Bug #126294 for PDL-CCS: libpdl-ccs-perl FTBFS on architectures where char is unsigned

Thu Aug 16 17:27:30 2018 gregoa [...] cpan.org - Ticket created

From:	gregoa [...] cpan.org
Subject:	libpdl-ccs-perl FTBFS on architectures where char is unsigned

We have the following bug reported to the Debian package of PDL-CCS (https://bugs.debian.org/902192): It doesn't seem to be a bug in the packaging, so you may want to take a look. Thanks! ------8<-----------8<-----------8<-----------8<-----------8<----- Source: libpdl-ccs-perl Version: 1.23.9-1 Severity: important Tags: ftbfs https://buildd.debian.org/status/package.php?p=libpdl-ccs-perl&suite=sid ... ... ------8<-----------8<-----------8<-----------8<-----------8<----- This can also be seen on http://matrix.cpantesters.org/?dist=PDL-CCS%201.23.9;os=linux;perl=5.26.2;reports=1 (the failing smoke test happens to be from me, from a raspberry pi :)) Thanks for considering, gregor herrmann, Debian Perl Group

Fri Aug 24 04:52:59 2018 moocow [...] cpan.org - Correspondence added

moin Gregor, Many thanks for the report, and apologies for the delayed response. The smoker output indicates that PDL::CCS::Ufunc::borover() is failing the presence of BAD values, which is very odd since the analogous tests for andover(), orover(), and bandover() are all passing (t/01_ufunc.t tests #44-46). Signedness of char shouldn't make any difference to the failing function, since the piddles ought to be auto-magically converted to the largest integer type available on the system (recorded in CCS/Config.pm on `perl Makefile.PL`). In fact, the failing function appears to be returning the *correct* values ("got=[-2 11 15 15 9 6]") and something is apparently wonky with PDL::borover() -- which I use to generate the "wanted" return values for testing -- on that platform. If so, I'd like to bring this to the PDL maintainers' attention and get the underlying problem resolved, but... Embarrassingly, I discovered a typo in the test code responsible for bogus "wanted=(undef)" diagnostics in the smoker output -- that bug has been fixed (and 'use strict' added) in PDL::CCS v1.23.10. Can you please repeat the smoker test for v1.23.10 (now on CPAN)? I still expect the borover() test to fail, but hopefully with more helpful diagnostics that I can pass on to the PDL folks. If all else fails, I can either add an exception for this particular test or just skip it on appropriate platforms. marmosets, Bryan / moocow On 2018-08-16 17:27:30, GREGOA wrote: Show quoted text

> We have the following bug reported to the Debian package of PDL-CCS > (https://bugs.debian.org/902192): > > It doesn't seem to be a bug in the packaging, so you may want to take > a look. Thanks! > > ------8<-----------8<-----------8<-----------8<-----------8<----- > > Source: libpdl-ccs-perl > Version: 1.23.9-1 > Severity: important > Tags: ftbfs > > https://buildd.debian.org/status/package.php?p=libpdl-ccs- > perl&suite=sid > > ... > ... > > > ------8<-----------8<-----------8<-----------8<-----------8<----- > > This can also be seen on > http://matrix.cpantesters.org/?dist=PDL- > CCS%201.23.9;os=linux;perl=5.26.2;reports=1 > (the failing smoke test happens to be from me, from a raspberry pi :)) > > > Thanks for considering, > gregor herrmann, > Debian Perl Group

Fri Aug 24 04:52:59 2018 The RT System itself - Status changed from 'new' to 'open'

Fri Aug 24 04:53:12 2018 moocow [...] cpan.org - Taken

Fri Aug 24 10:18:45 2018 gregoa [...] debian.org - Correspondence added

Subject:	Re: [rt.cpan.org #126294] libpdl-ccs-perl FTBFS on architectures where char is unsigned
Date:	Fri, 24 Aug 2018 16:18:22 +0200
To:	Bryan Jurish via RT <bug-PDL-CCS [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Fri, 24 Aug 2018 04:52:59 -0400, Bryan Jurish via RT wrote: Show quoted text

> Many thanks for the report, and apologies for the delayed response.

