* [PATCH] t/t7601: Modernize test scripts using functions
From: Dorcas AnonoLitunya @ 2023-10-16 15:21 UTC (permalink / raw)
To: christian.couder; +Cc: anonolitunya, git, gitster
The test script is currently using the command format 'test -f' to
check for existence or absence of files.
Replace it with new helper functions following the format
'test_path_is_file'.
Consequently, the patch also replaces the inverse command '! test -f' or
'test ! -f' with new helper function following the format
'test_path_is_missing'
This adjustment using helper functions makes the code more readable and
easier to understand.
Signed-off-by: Dorcas AnonoLitunya <anonolitunya@gmail.com>
---
t/t7601-merge-pull-config.sh | 24 ++++++++++++------------
1 file changed, 12 insertions(+), 12 deletions(-)
diff --git a/t/t7601-merge-pull-config.sh b/t/t7601-merge-pull-config.sh
index bd238d89b0..e08767df66 100755
--- a/t/t7601-merge-pull-config.sh
+++ b/t/t7601-merge-pull-config.sh
@@ -349,13 +349,13 @@ test_expect_success 'Cannot rebase with multiple heads' '
test_expect_success 'merge c1 with c2' '
git reset --hard c1 &&
- test -f c0.c &&
- test -f c1.c &&
- test ! -f c2.c &&
- test ! -f c3.c &&
+ test_path_is_file c0.c &&
+ test_path_is_file c1.c &&
+ test_path_is_missing c2.c &&
+ test_path_is_missing c3.c &&
git merge c2 &&
- test -f c1.c &&
- test -f c2.c
+ test_path_is_file c1.c &&
+ test_path_is_file c2.c
'
test_expect_success 'fast-forward pull succeeds with "true" in pull.ff' '
@@ -411,8 +411,8 @@ test_expect_success 'merge c1 with c2 (ours in pull.twohead)' '
git reset --hard c1 &&
git config pull.twohead ours &&
git merge c2 &&
- test -f c1.c &&
- ! test -f c2.c
+ test_path_is_file c1.c &&
+ test_path_is_missing c2.c
'
test_expect_success 'merge c1 with c2 and c3 (recursive in pull.octopus)' '
@@ -431,10 +431,10 @@ test_expect_success 'merge c1 with c2 and c3 (recursive and octopus in pull.octo
test "$(git rev-parse c2)" = "$(git rev-parse HEAD^2)" &&
test "$(git rev-parse c3)" = "$(git rev-parse HEAD^3)" &&
git diff --exit-code &&
- test -f c0.c &&
- test -f c1.c &&
- test -f c2.c &&
- test -f c3.c
+ test_path_is_file c0.c &&
+ test_path_is_file c1.c &&
+ test_path_is_file c2.c &&
+ test_path_is_file c3.c
'
conflict_count()
--
2.42.0.345.gaab89be2eb
^ permalink raw reply related
* [[Outreachy]PATCH v2] t/t7601: Modernize test scripts using functions
From: Dorcas AnonoLitunya @ 2023-10-16 15:50 UTC (permalink / raw)
To: christian.couder; +Cc: anonolitunya, git, gitster
The test script is currently using the command format 'test -f' to
check for existence or absence of files.
Replace it with new helper functions following the format
'test_path_is_file'.
Consequently, the patch also replaces the inverse command '! test -f' or
'test ! -f' with new helper function following the format
'test_path_is_missing'
This adjustment using helper functions makes the code more readable and
easier to understand.
Signed-off-by: Dorcas AnonoLitunya <anonolitunya@gmail.com>
---
Changes in v2:
- Add Outreachy to subject line
t/t7601-merge-pull-config.sh | 24 ++++++++++++------------
1 file changed, 12 insertions(+), 12 deletions(-)
diff --git a/t/t7601-merge-pull-config.sh b/t/t7601-merge-pull-config.sh
index bd238d89b0..e08767df66 100755
--- a/t/t7601-merge-pull-config.sh
+++ b/t/t7601-merge-pull-config.sh
@@ -349,13 +349,13 @@ test_expect_success 'Cannot rebase with multiple heads' '
test_expect_success 'merge c1 with c2' '
git reset --hard c1 &&
- test -f c0.c &&
- test -f c1.c &&
- test ! -f c2.c &&
- test ! -f c3.c &&
+ test_path_is_file c0.c &&
+ test_path_is_file c1.c &&
+ test_path_is_missing c2.c &&
+ test_path_is_missing c3.c &&
git merge c2 &&
- test -f c1.c &&
- test -f c2.c
+ test_path_is_file c1.c &&
+ test_path_is_file c2.c
'
test_expect_success 'fast-forward pull succeeds with "true" in pull.ff' '
@@ -411,8 +411,8 @@ test_expect_success 'merge c1 with c2 (ours in pull.twohead)' '
git reset --hard c1 &&
git config pull.twohead ours &&
git merge c2 &&
- test -f c1.c &&
- ! test -f c2.c
+ test_path_is_file c1.c &&
+ test_path_is_missing c2.c
'
test_expect_success 'merge c1 with c2 and c3 (recursive in pull.octopus)' '
@@ -431,10 +431,10 @@ test_expect_success 'merge c1 with c2 and c3 (recursive and octopus in pull.octo
test "$(git rev-parse c2)" = "$(git rev-parse HEAD^2)" &&
test "$(git rev-parse c3)" = "$(git rev-parse HEAD^3)" &&
git diff --exit-code &&
- test -f c0.c &&
- test -f c1.c &&
- test -f c2.c &&
- test -f c3.c
+ test_path_is_file c0.c &&
+ test_path_is_file c1.c &&
+ test_path_is_file c2.c &&
+ test_path_is_file c3.c
'
conflict_count()
--
2.42.0.345.gaab89be2eb
^ permalink raw reply related
* (no subject)
From: Dorcas Litunya @ 2023-10-16 15:57 UTC (permalink / raw)
To: christian.couder; +Cc: anonolitunya, git, gitster
Bcc:
Subject: Re: [PATCH] t/t7601: Modernize test scripts using functions
Reply-To:
In-Reply-To: <20231016152113.135970-1-anonolitunya@gmail.com>
On Mon, Oct 16, 2023 at 06:21:00PM +0300, Dorcas AnonoLitunya wrote:
> The test script is currently using the command format 'test -f' to
> check for existence or absence of files.
>
> Replace it with new helper functions following the format
> 'test_path_is_file'.
>
> Consequently, the patch also replaces the inverse command '! test -f' or
> 'test ! -f' with new helper function following the format
> 'test_path_is_missing'
>
> This adjustment using helper functions makes the code more readable and
> easier to understand.
