Bug #41369 for NetAddr-IP: NetAddr-IP v4.019-1 fails on Solaris-10

Tue Dec 02 07:07:59 2008 andrew.benham [...] thus.net - Ticket created

Subject:	NetAddr-IP v4.019-1 fails on Solaris-10
Date:	Tue, 02 Dec 2008 12:07:05 +0000
To:	bug-NetAddr-IP [...] rt.cpan.org
From:	Andrew Benham <andrew.benham [...] thus.net>

NetAddr-IP v4.019-1 Solaris-10 (both sparc and x86). Perl v5.8.4 Sun Studio 12 compiler 3 tests fail: Failed Test Stat Wstat Total Fail Failed List of Failed ------------------------------------------------------------------------------- t/splitref.t 0 139 ?? ?? % ?? t/v4-split-bulk.t 0 139 8 14 175.00% 2-8 t/v6-split-bulk.t 0 139 6 10 166.67% 2-6 Failed 3/27 test scripts, 88.89% okay. 7/9504 subtests failed, 99.93% okay. .... t/splitref.........ok 5/0 t/splitref.........dubious Test returned status 0 (wstat 139, 0x8b) test program seems to have generated a core .... t/v4-split-bulk....ok 1/8 t/v4-split-bulk....dubious Test returned status 0 (wstat 139, 0x8b) test program seems to have generated a core DIED. FAILED tests 2-8 Failed 7/8 tests, 12.50% okay .... t/v6-split-bulk....ok 1/6 t/v6-split-bulk....dubious Test returned status 0 (wstat 139, 0x8b) test program seems to have generated a core DIED. FAILED tests 2-6 Failed 5/6 tests, 16.67% okay If I run these tests with "make testdb" then they all pass, so it's difficult to debug. -- Andrew Benham Demon@Thus@Cable&Wireless andrew.benham@thus.net Finchley, London N3 2QQ, U.K. Tel: 020 8495 6343 Fax: 020 8495 6037

Tue Dec 09 19:25:38 2008 michael [...] bizsystems.com - Correspondence added 20 min

On Tue Dec 02 07:07:59 2008, andrew.benham@thus.net wrote: Show quoted text

> NetAddr-IP v4.019-1 > Solaris-10 (both sparc and x86). > Perl v5.8.4 > Sun Studio 12 compiler > > 3 tests fail: > Failed Test Stat Wstat Total Fail Failed List of Failed >

------------------------------------------------------------------------------- Show quoted text

> t/splitref.t 0 139 ?? ?? % ?? > t/v4-split-bulk.t 0 139 8 14 175.00% 2-8 > t/v6-split-bulk.t 0 139 6 10 166.67% 2-6 > Failed 3/27 test scripts, 88.89% okay. 7/9504 subtests failed, 99.93% > okay. > > .... > t/splitref.........ok 5/0 > > t/splitref.........dubious > Test returned status 0 (wstat 139, 0x8b) > test program seems to have generated a core > > .... > t/v4-split-bulk....ok 1/8 > > t/v4-split-bulk....dubious > Test returned status 0 (wstat 139, 0x8b) > test program seems to have generated a core > DIED. FAILED tests 2-8 > Failed 7/8 tests, 12.50% okay > > .... > t/v6-split-bulk....ok 1/6 > > t/v6-split-bulk....dubious > Test returned status 0 (wstat 139, 0x8b) > test program seems to have generated a core > DIED. FAILED tests 2-6 > Failed 5/6 tests, 16.67% okay > > > If I run these tests with "make testdb" then they all pass, so it's > difficult to debug. >

This test was run with perl 5.8.4 Unfortunately, this version of perl (also 5.8.0 - 5.8.5) have serious bugs in the handling of "@_". In particular, perl bug #[ 23429] causes massive failures in many programs that use "goto &sub". There are several other insidious bugs in these particular releases of perl. Down grade to 5.6.2 or upgrade to 5.8.6 or better. Yes, I know Sun recommends 5.8.4 for Solaris.... but is is very broken. I've added some work-arounds for the most egregious failures due to the "goto &sub" bug and it runs on my sun box now under 5.8.4, but I believe there are other problems with 5.8.4 in particular that may cause failures unrelated to the module code. Try version NetAddr::IP 4.020 and let me know the results. Michael

Tue Dec 09 19:25:40 2008 The RT System itself - Status changed from 'new' to 'open'

Tue Dec 09 19:25:40 2008 michael [...] bizsystems.com - Status changed from 'open' to 'rejected'