No problem. Show quoted text

> Embarrassingly, I discovered a typo in the test code responsible > for bogus "wanted=(undef)" diagnostics in the smoker output -- that > bug has been fixed (and 'use strict' added) in PDL::CCS v1.23.10. > Can you please repeat the smoker test for v1.23.10 (now on CPAN)? I > still expect the borover() test to fail, but hopefully with more > helpful diagnostics that I can pass on to the PDL folks.

I just started a test on my raspi, so you should get the report on cpantesters once it finishes. Additionally, I've uploaded 1.23.10 to Debian, so we will see build logs for all archtitectures soon at https://buildd.debian.org/status/package.php?p=libpdl-ccs-perl or for all versions at https://buildd.debian.org/status/logs.php?pkg=libpdl-ccs-perl Cheers, gregor -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- NP: Anouar Brahem: Leila au pays du carrousel

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Fri Aug 24 13:16:42 2018 gregoa [...] debian.org - Correspondence added

Subject:	Re: [rt.cpan.org #126294] libpdl-ccs-perl FTBFS on architectures where char is unsigned
Date:	Fri, 24 Aug 2018 19:16:22 +0200
To:	gregor herrmann via RT <bug-PDL-CCS [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Fri, 24 Aug 2018 10:18:45 -0400, gregor herrmann via RT wrote: Show quoted text

> I just started a test on my raspi, so you should get the report on > cpantesters once it finishes.

It's about the finish, and here's the test failure: make[2]: Entering directory '/srv/smoker/cpan/build/PDL-CCS-1.23.10-0/CCS/Ufunc' "/srv/smoker/perl5/perlbrew/perls/perl-5.28.0/bin/perl" -MExtUtils::Command::MM -e 'cp_nonempty' -- Ufunc.bs ../../blib/arch/auto/PDL/CCS/Ufunc/Ufunc.bs 644 PERL_DL_NONLAZY=1 "/srv/smoker/perl5/perlbrew/perls/perl-5.28.0/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, '../../blib/lib', '../../blib/arch')" t/*.t t/00_basic.t .. ok # Failed test 'borover:missing=BAD:vals # : got=[-2 11 15 15 9 6] # : wanted=[254 11 15 15 9 6]' # at /srv/smoker/cpan/build/PDL-CCS-1.23.10-0/t/common.plt line 11. # Looks like you failed 1 test of 52. t/01_ufunc.t .. Dubious, test returned 1 (wstat 256, 0x100) Failed 1/52 subtests Test Summary Report ------------------- t/01_ufunc.t (Wstat: 256 Tests: 52 Failed: 1) Failed test: 47 Non-zero exit status: 1 Files=2, Tests=53, 15 wallclock secs ( 0.72 usr 0.10 sys + 8.70 cusr 0.48 csys = 10.00 CPU) Result: FAIL Failed 1/2 test programs. 1/53 subtests failed. make[2]: *** [Makefile:851: test_dynamic] Error 1 make[2]: Leaving directory '/srv/smoker/cpan/build/PDL-CCS-1.23.10-0/CCS/Ufunc' make[1]: *** [Makefile:756: subdirs-test_dynamic] Error 2 make[1]: Leaving directory '/srv/smoker/cpan/build/PDL-CCS-1.23.10-0/CCS' Looks like the same as the ones from https://buildd.debian.org/status/logs.php?pkg=libpdl-ccs-perl Cheers, gregor -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- NP: Bruce Springsteen: Forever Young

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Mon Aug 27 08:53:09 2018 moocow [...] cpan.org - Correspondence added

Thanks for the update. I take it all back and assert the contrary: it *does* seem to be an (unsigned char) issue after all. I think I've identified the culprit in PDL::borover(), which does indeed use a temporary 'char' variable only in the presence of BAD values -- which is the only case where the CCS borover() variant is failing. I've reported the issue to the PDL maintainers as https://sourceforge.net/p/pdl/bugs/446/ . In the meantime, I've added some code in the test suite for PDL::CCS v1.23.11 (now on CPAN) to detect broken PDL::borover() and skip the failing test. That ought to eliminate the bogus failure reports and resolve the issue. I'll be watching the cpantesters reports in the next few days, and hopefully we'll see armv6l-linux go green for PDL::CCS. marmosets, Bryan On 2018-08-24 13:16:42, gregoa@debian.org wrote: Show quoted text

> # Failed test 'borover:missing=BAD:vals > # : got=[-2 11 15 15 9 6] > # : wanted=[254 11 15 15 9 6]' > # at /srv/smoker/cpan/build/PDL-CCS-1.23.10-0/t/common.plt line 11. > # Looks like you failed 1 test of 52.