>
> Signed-off-by: Dorcas AnonoLitunya <anonolitunya@gmail.com>
> ---
> t/t7601-merge-pull-config.sh | 24 ++++++++++++------------
> 1 file changed, 12 insertions(+), 12 deletions(-)
>
> diff --git a/t/t7601-merge-pull-config.sh b/t/t7601-merge-pull-config.sh
> index bd238d89b0..e08767df66 100755
> --- a/t/t7601-merge-pull-config.sh
> +++ b/t/t7601-merge-pull-config.sh
> @@ -349,13 +349,13 @@ test_expect_success 'Cannot rebase with multiple heads' '
>
> test_expect_success 'merge c1 with c2' '
> git reset --hard c1 &&
> - test -f c0.c &&
> - test -f c1.c &&
> - test ! -f c2.c &&
> - test ! -f c3.c &&
> + test_path_is_file c0.c &&
> + test_path_is_file c1.c &&
> + test_path_is_missing c2.c &&
> + test_path_is_missing c3.c &&
> git merge c2 &&
> - test -f c1.c &&
> - test -f c2.c
> + test_path_is_file c1.c &&
> + test_path_is_file c2.c
> '
>
> test_expect_success 'fast-forward pull succeeds with "true" in pull.ff' '
> @@ -411,8 +411,8 @@ test_expect_success 'merge c1 with c2 (ours in pull.twohead)' '
> git reset --hard c1 &&
> git config pull.twohead ours &&
> git merge c2 &&
> - test -f c1.c &&
> - ! test -f c2.c
> + test_path_is_file c1.c &&
> + test_path_is_missing c2.c
> '
>
> test_expect_success 'merge c1 with c2 and c3 (recursive in pull.octopus)' '
> @@ -431,10 +431,10 @@ test_expect_success 'merge c1 with c2 and c3 (recursive and octopus in pull.octo
> test "$(git rev-parse c2)" = "$(git rev-parse HEAD^2)" &&
> test "$(git rev-parse c3)" = "$(git rev-parse HEAD^3)" &&
> git diff --exit-code &&
> - test -f c0.c &&
> - test -f c1.c &&
> - test -f c2.c &&
> - test -f c3.c
> + test_path_is_file c0.c &&
> + test_path_is_file c1.c &&
> + test_path_is_file c2.c &&
> + test_path_is_file c3.c
> '
>
> conflict_count()
> --
> 2.42.0.345.gaab89be2eb
>
Please ignore this version as I had not indicated Outreachy in the
subject header. Thanks.
^ permalink raw reply
* Re: [PATCH v2 0/3] rev-list: add support for commits in `--missing`
From: Junio C Hamano @ 2023-10-16 16:24 UTC (permalink / raw)
To: Karthik Nayak; +Cc: git, ps
In-Reply-To: <20231016103830.56486-1-karthik.188@gmail.com>
Karthik Nayak <karthik.188@gmail.com> writes:
> ... some extra checks here, especially because we don't want to
> allow missing commit's tree and parent to be parsed.
Good changes relative to the previous round.
A bad news is that with this series (especially [PATCH 3/3]) applied
on top of Patrick's "always make sure the commit we found from the
commit-graph at least exists" change, t5310 and t5320 seem to
consistently fail for me. They pass with the first two steps
applied without [3/3] but that is to be expected as they are both
no-ops.
The change in 3/3 to list-objects.c::do_traverse() seems to be the
culprit. Reverting that single hunk makes t5310 and t5320 pass
again. What I find alarming is that two tests that are touched by
this series, t5318 and t6022, do not seem to fail with the hunk
reverted, which means the hunk has no meaningful test coverage for
the purpose of this series.
I didn't even try to understand what is going on, so there may be a
very good reason that the change must be there for do_traverse() to
work correctly. I dunno.
> +- if (commit->object.flags & ADDED)
> ++ if (commit->object.flags & ADDED || commit->object.flags & MISSING)
This and others are syntactically correct C, but some folks may find
it more readable to see each of the bitwise operations enclosed in a
pair of parentheses when combined by logical operations, i.e.,
if ((commit->object.flags & ADDED) || (commit->object.flags & MISSING))
In this particular case, we can even say
if (commit->object.flags & (ADDED | MISSING))
meaning, "if we have either the ADDED or the MISSING bit set", which
may make it even clearer.
Thanks.
^ permalink raw reply
* Re: [PATCH v8 2.5/3] fixup! unit tests: add TAP unit test framework
From: Junio C Hamano @ 2023-10-16 16:41 UTC (permalink / raw)
To: steadmon; +Cc: Phillip Wood, calvinwan, git, linusa, rsbecker
In-Reply-To: <20231016134421.21659-1-phillip.wood123@gmail.com>
Phillip Wood <phillip.wood123@gmail.com> writes:
> From: Phillip Wood <phillip.wood@dunelm.org.uk>
>
> Here are a couple of cleanups for the unit test framework that I
> noticed.
Thanks. I trust that this will be squashed into the next update,
but in the meantime, I'll include it in the copy of the series I
have (without squashing). Here is another one I noticed.
----- >8 --------- >8 --------- >8 -----
Subject: [PATCH] fixup! ci: run unit tests in CI
A CI job failed due to contrib/coccinelle/equals-null.cocci
and suggested this change, which seems sensible.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
t/unit-tests/t-strbuf.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/t/unit-tests/t-strbuf.c b/t/unit-tests/t-strbuf.c
index c2fcb0cbd6..8388442426 100644
--- a/t/unit-tests/t-strbuf.c
+++ b/t/unit-tests/t-strbuf.c
@@ -28,7 +28,7 @@ static void setup_populated(void (*f)(struct strbuf*, void*), char *init_str, vo
static int assert_sane_strbuf(struct strbuf *buf)
{
/* Initialized strbufs should always have a non-NULL buffer */
- if (buf->buf == NULL)
+ if (!buf->buf)
return 0;
/* Buffers should always be NUL-terminated */
if (buf->buf[buf->len] != '\0')
--
2.42.0-398-ga9ecda2788
^ permalink raw reply related
* Re: [PATCH] t/t7601: Modernize test scripts using functions
From: Junio C Hamano @ 2023-10-16 16:53 UTC (permalink / raw)
To: Dorcas AnonoLitunya; +Cc: christian.couder, git
In-Reply-To: <20231016152113.135970-1-anonolitunya@gmail.com>
Dorcas AnonoLitunya <anonolitunya@gmail.com> writes:
> Subject: Re: [PATCH] t/t7601: Modernize test scripts using functions
Let's try if we can pack a bit more information. For example
Subject: [PATCH] t7601: use "test_path_is_file" etc. instead of "test -f"
would clarify what kind of modernization is done by this patch.
> The test script is currently using the command format 'test -f' to
> check for existence or absence of files.
"is currently using" -> "uses".
> Replace it with new helper functions following the format
> 'test_path_is_file'.
I am not sure what role "the format" plays in this picture.
test_path_is_file is not new---it has been around for quite a while.
> Consequently, the patch also replaces the inverse command '! test -f' or
> 'test ! -f' with new helper function following the format
> 'test_path_is_missing'
A bit more on this later.
> This adjustment using helper functions makes the code more readable and
> easier to understand.
Looking good. If I were writing this, I'll make the whole thing
more like this, though:
t7601: use "test_path_is_file" etc. instead of "test -f"
Some tests in t7601 use "test -f" and "test ! -f" to see if a
path exists or is missing. Use test_path_is_file and
test_path_is_missing helper functions to clarify these tests a
bit better. This especially matters for the "missing" case,
because "test ! -f F" will be happy if "F" exists as a
directory, but the intent of the test is that "F" should not
exist, even as a directory.
> diff --git a/t/t7601-merge-pull-config.sh b/t/t7601-merge-pull-config.sh
> index bd238d89b0..e08767df66 100755
> --- a/t/t7601-merge-pull-config.sh
> +++ b/t/t7601-merge-pull-config.sh
> @@ -349,13 +349,13 @@ test_expect_success 'Cannot rebase with multiple heads' '
>
> test_expect_success 'merge c1 with c2' '
> git reset --hard c1 &&
> - test -f c0.c &&
> - test -f c1.c &&
> - test ! -f c2.c &&
> - test ! -f c3.c &&
> + test_path_is_file c0.c &&
> + test_path_is_file c1.c &&
> + test_path_is_missing c2.c &&
> + test_path_is_missing c3.c &&
The original says "We are happy if c2.c is not a file", so it would
have been happy if by some mistake "git reset" created a directory
there. But the _intent_ of the test is that we do not have anything
at c2.c, and the updated code expresses it better.
