[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Test failures with musl libc
From: |
arnold |
Subject: |
Re: Test failures with musl libc |
Date: |
Sun, 07 Feb 2021 00:36:13 -0700 |
User-agent: |
Heirloom mailx 12.5 7/5/10 |
Hi.
Thanks for the report.
Sören Tempel <soeren@soeren-tempel.net> wrote:
> Hello,
>
> I am currently trying to get the gawk test suite to pass as part of the
> Alpine Linux gawk package. Alpine uses musl libc instead of glibc. I
> only observed two failing tests with gawk-5.1.0 which I would preferably
> also get to pass on Alpine:
>
> On x86_64 the clos1way6 test case fails with the following error:
>
> *** clos1way6.ok 2019-08-28 18:54:15.000000000 +0000
> --- _clos1way6 2021-02-06 11:52:29.933319277 +0000
> ***************
> *** 1,3 ****
> gawk: clos1way6.awk:5: warning: fflush: cannot flush: two-way pipe
> `cat - 1>&2; sleep 2' has closed write end
> - test1
> Bad file descriptor
> --- 1,3 ----
> gawk: clos1way6.awk:5: warning: fflush: cannot flush: two-way pipe
> `cat - 1>&2; sleep 2' has closed write end
> Bad file descriptor
> + test1
>
> The test1 string seems to be output after the Bad file descriptor
> message on musl. Maybe a difference in libc flushing behaviour, any idea
> how this could be fixed?
I don't think it can be. This happens on Linux when running under
valgrind, and also I think on a few other non-GLIBC systems. I have tried
in the past to make it work across more systems, but not had any luck.
> Additionally, the following test case fails on the s390x architecture:
>
> *** sigpipe1.ok Wed Aug 28 18:43:58 2019
> --- _sigpipe1 Sat Feb 6 11:42:57 2021
> ***************
> *** 1,3 ****
> system
> pipe to command
> ! pipe from command
> --- 1,4 ----
> system
> pipe to command
> ! gawk: sigpipe1.awk:7: fatal: print to "yes | true" failed: Broken pipe
> ! EXIT CODE: 2
>
> Unfortunately, I don't have access to this architecture myself and can't
> properly debug this problem, so this is more of an FYI.
No clue why this would happen. I would also need access to such
a system, with compiler and GDB, in order to figure out what's going on.
Sorry I'm not of more help,
Arnold