perl 5.8.8 syslog tests failing on MacOS X (sometimes)

[kane@dwim.org]

On Fri, Dec 15, 2006 at 12:57:43PM +0000, Edmund von der Burg wrote:
> This is odd - and I think I know the cause but I want to check if
> anyone else has seen it:
> 
> evdb at cindy ~/cdb/ext/build/perl-5.8.8/t $ ./perl harness
> ../ext/Sys/Syslog/t/syslog.t
> ../ext/Sys/Syslog/t/syslog....ok
>        50/119 skipped: various reasons
> All tests successful, 50 subtests skipped.
> Files=1, Tests=119,  1 wallclock secs ( 0.07 cusr +  0.03 csys =  0.10 CPU)
> 
> And then again a few seconds later:
> 
> evdb at cindy ~/cdb/ext/build/perl-5.8.8/t $ ./perl harness
> ../ext/Sys/Syslog/t/syslog.t
> ../ext/Sys/Syslog/t/syslog....ok 1/119
> ../ext/Sys/Syslog/t/syslog....NOK 83#   Failed test 'syslog() called
> with level 'info''
> #   in ../ext/Sys/Syslog/t/syslog.t at line 108.
> #          got: 'no connection to syslog available
> #       - udp connect: nobody listening at
> ../ext/Sys/Syslog/t/syslog.t line 107# '
> #     expected: ''

[...]

> the test works / fails in roughly the same proportion.
> 
> My guess is that the tests should always be skipped, but that
> occasionally the error being returned is not recognized as valid: ie
> the 'udp connect' bit is unexpected.
> 
> Anyone confirm / deny this?
> 
> If that is it I'll submit a patch and become immortal :)

Yup, it's happened to me too. Sebastien is aware of the issue, but
unable to reproduce it. It seems sporadic failures, where one can't
even predict if it will always be the 'first time after ./Configure'
or 'after the last successful run' that it fails.

I'm sure a patch would be appreciated, but always skipping the test
seems the wrong way to go about this.

-- Jos