^ permalink raw reply
* (no subject)
From: Dorcas Litunya @ 2023-10-16 18:43 UTC (permalink / raw)
To: Junio C Hamano; +Cc: christian.couder, git
Bcc:
Subject: Re: [PATCH] t/t7601: Modernize test scripts using functions
Reply-To:
In-Reply-To: <xmqq1qdumrto.fsf@gitster.g>
On Mon, Oct 16, 2023 at 09:53:55AM -0700, Junio C Hamano wrote:
> Dorcas AnonoLitunya <anonolitunya@gmail.com> writes:
>
> > Subject: Re: [PATCH] t/t7601: Modernize test scripts using functions
>
> Let's try if we can pack a bit more information. For example
>
> Subject: [PATCH] t7601: use "test_path_is_file" etc. instead of "test -f"
>
> would clarify what kind of modernization is done by this patch.
>
> > The test script is currently using the command format 'test -f' to
> > check for existence or absence of files.
>
> "is currently using" -> "uses".
>
> > Replace it with new helper functions following the format
> > 'test_path_is_file'.
>
> I am not sure what role "the format" plays in this picture.
> test_path_is_file is not new---it has been around for quite a while.
>
> > Consequently, the patch also replaces the inverse command '! test -f' or
> > 'test ! -f' with new helper function following the format
> > 'test_path_is_missing'
>
> A bit more on this later.
>
So should I replace this in the next version or leave this as is?
> > This adjustment using helper functions makes the code more readable and
> > easier to understand.
>
> Looking good. If I were writing this, I'll make the whole thing
> more like this, though:
>
> t7601: use "test_path_is_file" etc. instead of "test -f"
>
> Some tests in t7601 use "test -f" and "test ! -f" to see if a
> path exists or is missing. Use test_path_is_file and
> test_path_is_missing helper functions to clarify these tests a
> bit better. This especially matters for the "missing" case,
> because "test ! -f F" will be happy if "F" exists as a
> directory, but the intent of the test is that "F" should not
> exist, even as a directory.
>
>
> > diff --git a/t/t7601-merge-pull-config.sh b/t/t7601-merge-pull-config.sh
> > index bd238d89b0..e08767df66 100755
> > --- a/t/t7601-merge-pull-config.sh
> > +++ b/t/t7601-merge-pull-config.sh
> > @@ -349,13 +349,13 @@ test_expect_success 'Cannot rebase with multiple heads' '
> >
> > test_expect_success 'merge c1 with c2' '
> > git reset --hard c1 &&
> > - test -f c0.c &&
> > - test -f c1.c &&
> > - test ! -f c2.c &&
> > - test ! -f c3.c &&
> > + test_path_is_file c0.c &&
> > + test_path_is_file c1.c &&
> > + test_path_is_missing c2.c &&
> > + test_path_is_missing c3.c &&
>
> The original says "We are happy if c2.c is not a file", so it would
> have been happy if by some mistake "git reset" created a directory
> there. But the _intent_ of the test is that we do not have anything
> at c2.c, and the updated code expresses it better.
^ permalink raw reply
* Re: [PATCH v2 0/3] rev-list: add support for commits in `--missing`
From: Karthik Nayak @ 2023-10-16 19:01 UTC (permalink / raw)
To: Junio C Hamano; +Cc: git, ps
In-Reply-To: <xmqqbkcyo7ro.fsf@gitster.g>
On Mon, Oct 16, 2023 at 6:24 PM Junio C Hamano <gitster@pobox.com> wrote:
> Good changes relative to the previous round.
>
> A bad news is that with this series (especially [PATCH 3/3]) applied
> on top of Patrick's "always make sure the commit we found from the
> commit-graph at least exists" change, t5310 and t5320 seem to
> consistently fail for me. They pass with the first two steps
> applied without [3/3] but that is to be expected as they are both
> no-ops.
>
> The change in 3/3 to list-objects.c::do_traverse() seems to be the
> culprit. Reverting that single hunk makes t5310 and t5320 pass
> again.
It seems that the new chunk introduced now causes collision because of
the bit field of the MISSING flag being the same as the pre-existing
NEEDS_BITMAP flag. Earlier this wasn't a concern, but now, their paths
have converged at this change.
So changing the bit field for the MISSING flag from `22` to an unused `28`
fixes the issue. I should have run the whole test suite again. Apologies.
> What I find alarming is that two tests that are touched by
> this series, t5318 and t6022, do not seem to fail with the hunk
> reverted, which means the hunk has no meaningful test coverage for
> the purpose of this series.
>
Well, actually the newly introduced tests t6022 require this block,
but this is specific
to when commit graph is enabled. So what you did see was correct, but
the test would
also fail with `GIT_TEST_COMMIT_GRAPH=1` if the block was removed. That's
exactly why in this series I increased the number of commits in the
test block from 2->3.
> > +- if (commit->object.flags & ADDED)
> > ++ if (commit->object.flags & ADDED || commit->object.flags & MISSING)
>
> This and others are syntactically correct C, but some folks may find
> it more readable to see each of the bitwise operations enclosed in a
> pair of parentheses when combined by logical operations, i.e.,
>
> if ((commit->object.flags & ADDED) || (commit->object.flags & MISSING))
>
> In this particular case, we can even say
>
> if (commit->object.flags & (ADDED | MISSING))
>
> meaning, "if we have either the ADDED or the MISSING bit set", which
> may make it even clearer.
I agree with this, I'll add it in the next block.
Thanks for the quick review, I'll wait a day/two and send v3 with the
changes to fix tests
and cleaner code.
^ permalink raw reply
* Git does not work on my machine!!!
From: jtrites @ 2023-10-16 20:20 UTC (permalink / raw)
To: git; +Cc: John Trites
I would appreciate some help resolving this issue!!!
JohnTrites@MSI MINGW64 /c/Users/JohnTrites
$ git config --global user.name "JohnTrites"
error: could not lock config file C:/Users/John
Trites/AppData/Roaming/SPB_16.6/.gitconfig: No such file or directory
^ permalink raw reply
* Re: [PATCH v2 0/3] rev-list: add support for commits in `--missing`
From: Junio C Hamano @ 2023-10-16 20:33 UTC (permalink / raw)
To: Karthik Nayak; +Cc: git, ps
In-Reply-To: <CAOLa=ZT4EuoB8GHMe+a2nfq8Pfg5x7xzrEa_qV39U1HqUyS+Eg@mail.gmail.com>
Karthik Nayak <karthik.188@gmail.com> writes:
> Well, actually the newly introduced tests t6022 require this block,
> but this is specific
> to when commit graph is enabled.
Ah, of course.
> Thanks for the quick review, I'll wait a day/two and send v3 with the
> changes to fix tests
> and cleaner code.
Thanks.
^ permalink raw reply
* Re: Method for Calculating Statistics of Developer Contribution to a Specified Branch.
From: brian m. carlson @ 2023-10-16 21:25 UTC (permalink / raw)
To: Hongyi Zhao; +Cc: Git List
In-Reply-To: <CAGP6POKg4mSFv-Z+dD1aXDFDbxH9Xu1WCdCA5TGfCAM3NUUQLw@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 2070 bytes --]
On 2023-10-16 at 14:10:01, Hongyi Zhao wrote:
> Dear Git Mailing List,
>
> I am a developer currently working on a project and I wanted to
> establish statistics for each team member's contribution to a specific
> branch.
>
> Say, for a user "JianboLin", I am currently using the following method:
>
> $ git clone https://github.com/OrderN/CONQUEST-release.git
> $ cd CONQUEST-release
> $ git log --author="JianboLin" --stat --summary origin/f-mlff | awk
> 'NF ==4 && $2 =="|" && $3 ~/[0-9]+/ && $4 ~/[+-]+|[+]+|[-]+/ {s+=$3}
> END {print s}'
>
> Using the above command, I am able to calculate the number of lines
> contributed by a specific author on a specific branch, which allows me
> to quantify the contribution to a branch by each team member.
