develooper Front page | perl.perl5.porters | Postings from May 2008

Re: helping 5.8.9

Thread Previous | Thread Next
From:
Reini Urban
Date:
May 6, 2008 12:05
Subject:
Re: helping 5.8.9
Message ID:
4820ABEB.4050600@x-ray.at
Nicholas Clark schrieb:
> On Mon, May 05, 2008 at 09:48:41PM +0200, Reini Urban wrote:
>> Nicholas Clark schrieb:
>>> So, does anyone feel like taking 
>>> rsync://ftp.linux.activestate.com/perl-5.8.x/
>> cygwin-thread-multi-64int smoke @33736:
>>
>> ext/DB_File/t/db-btree......................Can't call method 
>> "filter_fetch_key" on an undefined value at ../ext/DB_File/t/db-btree.t 
>> line 1465.
>> FAILED at test 172
>> ext/Time/HiRes/t/HiRes......................FAILED--non-zero wait status: 15
>> lib/DBM_Filter/t/02core.....................#   Failed test 'tied to 
>> SDBM_File'
>> #   at ../lib/DBM_Filter/t/02core.t line 583.
>> FAILED at test 139
>>
>> And then a hang at lib/ExtUtils/t/basic
>> because of loading 
>> /usr/lib/perl5/vendor_perl/5.8/cygwin/auto/version/vxs/vxs.dll
>> (from 5.8.8)
>> Attaching to the hung process revealed nothing obvious.
> 
> Thanks for testing this.
> 
> Erk. If you try to build a release 5.8.8 with the same toolchain, does it
> pass tests? Or fail in the same way?

An updated build is better now.
I removed otherlibdirs and inc_version_list, to get rid of the Cwd.dll 
call to vxs.dll which hangs.

Failed 2 tests out of 1157, 99.83% okay.
u=5.06  s=9.22  cu=634.27  cs=326.52  scripts=1157  tests=134755
	../ext/DB_File/t/db-btree.t
	../ext/Time/HiRes/t/HiRes.t
ext/DB_File/t/db-btree......................Can't call method 
"filter_fetch_key" on an undefined value at ../ext/DB_File/t/db-btree.t 
line 1465.
FAILED at test 172
ext/Time/HiRes/t/HiRes......................FAILED--non-zero wait status: 15

db-btree should be easy to fix:
    $db->filter_fetch_key   (sub { }) ;

   5.8.9 just adds support for DB_HASHVERSION compared to 5.10.0
   +  9     => [1, "4.6.x or greater"],
   I've got libdb-4.5.a. I'll debug into that. It looks like it's only
   1.817, which is problematic.

   5.10.0 $VERSION = "1.816_1"
   5.8.9  $VERSION = "1.817" ;
   5.8.8  $VERSION = "1.814" ;

The HiRes failure was introduced between 5.8.8 and 5.8.9 and is still 
present.

With 5.8.8 all tests pass.
All tests successful.
u=4.26  s=7.67  cu=611.68  cs=471.24  scripts=970  tests=118358
make[2]: Leaving directory `/usr/src/perl/perl-5.8.8/buildperl'
make[1]: Leaving directory `/usr/src/perl/perl-5.8.8/buildperl'

The real problem is now to check why 5.8.9 cannot load the 5.8.8 arch 
dlls. Same Configure line of course.
Oops: default cygwin hints changed. Not good.
   5.8.8 uses system malloc, 5.8.9 now the faster perl MYMALLOC
But I'll make no patch for the porters, just a private for the cygwin 
package.
People which build from source should be aware of that.
Is there a prominent warning with -de?

> And is blead currently passing? (ie are the fixes already there, but not
> yet merged?)

With 5.10.0 it's similar, but db-btree.t is fixed.
ext/threads/t/thread..........................................FAILED--expected 
34 tests, saw 13
ext/Time/HiRes/t/HiRes........................................FAILED--non-zero 
wait status: 15

blead is still running (cygwin is much slower than linux...)
The last blead test @33408 had this:
ext/Time/HiRes/t/HiRes........................................FAILED--non-zero 
wait status: 15
lib/perl5db...................................................# Failed 
at ../lib/perl5db.t line 69
-- 
Reini Urban
http://phpwiki.org/  http://murbreak.at/

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