From: Jeff King <peff@peff.net>
Date: Wed, 14 Jan 2026 17:30:55 GMT
Subject: Re: [PATCH v2] remote-curl: Use auth for probe_rpc() requests too
Message-ID: <20260114173055.GD885771@coredump.intra.peff.net>
In-Reply-To: <613b47af-5269-44e3-87f5-d29fba9d73b3@nvidia.com>

On Wed, Jan 14, 2026 at 08:33:43AM -0800, Aaron Plattner wrote:

> Ooh, neat. I guess I copied the wrong example. I verified that this works
> too, so I can send a v5 for that.
> 
> Is it worth changing up the other cases of this pattern, mostly in
> pack-refs-tests.sh? E.g.,
> 
> 	# Create 15 loose references.
> 	printf "create refs/heads/loose-%d HEAD\n" $(test_seq 15) >stdin &&
> 	git update-ref --stdin <stdin &&
> 
> [...]
> 
> 	# Create 99 packed refs. This should cause the heuristic
> 	# to require more than the minimum amount of loose refs.
> 	test_seq 99 |
> 	while read i
> 	do
> 		printf "create refs/heads/packed-%d HEAD\n" $i || return 1
> 	done >stdin &&
> 	git update-ref --stdin <stdin &&
> 
> I can put together a patch for those.

Yeah, I think they are worth updating. I looked for spots to convert
when I added the feature in b32c7ec02f (test-lib: teach test_seq the -f
option, 2025-06-23). But I missed those ones.

I think I grepped for "for i in $(test_seq ...)", but the use of the
while loop and the $()-substitution meant I didn't see them. Examining
every test_seq call didn't seem worthwhile, as there are hundreds. ;)

Grepping for "test_seq .* |" does yield a few more (e.g., p5551), but
most are false positives. Grepping for "printf.*$(test_seq" gets some
more.

I don't know that we need to exhaustively find all cases. ;) In the
first case above, it does save us a subshell. In the second case I think
it saves us a subshell _and_ the result is much easier to read (because
it avoids the loop and return). So IMHO it's a nice improvement, but
there's diminishing returns to investigating every test_seq call.

-Peff