>
> However, I would like to know if a more efficient or accurate method
> exists to carry out this task. Are there any other parameters,
> commands, or aspects I need to consider to get a more comprehensive
> measure of contribution?
Can you maybe explain what you want to measure and what your goal is in
doing so?
The problem is that lines of code isn't really that useful as a measure
of contribution value or developer productivity, which are the reasons
people typically measure that metric. For example, with three lines, a
colleague fixed a persistently difficult-to-reproduce problem which had
been affecting many of our largest customers. That was a very valuable
contribution, but not very large. I've made similar kinds of changes
myself, both at work and in open source projects.
Certainly you can compute the number of lines of code changed by a
developer, but that is not typically a very useful metric, since it
doesn't lead you to any interesting conclusions about the benefits or
value of the contributions or developer in question. However, perhaps
you have a different goal in mind, and if you can explain what that is,
we may be able to help you find a better way of doing it.
--
brian m. carlson (he/him or they/them)
Toronto, Ontario, CA
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 263 bytes --]
^ permalink raw reply
* [PATCH v3 0/1] bugreport: include +i in outfile suffix as needed
From: Jacob Stopak @ 2023-10-16 21:40 UTC (permalink / raw)
To: git; +Cc: Jacob Stopak, Junio C Hamano
In-Reply-To: <20231015034238.100675-2-jacob@initialcommit.io>
Similar functionality to the previous patch version but with the
following changes:
* Remove the default_option_suffix variable and clean up the way the
option_suffix value is set.
* Refactor code for building report path and diagnostics zip file path
into the new function build_path(...), which builds a fresh path
from scratch each time it's called. Although it does take quite a few
arguments, it reduces duplicated code and simplifies the code by
removing the need for splicing or inserting the incrementable suffix.
* Allow the increment value 'i' to grow as needed instead of stopping
at 9.
* Replace xopen() with open() so that the program doesn't die with an
error if the file already exists.
* Replace the while loop with a fallback loop to prevent TOCTOU.
* Clean up commit message verbiage.
Some additional comments:
> Perhaps we should refine the error message we give in this case and
> we are done, then?
I had thought about this but due to the variable timed behavior I had
trouble coming up with a message that would convey clearly what's going
on and what the user should do. Here were some things that popped into
my head but they all sounded a bit silly to me:
"A bugreport with this name already exists, try again shortly"
or
"File exists: wait 1 minute and try again or use a custom suffix with -s"
or
"File exists: try again in 1 to 60 seconds :-P"
I think the reason they sound silly to me is that the user is only being
made aware of this info because the program happens to be operating this
way - instead of the program working in a smoother way where this type of
operational info wouldn't need to be conveyed.
> What downside do you see in using 2023-10-15-1008+10 after you tried
> 9 of them? The code to limit the upper bound smells like a wasted
> effort that helps nobody in practice because it is "unlikely".
> And limiting the upper bound also means you now have to have extra
> code to deal with "we ran out---error out and help the user how to
> recover" anyway.
I completely agree with this and made these updates.
> Notice that you said "in a single minute" without "calendar" and the
> sentence is perfectly understandable?
I googled "calendar minute" and the only thing that came up was some
Java documentation... So I guess I just made it up... :'D
Jacob Stopak (1):
bugreport: include +i in outfile suffix as needed
builtin/bugreport.c | 83 +++++++++++++++++++++++++++++++--------------
1 file changed, 57 insertions(+), 26 deletions(-)
base-commit: 493f4622739e9b64f24b465b21aa85870dd9dc09
--
2.42.0.297.g36452639b8
^ permalink raw reply
* [PATCH v3 1/1] bugreport: include +i in outfile suffix as needed
From: Jacob Stopak @ 2023-10-16 21:40 UTC (permalink / raw)
To: git; +Cc: Jacob Stopak
In-Reply-To: <20231016214045.146862-1-jacob@initialcommit.io>
When the -s flag is absent, git bugreport includes the current hour and
minute values in the default bugreport filename (and diagnostics zip
filename if --diagnose is supplied).
If a user runs the bugreport command more than once within a minute, a
filename conflict with an existing file occurs and the program errors,
since the new output filename was already used for the previous file. If
the user waits anywhere from 1 to 60 seconds (depending on when during
the minute the first command was run) the command works again with no
error since the default filename is now unique, and multiple bug reports
are able to be created with default settings.
This is a minor thing but can cause confusion for first time users of
the bugreport command, who are likely to run it multiple times in quick
succession to learn how it works, (like I did). Or users who quickly
fill in a few details before closing and creating a new one.
Add a '+i' into the bugreport filename suffix where 'i' is an integer
starting at 1 and growing as needed until a unique filename is obtained.
This leads to default output filenames like:
git-bugreport-%Y-%m-%d-%H%M+1.txt
git-bugreport-%Y-%m-%d-%H%M+2.txt
...
git-bugreport-%Y-%m-%d-%H%M+i.txt
This means the user will end up with multiple bugreport files being
created if they run the command multiple times quickly, but that feels
more intuitive and consistent than an error arbitrarily occuring within
a minute, especially given that the time window in which the error
currently occurs is variable as described above.
If --diagnose is supplied, match the incremented suffix of the
diagnostics zip file to the bugreport.
Signed-off-by: Jacob Stopak <jacob@initialcommit.io>
---
builtin/bugreport.c | 83 +++++++++++++++++++++++++++++++--------------
1 file changed, 57 insertions(+), 26 deletions(-)
diff --git a/builtin/bugreport.c b/builtin/bugreport.c
index d2ae5c305d..ed65735873 100644
--- a/builtin/bugreport.c
+++ b/builtin/bugreport.c
@@ -11,6 +11,7 @@
#include "diagnose.h"
#include "object-file.h"
#include "setup.h"
+#include "dir.h"
static void get_system_info(struct strbuf *sys_info)
{
@@ -97,20 +98,41 @@ static void get_header(struct strbuf *buf, const char *title)
strbuf_addf(buf, "\n\n[%s]\n", title);
}
+static void build_path(struct strbuf *buf, const char *dir_path,
+ const char *prefix, const char *suffix,
+ time_t t, int *i, const char *ext)
+{
+ struct tm tm;
+
+ strbuf_reset(buf);
+ strbuf_addstr(buf, dir_path);
+ strbuf_complete(buf, '/');
+
+ strbuf_addstr(buf, prefix);
+ strbuf_addftime(buf, suffix, localtime_r(&t, &tm), 0, 0);
+
+ if (*i > 0)
+ strbuf_addf(buf, "+%d", *i);
+
+ strbuf_addstr(buf, ext);
+
+ (*i)++;
+}
+
int cmd_bugreport(int argc, const char **argv, const char *prefix)
{
struct strbuf buffer = STRBUF_INIT;
struct strbuf report_path = STRBUF_INIT;
int report = -1;
time_t now = time(NULL);
- struct tm tm;
enum diagnose_mode diagnose = DIAGNOSE_NONE;
char *option_output = NULL;
