 10 runs
[after]
Benchmark #1: git -C linux rev-list HEAD >/dev/null
Time (mean ± σ): 1.126 s ± 0.023 s [User: 896.5 ms, System: 229.0 ms]
Range (min … max): 1.106 s … 1.181 s 10 runs
Now that's more like it. We saved over 22% of the total time. Part of
that is because the runtime is shorter overall, but the absolute
improvement is also much larger. What's going on?
When we fill in a commit struct using the commit graph, we don't bother
to set the tree pointer, and instead lazy-load it when somebody calls
get_commit_tree(). So we're not only skipping the pointer write to the
pending queue, but we're skipping the lazy-load of the tree entirely.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Move a closing backtick that was placed one character too soon.
Signed-off-by: Cameron Steffen <cam.steffen94@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit b841d4ff43 (Add `human` format to test-tool, 2019-01-28) added
a get_time() function which allows $GIT_TEST_DATE_NOW in the
environment to override the current time. So we no longer need to
interpret that variable in cmd__date().
Therefore, we can stop passing the "now" parameter down through the
date functions, since nobody uses them. Note that we do need to make
sure all of the previous callers that took a "now" parameter are
correctly using get_time().
Signed-off-by: Stephen P. Smith <ischis2@cox.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-gui learned to revert selected lines and hunks, just like it can
stage selected lines and hunks. To provide a safety net for accidental
revert, the most recent revert can be undone.
* py/revert-hunks-lines:
git-gui: allow undoing last revert
git-gui: return early when patch fails to apply
git-gui: allow reverting selected hunk
git-gui: allow reverting selected lines
git-gui learned to switch focus between widgets "unstaged commits",
"staged commits", "diff", and "commit message" using the keyboard
shortcuts Alt+1, Alt+2, Alt+3, and Alt+4 respectively.
* bp/widget-focus-hotkeys:
git-gui: add hotkeys to set widget focus
The user cannot change focus between the list of files, the diff view and
the commit message widgets without using the mouse (clicking either of
the four widgets).
With this patch, the user may set ui focus to the previously selected path
in either the "Unstaged Changes" or "Staged Changes" widgets, using
ALT+1 or ALT+2.
The user may also set the ui focus to the diff view widget with
ALT+3, or to the commit message widget with ALT+4.
This enables the user to select/unselect files, view the diff and create a
commit in git-gui using keyboard-only.
Signed-off-by: Birger Skogeng Pedersen <birger.sp@gmail.com>
Signed-off-by: Pratyush Yadav <me@yadavpratyush.com>
The cache-tree datastructure is used to speed up the comparison
between the HEAD and the index, and when the index is updated by
a cherry-pick (for example), a tree object that would represent
the paths in the index in a directory is constructed in-core, to
see if such a tree object exists already in the object store.
When the lazy-fetch mechanism was introduced, we converted this
"does the tree exist?" check into an "if it does not, and if we
lazily cloned, see if the remote has it" call by mistake. Since
the whole point of this check is to repair the cache-tree by
recording an already existing tree object opportunistically, we
shouldn't even try to fetch one from the remote.
Pass the OBJECT_INFO_SKIP_FETCH_OBJECT flag to make sure we only
check for existence in the local object store without triggering the
lazy fetch mechanism.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
[jc: rewritten the proposed log message]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "git am" based backend of "git rebase" ignored the result of
updating ".gitattributes" done in one step when replaying
subsequent steps.
* bc/reread-attributes-during-rebase:
am: reload .gitattributes after patching it
path: add a function to check for path suffix
"for-each-ref" and friends that shows refs did not protect themselves
against ancient tags that did not record tagger names when asked to
show "%(taggername)", which have been corrected.
* mp/for-each-ref-missing-name-or-email:
ref-filter: initialize empty name or email fields
