Re: [PATCH] t3200: replace hardcoded null OID with $ZERO_OID

From: Patrick Steinhardt <ps@pks.im>
Date: Mar 11, 2026, 11:20 UTC
Message-ID: <abFP9vZOVHGI536G@pks.im>
In-Reply-To: <20260311105829.60508-1-r.siddharth.shrimali@gmail.com>

On Wed, Mar 11, 2026 at 04:28:10PM +0530, Siddharth Shrimali wrote:

Show 28 quoted lines

> Taking into consideration the SHA-256 transition, the test suite must
> be updated to support the length of the underlying hash algorithm.
> Tests that rely on hardcoded 40-character strings to represent the
> null object ID will fail when run in a SHA-256 environment, which
> expects a 64-character hash.
> 
> Replace the hardcoded 40-zero string in the 'git branch --merged' test
> with the '$ZERO_OID' variable which is provided by the test framework.
> This ensures the test dynamically adapts to the correct null OID
> length and functions correctly regardless of the active hash
> algorithm.
> 
> Signed-off-by: Siddharth Shrimali <r.siddharth.shrimali@gmail.com>
> ---
>  t/t3200-branch.sh | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/t/t3200-branch.sh b/t/t3200-branch.sh
> index c58e505c43..ed317a75f5 100755
> --- a/t/t3200-branch.sh
> +++ b/t/t3200-branch.sh
> @@ -1494,7 +1494,7 @@ test_expect_success 'refuse --edit-description on unborn branch for now' '
>  '
>  
>  test_expect_success '--merged catches invalid object names' '
> -	test_must_fail git branch --merged 0000000000000000000000000000000000000000
> +	test_must_fail git branch --merged $ZERO_OID
>  '

I expect that the failure reason before and after this change is different, right? And likewise, I expect that before the change, the failure with SHA1 is likely different than the one with SHA256.

Taking a peek, that's indeed the case. With SHA1 we get:

    error: option `merged' must point to a commit

But with SHA256 we get:

    fatal: malformed object name 0000000000000000000000000000000000000000

So the only reason why we didn't detect that the test is broken with SHA256 is that we didn't verify the error message. Do we maybe want to make the test a bit less fragile by using something like `test_grep "must point to a commit"` on the error message?

Patrick