- char *option_suffix = "%Y-%m-%d-%H%M";
+ char *option_suffix = "";
+ int option_suffix_is_from_user = 0;
const char *user_relative_path = NULL;
char *prefixed_filename;
- size_t output_path_len;
int ret;
+ int i = 0;
const struct option bugreport_options[] = {
OPT_CALLBACK_F(0, "diagnose", &diagnose, N_("mode"),
@@ -126,16 +148,16 @@ int cmd_bugreport(int argc, const char **argv, const char *prefix)
argc = parse_options(argc, argv, prefix, bugreport_options,
bugreport_usage, 0);
+ if (!strlen(option_suffix))
+ option_suffix = "%Y-%m-%d-%H%M";
+ else
+ option_suffix_is_from_user = 1;
+
/* Prepare the path to put the result */
prefixed_filename = prefix_filename(prefix,
option_output ? option_output : "");
- strbuf_addstr(&report_path, prefixed_filename);
- strbuf_complete(&report_path, '/');
- output_path_len = report_path.len;
-
- strbuf_addstr(&report_path, "git-bugreport-");
- strbuf_addftime(&report_path, option_suffix, localtime_r(&now, &tm), 0, 0);
- strbuf_addstr(&report_path, ".txt");
+ build_path(&report_path, prefixed_filename, "git-bugreport-",
+ option_suffix, now, &i, ".txt");
switch (safe_create_leading_directories(report_path.buf)) {
case SCLD_OK:
@@ -146,20 +168,6 @@ int cmd_bugreport(int argc, const char **argv, const char *prefix)
report_path.buf);
}
- /* Prepare diagnostics, if requested */
- if (diagnose != DIAGNOSE_NONE) {
- struct strbuf zip_path = STRBUF_INIT;
- strbuf_add(&zip_path, report_path.buf, output_path_len);
- strbuf_addstr(&zip_path, "git-diagnostics-");
- strbuf_addftime(&zip_path, option_suffix, localtime_r(&now, &tm), 0, 0);
- strbuf_addstr(&zip_path, ".zip");
-
- if (create_diagnostics_archive(&zip_path, diagnose))
- die_errno(_("unable to create diagnostics archive %s"), zip_path.buf);
-
- strbuf_release(&zip_path);
- }
-
/* Prepare the report contents */
get_bug_template(&buffer);
@@ -169,14 +177,37 @@ int cmd_bugreport(int argc, const char **argv, const char *prefix)
get_header(&buffer, _("Enabled Hooks"));
get_populated_hooks(&buffer, !startup_info->have_repository);
- /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
- report = xopen(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
+ again:
+ /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
+ report = open(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
+ if (report < 0 && errno == EEXIST && !option_suffix_is_from_user) {
+ build_path(&report_path, prefixed_filename,
+ "git-bugreport-", option_suffix, now, &i,
+ ".txt");
+ goto again;
+ } else if (report < 0) {
+ die_errno(_("unable to open '%s'"), report_path.buf);
+ }
if (write_in_full(report, buffer.buf, buffer.len) < 0)
die_errno(_("unable to write to %s"), report_path.buf);
close(report);
+ /* Prepare diagnostics, if requested */
+ if (diagnose != DIAGNOSE_NONE) {
+ struct strbuf zip_path = STRBUF_INIT;
+ i--; /* Undo last increment to match zipfile suffix to bugreport */
+ build_path(&zip_path, prefixed_filename, "git-diagnostics-",
+ option_suffix, now, &i, ".zip");
+
+ if (create_diagnostics_archive(&zip_path, diagnose))
+ die_errno(_("unable to create diagnostics archive %s"),
+ zip_path.buf);
+
+ strbuf_release(&zip_path);
+ }
+
/*
* We want to print the path relative to the user, but we still need the
* path relative to us to give to the editor.
--
2.42.0.297.g36452639b8
^ permalink raw reply related
* Re: [PATCH 00/25] Documentation fixes
From: Junio C Hamano @ 2023-10-16 21:54 UTC (permalink / raw)
To: Elijah Newren via GitGitGadget; +Cc: git, Elijah Newren
In-Reply-To: <pull.1595.git.1696747527.gitgitgadget@gmail.com>
"Elijah Newren via GitGitGadget" <gitgitgadget@gmail.com> writes:
> I did review every single change here, multiple times, and I have tried to
> split up this series in a way to make it easier to review. In particular I:
> ...
> (Note: every patch in this series, except for the whitespace fixes patch,
> are best viewed with --color-words.)
I didn't think of anything clever, so ended up reading a bit by bit
over days. I didn't find anything glaringly wrong ;-)
Let's mark it for 'next'.
Thanks.
^ permalink raw reply
* Re: [PATCH v2] status: fix branch shown when not only bisecting
From: Junio C Hamano @ 2023-10-16 22:08 UTC (permalink / raw)
To: Rubén Justo; +Cc: Git List
In-Reply-To: <2e24ca9b-9c5f-f4df-b9f8-6574a714dfb2@gmail.com>
Rubén Justo <rjusto@gmail.com> writes:
> In 83c750acde (wt-status.*: better advice for git status added,
> 2012-06-05), git-status received new informative messages to describe
> the ongoing work in a worktree.
>
> These messages were enhanced in 0722c805d6 (status: show the branch name
> if possible in in-progress info, 2013-02-03), to show, if possible, the
> branch where the operation was initiated.
>
> Since then, we show incorrect information when several operations are in
> progress and one of them is bisect:
>
> $ git checkout -b foo
> $ GIT_SEQUENCE_EDITOR='echo break >' git rebase -i HEAD~
> $ git checkout -b bar
> $ git bisect start
> $ git status
> ...
>
> You are currently editing a commit while rebasing branch 'bar' on '...'.
>
> You are currently bisecting, started from branch 'bar'.
>
> ...
>
> Note that we erroneously say "while rebasing branch 'bar'" when we
> should be referring to "foo".
>
> This must have gone unnoticed for so long because it must be unusual to
> start a bisection while another operation is in progress. And even less
> usual to involve different branches.
>
> It caught my attention reviewing a leak introduced in 8b87cfd000
> (wt-status: move strbuf into read_and_strip_branch(), 2013-03-16).
>
> A simple change to deal with this situation can be to record in struct
> wt_status_state, the branch where the bisect starts separately from the
> branch related to other operations.
>
> Let's do it and so we'll be able to display correct information and
> we'll avoid the leak as well.
>
> Signed-off-by: Rubén Justo <rjusto@gmail.com>
>
> ---
>
> Let's try again.
Sigh... nobody seems to be interested in making sure this correctly
improves the system X-<.
After a quick re-read, I didn't find anything glaringly wrong, so
let's see if anybody complains after the patch gets merged to
'next'.
Thanks.
^ permalink raw reply
* Re: [PATCH v3 1/1] bugreport: include +i in outfile suffix as needed
From: Junio C Hamano @ 2023-10-16 22:55 UTC (permalink / raw)
To: Jacob Stopak; +Cc: git
In-Reply-To: <20231016214045.146862-2-jacob@initialcommit.io>
Jacob Stopak <jacob@initialcommit.io> writes:
> builtin/bugreport.c | 83 +++++++++++++++++++++++++++++++--------------
> 1 file changed, 57 insertions(+), 26 deletions(-)
Looking good. It is not easy to do an automated and reliable test
for this one for obvious reasons ;-), so let's queue it as-is.
Thanks.
> - /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
> - report = xopen(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
> + again:
> + /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
> + report = open(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
> + if (report < 0 && errno == EEXIST && !option_suffix_is_from_user) {
> + build_path(&report_path, prefixed_filename,
> + "git-bugreport-", option_suffix, now, &i,
> + ".txt");
> + goto again;
> + } else if (report < 0) {
> + die_errno(_("unable to open '%s'"), report_path.buf);
> + }
I didn't expect a rewrite to add an extra level of indentation like
this, though ;-).
