develooper Front page | perl.perl5.porters | Postings from December 2014

Year-end EBCDIC porting status report

Thread Previous | Thread Next
From:
Karl Williamson
Date:
December 31, 2014 03:46
Subject:
Year-end EBCDIC porting status report
Message ID:
54A37184.3040204@khwilliamson.com
On 12/30/2014 12:10 AM, Yaroslav Kuzmin wrote:
> P.S. I am on vacation from 01.01.2015 to 12.01.2015

Have a good vacation

This would be a good time to update the current status.  There are 
currently 10 failing core tests, 99% pass rate.  These last few are 
proving somewhat intractable.  I'll go over each:

1) t/run/locale.t:  This appears to be some weird interaction with the 
shell.  The data indicates that PerlEnv_getenv("LC_ALL") is getting 
"perlio" when the test environment is testing with perlio, and "stdio" 
when the test environment is stdio.  This is instead of what these 
environment variables are actually getting set to (the string 'invalid') 
  and Perl code inside the run that does

   print STDERR "ENV{LC_ALL}=", $ENV{LC_ALL}

prints
   ENV{LC_ALL}=invalid

So it appears that retrieving the environment is corrupted.  That's why 
I would like this run manually from a terminal.  My guess is that it 
will work.

2) t/re/pat_advanced.t  This is now failing one test.  I've broken it 
recently with a "fix" to an unrelated bug.  And I just haven't had a 
chance to look at it, but it should be easy to fix.

3) t/op/pack.t.  This should again be easy to fix when I get a chance to 
look in real depth.

4) t/op/stat.t I've discussed this in other emails.  The OS is returning 
EBADF for a file descriptor that looks right.  I'm hoping that after the 
holidays someone can look at it who has expertise.

5) t/porting/readme.t  This is waiting for me to get around to looking 
at fixing a cpan module it depends on.

6) dist/Data-Dumper/t/dumper.t This is a parse bug.  I now have narrowed 
it down to a single line change in the .t that causes the problem, and 
now have to pore over the debug log.

7) dist/ExtUtils-CBuilder/t/04-base.t  This looks like an lstat() 
problem.  lstat is frequently failing with errno=146=EDC5146I Too many 
levels of symbolic links.  This happens frequently though this is the 
only time it is accompanied by a test failure.  I'm waiting for an 
expert on os390 to look at this.

8) dist/Net-Ping/t/450_service.t  This again looks like an OS 
interaction.  Nothing on this has changed from an earlier status report

9) ext/POSIX/t/sigaction.t  We have a suggestion to use a different 
library to see what happens.  Yaroslav Have you tried that?

10) lib/open.t  I think this will be fixed after a cpan module it uses 
gets fixed.

When tests for the cpan modules shipped with core are added, we drop to 
a 91% pass rate.

I'm hoping Yaroslave can manually run t/run/locale.t on the 
already-existing bracn before he leaves.

There is a new branch as well, but it has re-enabled some code that I 
suspect will cause it not to work, so that smoking this likely will be 
very fast as it likely won't fully compile.  I don't understand what the 
problem has been, and so it might be possible that something done in the 
mean time will have fixed whatever is causing the breakage, but I don't 
really have any candidate fixes that might have done that.  But we need 
to try it out at some point, and doing so just before going on vacation 
seems like the best time.

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About