bug#51286: [core-updates-frozen] Gawk 5.1.0 test failure (non-deterministic)

[Maxim Cournoyer]

Hello,

Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:

> Hello,
>
> While rebuilding the core-updates-core-frozen branch locally, I stumbled
> upon:
>
> 1 TESTS FAILED
> make[3]: *** [Makefile:5145: pass-fail] Error 1
> make[3]: Leaving directory '/tmp/guix-build-gawk-5.1.0.drv-0/gawk-5.1.0/test'
> make[3]: Entering directory '/tmp/guix-build-gawk-5.1.0.drv-0/gawk-5.1.0/test'
> for i in _* ; \
> do  \
>         if [ "$i" != "_*" ]; then \
>         echo ============== $i ============= ; \
>         base=`echo $i | sed 's/^_//'` ; \
>         if [ -r ${base}.ok ]; then \
>         diff -c ${base}.ok $i ; \
>         else \
>         diff -c "."/${base}.ok  $i ; \
>         fi ; \
>         fi ; \
> done 
> ============== _clos1way2 =============
> *** clos1way2.ok        2019-08-28 18:43:58.000000000 +0000
> --- _clos1way2  2021-10-19 04:25:01.526475198 +0000
> ***************
> *** 1,4 ****
> ! gawk: clos1way2.awk:5: (FILENAME=- FNR=1) warning: fflush: cannot flush: 
> two-way pipe `cat - 1>&2 && sleep 2' has closed write end
> ! test
>   gawk: clos1way2.awk:6: (FILENAME=- FNR=1) fatal: print: attempt to write to 
> closed write end of two-way pipe
>   EXIT CODE: 2
> --- 1,4 ----
> ! gawk: clos1way2.awk:5: test
> ! (FILENAME=- FNR=1) warning: fflush: cannot flush: two-way pipe `cat - 1>&2 
> && sleep 2' has closed write end
>   gawk: clos1way2.awk:6: (FILENAME=- FNR=1) fatal: print: attempt to write to 
> closed write end of two-way pipe
>   EXIT CODE: 2
> make[3]: *** [Makefile:5154: diffout] Error 1
> make[3]: Leaving directory '/tmp/guix-build-gawk-5.1.0.drv-0/gawk-5.1.0/test'
> make[2]: *** [Makefile:2034: check] Error 1
> make[2]: Leaving directory '/tmp/guix-build-gawk-5.1.0.drv-0/gawk-5.1.0/test'
> make[1]: *** [Makefile:747: check-recursive] Error 1
> make[1]: Leaving directory '/tmp/guix-build-gawk-5.1.0.drv-0/gawk-5.1.0'
> make: *** [Makefile:1047: check] Error 2
>
> Test suite failed, dumping logs.
>
> To be investigated, but we should probably:
>
> 1. report the problem upstream;
> 2. disable the test.
>
> Thank you,
>
> Maxim

Seems it had to do with using a minimal variant of Bash instead of the
regular full-fledged Bash.  I've fixed it on the
core-updates-frozen-batched-changes branch.

Closing.

Maxim