This behavior of git-rev-parse is observed since git 1.8.3.1
at least(*), and likely earlier versions.
At least one git-reliant project in-the-wild relies on this
current behavior of git-rev-parse being able to handle multiple
--since= arguments without squeezing identical results together.
So add a test to prevent the potential for regression in
downstream projects.
(*) 1.8.3.1 the version packaged for CentOS 7.x
Signed-off-by: Eric Wong <redacted>
---
Said git-reliant project does use allow /optional/ use of libgit2
via Perl Inline::C; but libgit2's git__date_parse() isn't part of
the public API; and it's unlikely anybody using CentOS 7.x would
run the latest libgit2 even if it were added today.
t/t1500-rev-parse.sh | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/t/t1500-rev-parse.sh b/t/t1500-rev-parse.sh
index abdda360d5..deae916707 100755
--- a/t/t1500-rev-parse.sh
+++ b/t/t1500-rev-parse.sh
@@ -243,4 +243,19 @@ test_expect_success 'showing the superproject correctly' '
test_cmp expect out
'
+# at least one external project depends on this behavior:
+test_expect_success 'rev-parse --since= unsqueezed ordering' '
+ x1=--since=1970-01-01T00:00:01Z &&
+ x2=--since=1970-01-01T00:00:02Z &&
+ x3=--since=1970-01-01T00:00:03Z &&
+ git rev-parse $x1 $x1 $x3 $x2 >actual &&
+ cat >expect <<-EOF &&
+ --max-age=1
+ --max-age=1
+ --max-age=3
+ --max-age=2
+ EOF
+ test_cmp expect actual
+'
+
test_done
On Wed, Feb 10, 2021 at 09:55:43PM +0000, Eric Wong wrote:
This behavior of git-rev-parse is observed since git 1.8.3.1
at least(*), and likely earlier versions.
At least one git-reliant project in-the-wild relies on this
current behavior of git-rev-parse being able to handle multiple
--since= arguments without squeezing identical results together.
So add a test to prevent the potential for regression in
downstream projects.
I had to read this a few times to understand what "this behavior" meant.
It is just: when given multiple --since options, output a --max-age for
each of them, even though internally, Git's revision traversal will only
use one (in the usual last-one-wins fashion).
I'm not sure if I was just being dense, or if this could be spelled out
more clearly. :)
Out of curiosity, why does the other project want that? From your
mention of libgit2's git__date_parse(), I assume it's something that
wants to parse approxidates into timestamps in a script. Maybe we ought
to provide a more direct and robust way of doing that. We have a similar
need in t0006, but we use a test-helper program for it.
(I have no problem in the meantime with this patch, though; any new
method for accomplishing this would want to give other projects time to
adapt to its use).
-Peff
Jeff King [off-list ref] wrote:
On Wed, Feb 10, 2021 at 09:55:43PM +0000, Eric Wong wrote:
quoted
This behavior of git-rev-parse is observed since git 1.8.3.1
at least(*), and likely earlier versions.
At least one git-reliant project in-the-wild relies on this
current behavior of git-rev-parse being able to handle multiple
--since= arguments without squeezing identical results together.
So add a test to prevent the potential for regression in
downstream projects.
I had to read this a few times to understand what "this behavior" meant.
It is just: when given multiple --since options, output a --max-age for
each of them, even though internally, Git's revision traversal will only
use one (in the usual last-one-wins fashion).
I'm not sure if I was just being dense, or if this could be spelled out
more clearly. :)
*shrug* :> My brain struggles with coherent thought so I'm
surprised anybody is able to understand me at all :x
Out of curiosity, why does the other project want that? From your
mention of libgit2's git__date_parse(), I assume it's something that
wants to parse approxidates into timestamps in a script. Maybe we ought
to provide a more direct and robust way of doing that. We have a similar
need in t0006, but we use a test-helper program for it.
It takes about 5ms for my system to run git-rev-parse once. I
may be getting multiple approxidates at once, 1-2 in a typical
input (start..end); but a strange or malicious input could have
hundreds/thousands of approxidates to parse.
Thus, I'm batching up all the approxidates into one rev-parse
invocation (up to system argv limits right now). With the
output lines split into an array, walking the output/input
arrays in parallel will match them up. Hypothetically, if
rev-parse were to get clever and deduplicate or reject identical
inputs; then the parallel walk would be broken.
(I have no problem in the meantime with this patch, though; any new
method for accomplishing this would want to give other projects time to
adapt to its use).
Yes. I think I've mentioned some years/decade ago having
general functionality along the lines of "git cat-file --batch"
or fast-import would be nice (even for some existing scripts and
tests shipped with git). (v)fork+execve is painful even on
Linux.