Tue Dec 09 19:25:42 2008 michael [...] bizsystems.com - Given to MIKER

Wed Dec 10 07:12:21 2008 paul [...] city-fan.org - Correspondence added

From:

paul [...] city-fan.org

On Tue Dec 09 19:25:38 2008, MIKER wrote: Show quoted text

> On Tue Dec 02 07:07:59 2008, andrew.benham@thus.net wrote:

> > NetAddr-IP v4.019-1 > > Solaris-10 (both sparc and x86). > > Perl v5.8.4 > > Sun Studio 12 compiler > > > > 3 tests fail: > > Failed Test Stat Wstat Total Fail Failed List of Failed > >

>

------------------------------------------------------------------------------- Show quoted text

> > t/splitref.t 0 139 ?? ?? % ?? > > t/v4-split-bulk.t 0 139 8 14 175.00% 2-8 > > t/v6-split-bulk.t 0 139 6 10 166.67% 2-6 > > Failed 3/27 test scripts, 88.89% okay. 7/9504 subtests failed,

> 99.93%

> > okay. > > > > .... > > t/splitref.........ok 5/0 > > > > t/splitref.........dubious > > Test returned status 0 (wstat 139, 0x8b) > > test program seems to have generated a core > > > > .... > > t/v4-split-bulk....ok 1/8 > > > > t/v4-split-bulk....dubious > > Test returned status 0 (wstat 139, 0x8b) > > test program seems to have generated a core > > DIED. FAILED tests 2-8 > > Failed 7/8 tests, 12.50% okay > > > > .... > > t/v6-split-bulk....ok 1/6 > > > > t/v6-split-bulk....dubious > > Test returned status 0 (wstat 139, 0x8b) > > test program seems to have generated a core > > DIED. FAILED tests 2-6 > > Failed 5/6 tests, 16.67% okay > > > > > > If I run these tests with "make testdb" then they all pass, so it's > > difficult to debug. > >

> > This test was run with perl 5.8.4 > > Unfortunately, this version of perl (also 5.8.0 - 5.8.5) have serious > bugs in the handling of "@_". In particular, perl bug #[ 23429] causes > massive failures in many programs that use "goto &sub". There are > several other insidious bugs in these particular releases of perl. > > Down grade to 5.6.2 or upgrade to 5.8.6 or better. Yes, I know Sun > recommends 5.8.4 for Solaris.... but is is very broken. > > I've added some work-arounds for the most egregious failures due to > the > "goto &sub" bug and it runs on my sun box now under 5.8.4, but I > believe > there are other problems with 5.8.4 in particular that may cause > failures unrelated to the module code. > > Try version NetAddr::IP 4.020 and let me know the results. > > Michael

This problem also affected Red Hat Enterprise Linux 4 (and clones), which ship with perl 5.8.5. Version 4.015 of NetAddr-IP was the last one that got through the test suite successfully on this platform until 4.020, which has resolved the issue. However, the warning about broken perl versions produced during the build looks like it needs some work. Running on the current Fedora development tree with perl 5.10.0 produces this, which I'm confident isn't intentional: ######################### WARNING ############################### # # # YOUR VERSION OF PERL = 5.10.0 HAS SERIOUS BUGS # # # # Early versions of perl 5.8.x contain bugs that cause certain # # @_ operations to fail. Please upgrade to at least perl 5.8.6 # # # #################################################################

Wed Dec 10 07:12:26 2008 The RT System itself - Status changed from 'rejected' to 'open'

Wed Dec 10 14:32:43 2008 michael [...] bizsystems.com - Correspondence added 10 min

RT-Send-CC:

paul [...] city-fan.org

<snip> Show quoted text

> This problem also affected Red Hat Enterprise Linux 4 (and clones), > which ship with perl 5.8.5. Version 4.015 of NetAddr-IP was the last one > that got through the test suite successfully on this platform until > 4.020, which has resolved the issue. > > However, the warning about broken perl versions produced during the > build looks like it needs some work. Running on the current Fedora > development tree with perl 5.10.0 produces this, which I'm confident > isn't intentional: > > ######################### WARNING ############################### > # # > # YOUR VERSION OF PERL = 5.10.0 HAS SERIOUS BUGS # > # # > # Early versions of perl 5.8.x contain bugs that cause certain # > # @_ operations to fail. Please upgrade to at least perl 5.8.6 # > # # > ################################################################# >

aaarrggghhhhHHH! I left in a test edit to force the failure on my development box that does not have the "bad" perl version. I've uploaded version 4.021 with the corrected perl version test to CPAN.

Wed Dec 10 14:33:02 2008 michael [...] bizsystems.com - Status changed from 'open' to 'resolved'