Mon Aug 27 15:50:01 2018 gregoa [...] debian.org - Correspondence added

Subject:	Re: [rt.cpan.org #126294] libpdl-ccs-perl FTBFS on architectures where char is unsigned
Date:	Mon, 27 Aug 2018 21:49:41 +0200
To:	Bryan Jurish via RT <bug-PDL-CCS [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Mon, 27 Aug 2018 08:53:10 -0400, Bryan Jurish via RT wrote: Show quoted text

> In the meantime, I've added some code in the test suite for > PDL::CCS v1.23.11 (now on CPAN) to detect broken PDL::borover() and > skip the failing test. That ought to eliminate the bogus failure > reports and resolve the issue. I'll be watching the cpantesters > reports in the next few days, and hopefully we'll see armv6l-linux > go green for PDL::CCS.

Thanks for your work on this issue! Unfortunately it didn't work out as planned on my raspi: PERL_DL_NONLAZY=1 "/srv/smoker/perl5/perlbrew/perls/perl-5.28.0/bin/perl" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_harness(0, '../blib/lib', '../blib/arch')" t/*.t t/00_basic.t ..... ok t/01_encode.t .... ok t/02_indexing.t .. ok # Failed test 'borover:missing=BAD:vals # : got=[-2 11 15 15 15 6] # : wanted=[254 11 15 15 15 6]' # at /srv/smoker/cpan/build/PDL-CCS-1.23.11-0/t/common.plt line 11. # Looks like you failed 1 test of 136. t/03_ufuncs.t .... Dubious, test returned 1 (wstat 256, 0x100) Failed 1/136 subtests t/04_unops.t ..... ok t/05_binops.t .... ok t/06_matops.t .... ok Test Summary Report ------------------- t/03_ufuncs.t (Wstat: 256 Tests: 136 Failed: 1) Failed test: 118 Non-zero exit status: 1 Files=7, Tests=4881, 192 wallclock secs (13.27 usr 0.57 sys + 151.62 cusr 3.07 csys = 168.53 CPU) Result: FAIL Failed 1/7 test programs. 1/4881 subtests failed. make[1]: *** [Makefile:762: test_dynamic] Error 255 make[1]: Leaving directory '/srv/smoker/cpan/build/PDL-CCS-1.23.11-0/CCS' make: *** [Makefile:963: subdirs-test_dynamic] Error 2 (/usr/bin/make test exited with 512) CPAN::Reporter: Test result is 'fail', 'make test' error detected. CPAN::Reporter: preparing a CPAN Testers report for PDL-CCS-1.23.11 Do you want to review or edit the test report? (yes/no) [no] no Do you want to send the report? (yes/no) [yes] yes CPAN::Reporter: sending test report with 'fail' via Metabase Cheers, gregor -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- NP: Ludwig Hirsch: Asta

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Tue Aug 28 02:57:10 2018 moocow [...] cpan.org - Correspondence added

On 2018-08-27 15:50:01, gregoa@debian.org wrote: Show quoted text

> Thanks for your work on this issue!

No worries, just trying to do things "right". Show quoted text

> Unfortunately it didn't work out as planned on my raspi:

... Show quoted text

> # Failed test 'borover:missing=BAD:vals > # : got=[-2 11 15 15 15 6] > # : wanted=[254 11 15 15 15 6]' > # at /srv/smoker/cpan/build/PDL-CCS-1.23.11-0/t/common.plt line 11. > # Looks like you failed 1 test of 136.

Argh. I just saw the cpantesters report fot this too, and am thoroughly confused. Can you run: $ perl -M5.10.0 -MPDL -e 'say pdl(double,[10,0,-2])->setvaltobad(0)->borover->sclr' on the raspi and tell me what the output is (presumably either 254 or -2)?

