Bug#825018: perl: FTBFS on hurd-i386: hardcoded E2BIG errno value

[Dominic Hargreaves]

On Sun, May 22, 2016 at 01:56:53PM +0000, Samuel Thibault wrote:
> The following testsuite failure on hurd-i386:
> 
>     dist/constant/t/constant ...................................... #   Failed test at t/constant.t line 99.
>     #     '1'
>     #         >
>     #     '6'
>     # Looks like you failed 1 test of 109.
>     FAILED at test 31
> 
> is actually a bug in the test:
> 
>     use constant E2BIG => $! = 7;
>     cmp_ok E2BIG, '==', 7;
>     # This is something like "Arg list too long", but the actual message
>     # text may vary, so we can't test much better than this.
>     cmp_ok length(E2BIG), '>', 6;
> 
> This hardcodes that E2BIG is supposed to be 7, but nothing asserts that
> in C and POSIX, and hurd-i386 happens to be using other values, and thus
> the test fails, but that's not a bug in hurd-i386.

Hmm, this test has remained unchanged for a long time, and doesn't fail
for me on exodar:

dom at exodar:~/perl$ ./perl -I. -T -MTestInit dist/constant/t/constant.t |grep ^not
not ok 103 - list constant has constant elements (1) # TODO disabled for now; breaks CPAN; see perl #119045
not ok 104 - list constant has constant elements (2) # TODO disabled for now; breaks CPAN; see perl #119045
dom at exodar:~/perl$ 

Admittedly, the test is flawed; I get:

dom at exodar:~$ cat test.pl 
#!/usr/bin/perl

$! = 7;
print "errno 7 stringifies to '$!'\n";
dom at exodar:~$ ./test.pl 
errno 7 stringifies to '(os/kern) not receiver'

but because that error message is also long enough, the test passes.
Can you shed any more light about how to reproduce this test failure?

Thanks,
Dominic.