RE: [BUG] t5300 fails tests on NonStop x86 at 2.32.0-rc2

From: Randall S. Becker <rsbecker@nexbridge.com>
Date: Jun 1, 2021, 15:09 UTC
Message-ID: <00c901d756f8$275a5ab0$760f1010$@nexbridge.com>

On June 1, 2021 10:52 AM, I wrote:

Show 32 quoted lines

>To: 'git@vger.kernel.org' <git@vger.kernel.org>
>Subject: RE: [BUG] t5300 fails tests on NonStop x86 at 2.32.0-rc2
>
>On May 31, 2021 2:36 PM, I wrote:
>>To: 'git@vger.kernel.org' <git@vger.kernel.org>
>>Subject: [BUG] t5300 fails tests on NonStop x86 at 2.32.0-rc2
>>
>>This one started at 2.32.0-rc1, AFIAK:
>>
>>expecting success of 5300.2 'pack without delta':
>>        packname_1=$(git pack-objects --progress --window=0 test-1 \
>>                        <obj-list 2>stderr) &&
>>        check_deltas stderr = 0
>>
>>+ + git pack-objects --progress --window=0 test-1
>>+ 0< obj-list 2> stderr
>>packname_1=
>>error: last command exited with $?=128
>>not ok 2 - pack without delta
>>#
>>#               packname_1=$(git pack-objects --progress --window=0
>>test-1 \ #                               <obj-list 2>stderr) && #
>>check_deltas stderr = 0 #
>>
>>Stderr contains:
>>
>>Enumerating objects: 8, done.
>>Counting objects: 100% (8/8), done.
>>fatal: fsync error on '.git/objects/pack/tmp_pack_Y6t3Br': Interrupted
>>system call
>
>I thought this was limited to the NonStop x86, but it also apples to NonStop ia64. Any wisdom on where this might be, so I can

debug?

This may be a recurrence of a MAX_SIZE problem from a few years ago with xread, xwrite. The I/Os that are failing appear to be using using IO sizes larger than the platform maximum. This will cause fsync to fail with EINTR. Pointers on where this might be, please?

Thanks, Randall