^ permalink raw reply
* Re: [PATCH v2 1/2] t: add a test helper to truncate files
From: Jeff King @ 2023-10-16 23:53 UTC (permalink / raw)
To: brian m. carlson; +Cc: git, Junio C Hamano, Taylor Blau, Jason Hatton
In-Reply-To: <20231012160930.330618-2-sandals@crustytoothpaste.net>
On Thu, Oct 12, 2023 at 04:09:29PM +0000, brian m. carlson wrote:
> +int cmd__truncate(int argc, const char **argv)
> +{
> + char *p = NULL;
> + uintmax_t sz = 0;
> + int fd = -1;
> +
> + if (argc != 3)
> + die("expected filename and size");
> +
> + sz = strtoumax(argv[2], &p, 0);
> + if (*p)
> + die("invalid size");
> +
> + fd = open(argv[1], O_WRONLY | O_CREAT, 0600);
> + if (fd < 0)
> + die_errno("failed to open file %s", argv[1]);
> +
> + if (ftruncate(fd, (off_t) sz) < 0)
> + die_errno("failed to truncate file");
> + return 0;
> +}
Coverity flagged this as leaking the descriptor "fd" (which is obviously
true). I guess it doesn't matter much in practice, since we're exiting
the process directly afterwards. If it were a memory leak we'd free() it
anyway to shut up the leak detector, but I don't know if we want to be
as careful here (in theory it creates noise that distracts from real
problems, but Coverity is noisy enough that I don't even bother looking
at existing problems).
-Peff
^ permalink raw reply
* Re: [PATCH v2 2/2] Prevent git from rehashing 4GiB files
From: Jeff King @ 2023-10-17 0:00 UTC (permalink / raw)
To: brian m. carlson; +Cc: git, Junio C Hamano, Taylor Blau, Jason Hatton
In-Reply-To: <20231012160930.330618-3-sandals@crustytoothpaste.net>
On Thu, Oct 12, 2023 at 04:09:30PM +0000, brian m. carlson wrote:
> +static unsigned int munge_st_size(off_t st_size) {
> + unsigned int sd_size = st_size;
> +
> + /*
> + * If the file is an exact multiple of 4 GiB, modify the value so it
> + * doesn't get marked as racily clean (zero).
> + */
> + if (!sd_size && st_size)
> + return 0x80000000;
> + else
> + return sd_size;
> +}
Coverity complained that the "true" side of this conditional is
unreachable, since sd_size is assigned from st_size, so the two values
cannot be both true and false. But obviously we are depending here on
the truncation of off_t to "unsigned int". I'm not sure if Coverity is
just dumb, or if it somehow has a different size for off_t.
I don't _think_ this would ever cause confusion in a real compiler, as
assignment from a larger type to a smaller has well-defined truncation,
as far as I know.
But I do wonder if an explicit "& 0xFFFFFFFF" would make it more obvious
what is happening (which would also do the right thing if in some
hypothetical platform "unsigned int" ended up larger than 32 bits).
-Peff
^ permalink raw reply
* Bug: dir.c traversing the filesystem: unindexed directories do not get recursed into when there is a (non-excluding) pathspec
From: Joanna Wang @ 2023-10-17 0:08 UTC (permalink / raw)
To: git
What did you do before the bug happened? (Steps to reproduce your issue)
I created a new untracked directory `foo/`.
In .gitattributes i added `bar_yes* text`
and created files `foo/bar_yes` and `foo/bar_no` with content 'chicken'
I tried the following commands:
git grep --no-index chicken -- ':(attr:text)foo/*'
git ls-files --others --exclude-standard ':(attr:text)foo/*'
git status ':(attr:text)foo/*'
What did you expect to happen? (Expected behavior)
For all of the above, I expected `foo/bar_yes` to be returned.
What happened instead? (Actual behavior)
For all three commands, nothing was returned.
Anything else you want to add:
If I change the pathspec to ':(exclude,attr:text)foo/*' it works as expected.
`foo/bar_no` is returned.
(except for git status where the output is just `foo/`, but I think
this is expected.)
This behavior seems due to this section in dir.c:
https://github.com/git/git/blob/master/dir.c#L1893-L1908
for a pathspec like ':(attr:text)foo/*', this section of code will prevent
`foo/` from getting recursed.
It only checks for a pathspec match against the directory name, so any
files under the dir aren't checked.
Commenting that part of the code out, seems to fix this issue.
(for `git status`, `foo/` is returned both with `exclude` in the
pathspec and without)
But I am not sure if simply removing that code is the right fix.
[Enabled Hooks]
commit-msg
pre-commit
prepare-commit-msg
^ permalink raw reply
* Re: [PATCH v3 1/1] bugreport: include +i in outfile suffix as needed
From: Jacob Stopak @ 2023-10-17 3:17 UTC (permalink / raw)
To: Junio C Hamano; +Cc: git
In-Reply-To: <xmqq4jiqkwi1.fsf@gitster.g>
On Mon, Oct 16, 2023 at 03:55:50PM -0700, Junio C Hamano wrote:
> Jacob Stopak <jacob@initialcommit.io> writes:
>
> > builtin/bugreport.c | 83 +++++++++++++++++++++++++++++++--------------
> > 1 file changed, 57 insertions(+), 26 deletions(-)
>
> Looking good. It is not easy to do an automated and reliable test
> for this one for obvious reasons ;-), so let's queue it as-is.
>
> Thanks.
>
Awesome! Thank you!
> > - /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
> > - report = xopen(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
> > + again:
> > + /* fopen doesn't offer us an O_EXCL alternative, except with glibc. */
> > + report = open(report_path.buf, O_CREAT | O_EXCL | O_WRONLY, 0666);
> > + if (report < 0 && errno == EEXIST && !option_suffix_is_from_user) {
> > + build_path(&report_path, prefixed_filename,
> > + "git-bugreport-", option_suffix, now, &i,
> > + ".txt");
> > + goto again;
> > + } else if (report < 0) {
> > + die_errno(_("unable to open '%s'"), report_path.buf);
> > + }
>
> I didn't expect a rewrite to add an extra level of indentation like
> this, though ;-).
Whoops... I looked in another file to check the indentation around a
goto label and misread how it should be. Let me know if I should submit
v4 with that corrected.
^ permalink raw reply
* [PATCH] Include gettext.h in MyFirstContribution tutorial
From: Jacob Stopak @ 2023-10-17 4:15 UTC (permalink / raw)
To: git; +Cc: Jacob Stopak
The tutorial in Documentation/MyFirstContribution.txt has steps to print
some text using the "_" function. However, this leads to compiler errors
when running "make" since "gettext.h" is not #included.
Update docs with a note to #include "gettext.h" in "builtin/psuh.c".
Signed-off-by: Jacob Stopak <jacob@initialcommit.io>
---
Documentation/MyFirstContribution.txt | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/Documentation/MyFirstContribution.txt b/Documentation/MyFirstContribution.txt
index 62d11a5cd7..7cfed60c2e 100644
--- a/Documentation/MyFirstContribution.txt
+++ b/Documentation/MyFirstContribution.txt
@@ -160,10 +160,11 @@ in order to keep the declarations alphabetically sorted:
int cmd_psuh(int argc, const char **argv, const char *prefix);
----
-Be sure to `#include "builtin.h"` in your `psuh.c`.
+Be sure to `#include "builtin.h"` in your `psuh.c`. You'll also need to
+`#include "gettext.h"` to use functions related to printing output text.
-Go ahead and add some throwaway printf to that function. This is a decent
-starting point as we can now add build rules and register the command.
+Go ahead and add some throwaway printf to the `cmd_psuh` function. This is a
+decent starting point as we can now add build rules and register the command.
NOTE: Your throwaway text, as well as much of the text you will be adding over
the course of this tutorial, is user-facing. That means it needs to be
--
2.42.0.398.ga9ecda2788.dirty
^ permalink raw reply related
* Proxy Flag for git-clone, push, pull
From: Jaydeep Das @ 2023-10-17 5:15 UTC (permalink / raw)
To: Git Mailing List
To make git work through proxy, we need to set the `http.proxy` config
or need to set `http_proxy` environment variable.
