Re: [Failure] Re: git 2.40.0-rc1

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 2, 2023, 20:00 UTC
Message-ID: <xmqq356mkjp1.fsf@gitster.g>
In-Reply-To: <001a01d94d3f$04c2e460$0e48ad20$@nexbridge.com>

<rsbecker@nexbridge.com> writes:

> On Thursday, March 2, 2023 1:44 PM, Junio C Hamano wrote:
>><rsbecker@nexbridge.com> writes:
>>
>>> Git 2.40.0-rc2 was a regression from rc1. The following new tests failed:

I think you are reporting -rc1 relative to -rc0.

Show 19 quoted lines

>>> t3206 - virtually all subtests failed. An example of this failure is a
>>> results comparison in subtest 1 (seemingly relating to abbreviation):
>>> + test_cmp expect actual
>>> --- expect      2023-03-02 16:39:23 +0000
>>> +++ actual      2023-03-02 16:39:23 +0000
>>> @@ -1,4 +1,4 @@
>>> -1:  4de457d = 1:  35b9b25 s/5/A/
>>> -2:  fccce22 = 2:  de345ab s/4/A/
>>> -3:  147e64e = 3:  9af6654 s/11/B/
>>> -4:  a63e992 = 4:  2901f77 s/12/B/
>>> +1:  4de457d2c0d218f48d66f45f9b30f3aa62562105 = 1:
>>> 35b9b25f76d404d09a23e6c8efa96c3ce19e19aa s/5/A/
>>
>>Are any big endian machines involved?
>>
>>cf. https://lore.kernel.org/git/Y%2F+paI8WGSmEbv%2Fw@pobox.com/
>
> Oh yes. All NonStop variants are big-endian. That hasn't changed the SHA-1
> signatures though.

There is a stupid sequence that slipped through the review

	char c = structPtr->intMember;

	if (c < 0)
		... do something ...

and where "char" is "unsigned char", the compiler thinks "do something" will never trigger and from there everything goes haywire.

The tip of 'master' already has a fix (and no other changes as of this writing), and will be part of -rc2.

Thanks.