Tue Aug 28 03:17:08 2018 gregoa [...] debian.org - Correspondence added

Subject:	Re: [rt.cpan.org #126294] libpdl-ccs-perl FTBFS on architectures where char is unsigned
Date:	Tue, 28 Aug 2018 09:16:51 +0200
To:	Bryan Jurish via RT <bug-PDL-CCS [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Tue, 28 Aug 2018 02:57:10 -0400, Bryan Jurish via RT wrote: Show quoted text

> Argh. I just saw the cpantesters report fot this too, and am > thoroughly confused. Can you run: > > $ perl -M5.10.0 -MPDL -e 'say pdl(double,[10,0,-2])->setvaltobad(0)->borover->sclr' > > on the raspi and tell me what the output is (presumably either 254 > or -2)?

Sure, here we go: $ PERL5LIB=~/cpan/build/PDL-2.019-0/blib/lib:~/cpan/build/PDL-2.019-0/blib/arch perl -M5.10.0 -MPDL -e 'say pdl(double,[10,0,-2])->setvaltobad(0)->borover->sclr' 254 Cheers, gregor -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- BOFH excuse #412: Radial Telemetry Infiltration

Tue Aug 28 03:35:28 2018 moocow [...] cpan.org - Correspondence added

On 2018-08-28 03:17:08, gregoa@debian.org wrote: Show quoted text

> $ PERL5LIB=~/cpan/build/PDL-2.019-0/blib/lib:~/cpan/build/PDL-2.019- > 0/blib/arch perl -M5.10.0 -MPDL -e 'say pdl(double,[10,0,-2])-

> >setvaltobad(0)->borover->sclr'

> 254

Thanks for the speedy reply! It turns out that it's a *different* test failing now, and not the one for which I had added the skip-check: my bad for not remembering that borover() gets tested twice (once for the low-level function PDL::CCS::Ufunc::borover(), once for the OO wrappers PDL::CCS::Nd::borover()). I've duplicated the skip-check from the (previously failing but now apparently succeeding) CCS/Ufunc/t/01_ufunc.t to the (now failing) CCS/t/03_ufuncs.t, and uploaded to CPAN as PDL-CCS v1.23.12. Crossing fingers. marmosets, Bryan

Tue Aug 28 10:36:13 2018 gregoa [...] debian.org - Correspondence added

Subject:	Re: [rt.cpan.org #126294] libpdl-ccs-perl FTBFS on architectures where char is unsigned
Date:	Tue, 28 Aug 2018 16:35:51 +0200
To:	Bryan Jurish via RT <bug-PDL-CCS [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Tue, 28 Aug 2018 03:35:28 -0400, Bryan Jurish via RT wrote: Show quoted text

> Thanks for the speedy reply! It turns out that it's a *different* > test failing now, and not the one for which I had added the > skip-check: my bad for not remembering that borover() gets tested > twice (once for the low-level function PDL::CCS::Ufunc::borover(), > once for the OO wrappers PDL::CCS::Nd::borover()).

Heh :) Show quoted text

> I've duplicated the skip-check from the (previously failing but now > apparently succeeding) CCS/Ufunc/t/01_ufunc.t to the (now failing) > CCS/t/03_ufuncs.t, and uploaded to CPAN as PDL-CCS v1.23.12. > Crossing fingers.

And success! All tests pass, cpantesters report sent: CPAN::Reporter: Test result is 'pass', 'make test' no errors. CPAN::Reporter: preparing a CPAN Testers report for PDL-CCS-1.23.12 CPAN::Reporter: sending test report with 'pass' via Metabase Cheers, gregor -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- NP: Tori Amos: Cornflake Girl

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Wed Aug 29 02:46:51 2018 moocow [...] cpan.org - Correspondence added

yay! The debian logs https://buildd.debian.org/status/package.php?p=libpdl-ccs-perl seem to indicate that the skip-check was successful for other (unsigned char) architectures too (e.g. https://buildd.debian.org/status/logs.php?pkg=libpdl-ccs-perl&arch=powerpc is now succeeding too), so I'm marking this issue as resolved. Thanks again for following up on this (and of course for providing the rpi smoker!) marmosets, Bryan

Wed Aug 29 02:46:51 2018 moocow [...] cpan.org - Status changed from 'open' to 'resolved'

Wed Aug 29 02:46:52 2018 moocow [...] cpan.org - Fixed in 1.23.12 added