However wouldn't it be better if there was a flag in the command
itself (like npm) which
overrides whatever proxy is set. Something like
git clone --proxy "http://..." <url>
^ permalink raw reply
* Re: Proxy Flag for git-clone, push, pull
From: Eric Sunshine @ 2023-10-17 5:19 UTC (permalink / raw)
To: Jaydeep Das; +Cc: Git Mailing List
In-Reply-To: <CACaPSotRhyFZ-eBZ9KNKRUjLFHKo09P-Un+sitDXEktzmwuaxA@mail.gmail.com>
On Tue, Oct 17, 2023 at 1:15 AM Jaydeep Das <jaydeepjd.8914@gmail.com> wrote:
> To make git work through proxy, we need to set the `http.proxy` config
> or need to set `http_proxy` environment variable.
>
> However wouldn't it be better if there was a flag in the command
> itself (like npm) which
> overrides whatever proxy is set. Something like
>
> git clone --proxy "http://..." <url>
The -c option allows you to specify configuration on the command-line, so:
git -c http.proxy="http://..." clone <url>
should do what you want.
^ permalink raw reply
* Re: [PATCH] commit: detect commits that exist in commit-graph but not in the ODB
From: Patrick Steinhardt @ 2023-10-17 6:37 UTC (permalink / raw)
To: Junio C Hamano; +Cc: git, Karthik Nayak, Taylor Blau
In-Reply-To: <xmqq1qdy1iyr.fsf@gitster.g>
[-- Attachment #1: Type: text/plain, Size: 5942 bytes --]
On Fri, Oct 13, 2023 at 11:21:48AM -0700, Junio C Hamano wrote:
> Patrick Steinhardt <ps@pks.im> writes:
>
> > @@ -572,8 +572,13 @@ int repo_parse_commit_internal(struct repository *r,
> > return -1;
> > if (item->object.parsed)
> > return 0;
> > - if (use_commit_graph && parse_commit_in_graph(r, item))
> > + if (use_commit_graph && parse_commit_in_graph(r, item)) {
> > + if (!has_object(r, &item->object.oid, 0))
> > + return quiet_on_missing ? -1 :
> > + error(_("commit %s exists in commit-graph but not in the object database"),
> > + oid_to_hex(&item->object.oid));
> > return 0;
> > + }
>
> Ever since this codepath was introduced by 177722b3 (commit:
> integrate commit graph with commit parsing, 2018-04-10), we blindly
> trusted what commit-graph file says. This change is a strict
> improvement in the correctness department, but there are two things
> that are a bit worrying.
>
> One. The additional check should certainly be cheaper than a full
> reading and parsing of an object, either from a loose object file or
> from a pack entry. It may not hurt performance too much, but it
> still would give us more confidence if we know by how much we are
> pessimising good cases where the commit-graph does match reality.
> Our stance on these secondary files that store precomputed values
> for optimization purposes is in general to use them blindly unless
> in exceptional cases where the operation values the correctness even
> when the validity of these secondary files is dubious (e.g., "fsck"),
> and doing this extra check regardless of the caller at this low level
> of the callchain is a bit worrying.
Fair point indeed. The following is a worst-case scenario benchmark of
of the change where we do a full topological walk of all reachable
commits in the graph, executed in linux.git. We parse commit parents via
`repo_parse_commit_gently()`, so the new code path now basically has to
check for object existence of every reachable commit:
Benchmark 1: git -c core.commitGraph=true rev-list --topo-order --all (git = master)
Time (mean ± σ): 2.913 s ± 0.018 s [User: 2.363 s, System: 0.548 s]
Range (min … max): 2.894 s … 2.950 s 10 runs
Benchmark 2: git -c core.commitGraph=true rev-list --topo-order --all (git = pks-commit-graph-inconsistency)
Time (mean ± σ): 3.834 s ± 0.052 s [User: 3.276 s, System: 0.556 s]
Range (min … max): 3.780 s … 3.961 s 10 runs
Benchmark 3: git -c core.commitGraph=false rev-list --topo-order --all (git = master)
Time (mean ± σ): 13.841 s ± 0.084 s [User: 13.152 s, System: 0.687 s]
Range (min … max): 13.714 s … 13.995 s 10 runs
Benchmark 4: git -c core.commitGraph=false rev-list --topo-order --all (git = pks-commit-graph-inconsistency)
Time (mean ± σ): 13.762 s ± 0.116 s [User: 13.094 s, System: 0.667 s]
Range (min … max): 13.645 s … 14.038 s 10 runs
Summary
git -c core.commitGraph=true rev-list --topo-order --all (git = master) ran
1.32 ± 0.02 times faster than git -c core.commitGraph=true rev-list --topo-order --all (git = pks-commit-graph-inconsistency)
4.72 ± 0.05 times faster than git -c core.commitGraph=false rev-list --topo-order --all (git = pks-commit-graph-inconsistency)
4.75 ± 0.04 times faster than git -c core.commitGraph=false rev-list --topo-order --all (git = master)
The added check does lead to a performance regression indeed, which is
not all that unexpected. That being said, the commit-graph still results
in a significant speedup compared to the case where we don't have it.
> Another is that by the time parse_commit_in_graph() returns true and
> we realize that the answer we got is bogus by asking has_object(),
> item->object.parsed has already been toggled on, so the caller now
> has a commit object that claimed it was already parsed and does not
> match reality. Hopefully the caller takes an early exit upon seeing
> a failure from parse_commit_gently() and the .parsed bit does not
> matter, but maybe I am missing a case where it does. I dunno.
We could also call `unparse_commit()` when we notice the stale commit
graph item. This would be in the same spirit as the rest of this patch
as it would lead to an overall safer end state.
In any case I'll wait for additional input before sending a v2, most
importantly to see whether we think that consistency trumps performance
in this case. Personally I'm still of the mind that it should, which
also comes from the fact that we were fighting with stale commit graphs
several times in production data.
Patrick
> Other than that, sounds very sensible and the code change is clean.
>
> Will queue. Thanks.
>
> > diff --git a/t/t5318-commit-graph.sh b/t/t5318-commit-graph.sh
> > index ba65f17dd9..25f8e9e2d3 100755
> > --- a/t/t5318-commit-graph.sh
> > +++ b/t/t5318-commit-graph.sh
> > @@ -821,4 +821,27 @@ test_expect_success 'overflow during generation version upgrade' '
> > )
> > '
> >
> > +test_expect_success 'commit exists in commit-graph but not in object database' '
> > + test_when_finished "rm -rf repo" &&
> > + git init repo &&
> > + (
> > + cd repo &&
> > +
> > + test_commit A &&
> > + test_commit B &&
> > + test_commit C &&
> > + git commit-graph write --reachable &&
> > +
> > + # Corrupt the repository by deleting the intermittent commit
> > + # object. Commands should notice that this object is absent and
> > + # thus that the repository is corrupt even if the commit graph
> > + # exists.
> > + oid=$(git rev-parse B) &&
> > + rm .git/objects/"$(test_oid_to_path "$oid")" &&
> > +
> > + test_must_fail git rev-parse HEAD~2 2>error &&
> > + grep "error: commit $oid exists in commit-graph but not in the object database" error
> > + )
> > +'
> > +
> > test_done
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH v3 03/17] commit-graph: ensure Bloom filters are read with consistent settings
From: Patrick Steinhardt @ 2023-10-17 8:45 UTC (permalink / raw)
To: Taylor Blau
Cc: git, Jonathan Tan, Junio C Hamano, Jeff King, SZEDER Gábor
In-Reply-To: <2ecc0a2d58432b149d73a3e2abfa948eb1f0aa0b.1696969994.git.me@ttaylorr.com>
[-- Attachment #1: Type: text/plain, Size: 8299 bytes --]
On Tue, Oct 10, 2023 at 04:33:26PM -0400, Taylor Blau wrote:
> The changed-path Bloom filter mechanism is parameterized by a couple of
> variables, notably the number of bits per hash (typically "m" in Bloom
> filter literature) and the number of hashes themselves (typically "k").
>
> It is critically important that filters are read with the Bloom filter
> settings that they were written with. Failing to do so would mean that
> each query is liable to compute different fingerprints, meaning that the
> filter itself could return a false negative. This goes against a basic
> assumption of using Bloom filters (that they may return false positives,
> but never false negatives) and can lead to incorrect results.
>
> We have some existing logic to carry forward existing Bloom filter
> settings from one layer to the next. In `write_commit_graph()`, we have
> something like:
>
> if (!(flags & COMMIT_GRAPH_NO_WRITE_BLOOM_FILTERS)) {
> struct commit_graph *g = ctx->r->objects->commit_graph;
>
> /* We have changed-paths already. Keep them in the next graph */
> if (g && g->chunk_bloom_data) {
> ctx->changed_paths = 1;
> ctx->bloom_settings = g->bloom_filter_settings;
> }
> }
>
> , which drags forward Bloom filter settings across adjacent layers.
>
> This doesn't quite address all cases, however, since it is possible for
> intermediate layers to contain no Bloom filters at all. For example,
> suppose we have two layers in a commit-graph chain, say, {G1, G2}. If G1
> contains Bloom filters, but G2 doesn't, a new G3 (whose base graph is
> G2) may be written with arbitrary Bloom filter settings, because we only
> check the immediately adjacent layer's settings for compatibility.
>
> This behavior has existed since the introduction of changed-path Bloom
> filters. But in practice, this is not such a big deal, since the only
> way up until this point to modify the Bloom filter settings at write
> time is with the undocumented environment variables:
>
> - GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY
> - GIT_TEST_BLOOM_SETTINGS_NUM_HASHES
> - GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS
>
> (it is still possible to tweak MAX_CHANGED_PATHS between layers, but
> this does not affect reads, so is allowed to differ across multiple
> graph layers).
>
> But in future commits, we will introduce another parameter to change the
> hash algorithm used to compute Bloom fingerprints itself. This will be
> exposed via a configuration setting, making this foot-gun easier to use.
>
> To prevent this potential issue, validate that all layers of a split
> commit-graph have compatible settings with the newest layer which
> contains Bloom filters.
>
> Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
> Original-test-by: SZEDER Gábor <szeder.dev@gmail.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
> commit-graph.c | 25 +++++++++++++++++
> t/t4216-log-bloom.sh | 64 ++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 89 insertions(+)
>
> diff --git a/commit-graph.c b/commit-graph.c
> index 1a56efcf69..ae0902f7f4 100644
> --- a/commit-graph.c
> +++ b/commit-graph.c
> @@ -498,6 +498,30 @@ static int validate_mixed_generation_chain(struct commit_graph *g)
> return 0;
> }
>
> +static void validate_mixed_bloom_settings(struct commit_graph *g)
> +{
> + struct bloom_filter_settings *settings = NULL;
> + for (; g; g = g->base_graph) {
> + if (!g->bloom_filter_settings)
> + continue;
> + if (!settings) {
> + settings = g->bloom_filter_settings;
> + continue;
> + }
> +
> + if (g->bloom_filter_settings->bits_per_entry != settings->bits_per_entry ||
> + g->bloom_filter_settings->num_hashes != settings->num_hashes) {
> + g->chunk_bloom_indexes = NULL;
> + g->chunk_bloom_data = NULL;
> + FREE_AND_NULL(g->bloom_filter_settings);
> +
> + warning(_("disabling Bloom filters for commit-graph "
> + "layer '%s' due to incompatible settings"),
> + oid_to_hex(&g->oid));
> + }
> + }
> +}
> +
> static int add_graph_to_chain(struct commit_graph *g,
> struct commit_graph *chain,
> struct object_id *oids,
> @@ -614,6 +638,7 @@ struct commit_graph *load_commit_graph_chain_fd_st(struct repository *r,
> }
>
> validate_mixed_generation_chain(graph_chain);
> + validate_mixed_bloom_settings(graph_chain);
>
> free(oids);
> fclose(fp);
> diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
> index 322640feeb..f49a8f2fbf 100755
> --- a/t/t4216-log-bloom.sh
> +++ b/t/t4216-log-bloom.sh
> @@ -420,4 +420,68 @@ test_expect_success 'Bloom generation backfills empty commits' '
> )
> '
>
> +graph=.git/objects/info/commit-graph
> +graphdir=.git/objects/info/commit-graphs
> +chain=$graphdir/commit-graph-chain
> +
> +test_expect_success 'setup for mixed Bloom setting tests' '
> + repo=mixed-bloom-settings &&
> +
> + git init $repo &&
> + for i in one two three
> + do
> + test_commit -C $repo $i file || return 1
> + done
> +'
> +
> +test_expect_success 'split' '
> + # Compute Bloom filters with "unusual" settings.
> + git -C $repo rev-parse one >in &&
> + GIT_TEST_BLOOM_SETTINGS_NUM_HASHES=3 git -C $repo commit-graph write \
> + --stdin-commits --changed-paths --split <in &&
> + layer=$(head -n 1 $repo/$chain) &&
> +
> + # A commit-graph layer without Bloom filters "hides" the layers
> + # below ...
> + git -C $repo rev-parse two >in &&
> + git -C $repo commit-graph write --stdin-commits --no-changed-paths \
> + --split=no-merge <in &&
> +
> + # Another commit-graph layer that has Bloom filters, but with
> + # standard settings, and is thus incompatible with the base
> + # layer written above.
> + git -C $repo rev-parse HEAD >in &&
> + git -C $repo commit-graph write --stdin-commits --changed-paths \
> + --split=no-merge <in &&
> +
> + test_line_count = 3 $repo/$chain &&
> +
> + # Ensure that incompatible Bloom filters are ignored.
> + git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
> + >expect 2>err &&
> + git -C $repo log --oneline --no-decorate -- file >actual 2>err &&
> + test_cmp expect actual &&
> + grep "disabling Bloom filters for commit-graph layer .$layer." err
> +'
Up to this point everything looks sensible to me.
> +test_expect_success 'merge graph layers with incompatible Bloom settings' '
> + # Ensure that incompatible Bloom filters are ignored when
> + # generating new layers.
> + git -C $repo commit-graph write --reachable --changed-paths 2>err &&
> + grep "disabling Bloom filters for commit-graph layer .$layer." err &&
> +
> + test_path_is_file $repo/$graph &&
> + test_dir_is_empty $repo/$graphdir &&
> +
> + # ...and merging existing ones.
> + git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
> + >expect 2>err &&
> + GIT_TRACE2_PERF="$(pwd)/trace.perf" \
> + git -C $repo log --oneline --no-decorate -- file >actual 2>err &&
But this test is a bit confusing to me, to be honest, also because the
comment for the second block here reads funny. We don't really merge
anything, do we? We only generate logs and compare that the log with and
without the resulting merged commit graph is the same. The actual logic
happened before.
> + test_cmp expect actual && cat err &&
The `cat err` looks like a leftover from debugging.
> + grep "statistics:{\"filter_not_present\":0" trace.perf &&
Also, why should the filter not be present here? If we merge the
commit-graphs with `--changed-paths` I'd have expected that we either
carry over bloom filters from preexisting commit graphs if compatible,
or otherwise generate them if they are either incompatible or don't
exist.
I feel like I'm missing something obvious, so this may be me just
missing the bigger picture.
> + ! grep "disabling Bloom filters" err
Can we make this assertion stricter and verify that `err` is empty? I
always think that `! grep` is quite a fragile pattern as it is quite
prone to becoming stale, e.g. when the error message itself would
change.
Patrick
> +'
> +
> test_done
> --
> 2.42.0.342.g8bb3a896ee
>
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox