Bug #118730 for Sys-CpuAffinity: libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'

Sat Nov 12 15:32:55 2016 gregoa [...] cpan.org - Ticket created

From:	gregoa [...] cpan.org
Subject:	libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'

Message body is not shown because it is too large.

Mon Nov 14 11:02:12 2016 MOB [...] cpan.org - Correspondence added

Hi Gregor, Thank you very much for this report. I'm not doing anything fancy with the CPUs and I wouldn't expect the TSX extensions to be a problem. One thing I suspect is a bug around how the number 2**64 is represented in the Perl module. That may not be the most likely cause of this issue, but it is the easiest thing to check. Can you tell me if you have tested this module successfully on any systems without TSX extensions that had exactly 64 cpus? More than 64 cpus? Thanks, Marty O'Brien On Sat Nov 12 15:32:55 2016, GREGOA wrote: Show quoted text

> We have the following bug reported to the Debian package of > Sys-CpuAffinity (https://bugs.debian.org/844141): > > Please note the remark about a CPU with TSX extensions at the bottom. > > 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: libsys-cpuaffinity-perl > Version: 1.10-1 > Severity: serious > Tags: stretch sid > User: debian-qa@lists.debian.org > Usertags: qa-ftbfs-20161111 qa-ftbfs > Justification: FTBFS on amd64 > > Hi, > > During a rebuild of all packages in sid, your package failed to build > on > amd64. > > Relevant part (hopefully):

> >

... Show quoted text

> > # Failed test 'bind to all processors successful > > 18446744073709551615 == 18446744073709551616-1' > > # at t/10-exercise.t line 53. > > > > # Failed test 'setAffinity(-1) binds to all processors' > > # at t/10-exercise.t line 83. > > result: -1 3 ESRCH > > result: -1 3 ESRCH > > Sys::CpuAffinity: invalid mask 0 in call to setAffinty > > at t/10-exercise.t line 94. > > Sys::CpuAffinity: mask 18446744073709551616 is not valid for system > > with 64 processors. > > at t/10-exercise.t line 95. > > # Looks like you failed 2 tests of 14. > > t/10-exercise.t ...... > > 1..14

... Show quoted text

> > This failure happens on a CPU with TSX extensions available, but is > not > reproducible on a machine without them. For context, I recommend > reading the > thread starting at https://lists.debian.org/debian- > devel/2016/11/msg00210.html > > The node used is an Amazon EC2 VM with 64 cores. /proc/cpuinfo says: > model : 79 > model name : Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz > stepping : 1 > > A list of current common problems and possible solutions is available > at > http://wiki.debian.org/qa.debian.org/FTBFS . You're welcome to > contribute! > > About the archive rebuild: The rebuild was done on EC2 VM instances > from > Amazon Web Services, using a clean, minimal and up-to-date chroot. > Every > failed build was retried once to eliminate random failures. > > ------8<-----------8<-----------8<-----------8<-----------8<----- > > > Thanks for considering, > gregor herrmann, > Debian Perl Group

Mon Nov 14 11:02:12 2016 The RT System itself - Status changed from 'new' to 'open'

Mon Nov 14 15:17:03 2016 gregoa [...] debian.org - Correspondence added

CC:	Lucas Nussbaum <lucas [...] debian.org>
Subject:	Re: [rt.cpan.org #118730] libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'
Date:	Mon, 14 Nov 2016 21:16:45 +0100
To:	Marty O'Brien via RT <bug-Sys-CpuAffinity [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Mon, 14 Nov 2016 11:02:12 -0500, Marty O'Brien via RT wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 >

Show quoted text

> On Sat Nov 12 15:32:55 2016, GREGOA wrote:

> > We have the following bug reported to the Debian package of > > Sys-CpuAffinity (https://bugs.debian.org/844141):

Show quoted text

> Hi Gregor,

Show quoted text

> Thank you very much for this report. I'm not doing anything fancy > with the CPUs and I wouldn't expect the TSX extensions to be a > problem. One thing I suspect is a bug around how the number 2**64 > is represented in the Perl module. That may not be the most likely > cause of this issue, but it is the easiest thing to check. Can you > tell me if you have tested this module successfully on any systems > without TSX extensions that had exactly 64 cpus? More than 64 cpus?

Personally, I've tested the module only on my laptop :) Lucas who ran these specific tests added some information in https://lists.debian.org/debian-devel/2016/11/msg00486.html but not about the number of cores of the other test run. (CC'in Lucas, maybe he can share more information.) In the "normal" Debian infrastructure the package looks good: buildds: https://buildd.debian.org/status/logs.php?pkg=libsys-cpuaffinity-perl ci/autopkgtests: https://ci.debian.net/packages/libs/libsys-cpuaffinity-perl/unstable/amd64/ Oh, but there's a test failure on i386 on reproducible build machine: https://tests.reproducible-builds.org/debian/rb-pkg/unstable/i386/libsys-cpuaffinity-perl.html although that's a different failure. 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 of VIBE!AT & SPI, fellow of the Free Software Foundation Europe `- NP: Funny van Dannen: Sei nicht Traurig

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Mon Nov 14 15:56:15 2016 lucas [...] debian.org - Correspondence added

CC:	Marty O'Brien via RT <bug-Sys-CpuAffinity [...] rt.cpan.org>
Subject:	Re: [rt.cpan.org #118730] libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'
Date:	Mon, 14 Nov 2016 21:55:55 +0100
To:	gregor herrmann <gregoa [...] debian.org>
From:	Lucas Nussbaum <lucas [...] debian.org>

Hi, On 14/11/16 at 21:16 +0100, gregor herrmann wrote: Show quoted text

> On Mon, 14 Nov 2016 11:02:12 -0500, Marty O'Brien via RT wrote: >

> > <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 >

> > On Sat Nov 12 15:32:55 2016, GREGOA wrote:

> > > We have the following bug reported to the Debian package of > > > Sys-CpuAffinity (https://bugs.debian.org/844141):

> > Hi Gregor,

> > Thank you very much for this report. I'm not doing anything fancy > > with the CPUs and I wouldn't expect the TSX extensions to be a > > problem. One thing I suspect is a bug around how the number 2**64 > > is represented in the Perl module. That may not be the most likely > > cause of this issue, but it is the easiest thing to check. Can you > > tell me if you have tested this module successfully on any systems > > without TSX extensions that had exactly 64 cpus? More than 64 cpus?

> > Personally, I've tested the module only on my laptop :) > > Lucas who ran these specific tests added some information in > https://lists.debian.org/debian-devel/2016/11/msg00486.html > but not about the number of cores of the other test run. > (CC'in Lucas, maybe he can share more information.)

There are indeed 64 cores on this (virtual) machine. Good catch! Lucas

Tue Nov 22 12:39:23 2016 MOB [...] cpan.org - Correspondence added

RT-Send-CC:

gregoa [...] debian.org, lucas [...] debian.org

Would someone like to give http://search.cpan.org/CPAN/authors/id/M/MO/MOB/Sys-CpuAffinity-1.11_02.tar.gz a try? -- Marty On Mon Nov 14 15:56:15 2016, lucas@debian.org wrote: Show quoted text

> Hi, > > On 14/11/16 at 21:16 +0100, gregor herrmann wrote:

> > On Mon, 14 Nov 2016 11:02:12 -0500, Marty O'Brien via RT wrote: > >

> > > <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 >

> >

> > > On Sat Nov 12 15:32:55 2016, GREGOA wrote:

> > > > We have the following bug reported to the Debian package of > > > > Sys-CpuAffinity (https://bugs.debian.org/844141):

> >

> > > Hi Gregor,

> >

> > > Thank you very much for this report. I'm not doing anything fancy > > > with the CPUs and I wouldn't expect the TSX extensions to be a > > > problem. One thing I suspect is a bug around how the number 2**64 > > > is represented in the Perl module. That may not be the most likely > > > cause of this issue, but it is the easiest thing to check. Can you > > > tell me if you have tested this module successfully on any systems > > > without TSX extensions that had exactly 64 cpus? More than 64 cpus?

> > > > Personally, I've tested the module only on my laptop :) > > > > Lucas who ran these specific tests added some information in > > https://lists.debian.org/debian-devel/2016/11/msg00486.html > > but not about the number of cores of the other test run. > > (CC'in Lucas, maybe he can share more information.)

> > There are indeed 64 cores on this (virtual) machine. Good catch! > > Lucas

Tue Nov 22 13:26:21 2016 gregoa [...] debian.org - Correspondence added

CC:	lucas [...] debian.org
Subject:	Re: [rt.cpan.org #118730] libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'
Date:	Tue, 22 Nov 2016 19:25:53 +0100
To:	Marty O'Brien via RT <bug-Sys-CpuAffinity [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Tue, 22 Nov 2016 12:39:24 -0500, Marty O'Brien via RT wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 > > > Would someone like to give http://search.cpan.org/CPAN/authors/id/M/MO/MOB/Sys-CpuAffinity-1.11_02.tar.gz a try?

Sure, uploaded to Debian/unstable. Thanks for working on a fix! "Normal" results should show up at https://buildd.debian.org/status/package.php?p=libsys-cpuaffinity-perl and https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/libsys-cpuaffinity-perl.html https://tests.reproducible-builds.org/debian/rb-pkg/unstable/i386/libsys-cpuaffinity-perl.html https://tests.reproducible-builds.org/debian/rb-pkg/unstable/armhf/libsys-cpuaffinity-perl.html and https://ci.debian.net/packages/libs/libsys-cpuaffinity-perl/ Lucas, do you have a chance to test libsys-cpuaffinity-perl_1.11~02-1 on this 64 core VM where you originally detected the issue (#844141)? 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 of VIBE!AT & SPI, fellow of the Free Software Foundation Europe `- NP: John Lennon: Power To The People

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.

Mon Dec 05 04:19:08 2016 lucas [...] debian.org - Correspondence added

CC:	Marty O'Brien via RT <bug-Sys-CpuAffinity [...] rt.cpan.org>
Subject:	Re: [rt.cpan.org #118730] libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'
Date:	Mon, 5 Dec 2016 10:16:56 +0100
To:	gregor herrmann <gregoa [...] debian.org>
From:	Lucas Nussbaum <lucas [...] debian.org>

On 22/11/16 at 19:25 +0100, gregor herrmann wrote: Show quoted text

> On Tue, 22 Nov 2016 12:39:24 -0500, Marty O'Brien via RT wrote: >

> > <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 > > > > > Would someone like to give http://search.cpan.org/CPAN/authors/id/M/MO/MOB/Sys-CpuAffinity-1.11_02.tar.gz a try?

> > > Sure, uploaded to Debian/unstable. > Thanks for working on a fix! > > > "Normal" results should show up at > https://buildd.debian.org/status/package.php?p=libsys-cpuaffinity-perl > and > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/libsys-cpuaffinity-perl.html > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/i386/libsys-cpuaffinity-perl.html > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/armhf/libsys-cpuaffinity-perl.html > and > https://ci.debian.net/packages/libs/libsys-cpuaffinity-perl/ > > Lucas, do you have a chance to test libsys-cpuaffinity-perl_1.11~02-1 > on this 64 core VM where you originally detected the issue (#844141)?

Hi, Sorry for not replying earlier. It still fails. Build log attached. Lucas

Message body is not shown because sender requested not to inline it.

Mon Dec 05 10:53:15 2016 MOB [...] cpan.org - Correspondence added

Thanks, Lucas. This is very interesting. To me this looks like an issue with Math::BigInt. It seems to be evaluating the expression 18446744073709551615 + 1 == 18446744073709551616 as false, even as it evaluates 18446744073709551615 == 18446744073709551616 - 1 as true. I've uploaded Sys-CpuAffinity-1.12_01.tar.gz which performs some additional sanity checks on Math::BigInt, and which replaces tests of the form $x+1 == $y with the form $x == $y-1 . Hopefully this will either work or give me some new information about what is going on. -- Marty

Mon Dec 05 13:17:19 2016 gregoa [...] debian.org - Correspondence added

CC:	Lucas Nussbaum <lucas [...] debian.org>
Subject:	Re: [rt.cpan.org #118730] libsys-cpuaffinity-perl: FTBFS: Failed test 'bind to all processors successful 18446744073709551615 == 18446744073709551616-1'
Date:	Mon, 5 Dec 2016 19:16:56 +0100
To:	Marty O'Brien via RT <bug-Sys-CpuAffinity [...] rt.cpan.org>
From:	gregor herrmann <gregoa [...] debian.org>

On Mon, 05 Dec 2016 10:53:15 -0500, Marty O'Brien via RT wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=118730 > > > Thanks, Lucas. This is very interesting. > > I've uploaded Sys-CpuAffinity-1.12_01.tar.gz which performs some > additional sanity checks on Math::BigInt, and which replaces tests > of the form $x+1 == $y with the form $x == $y-1 . Hopefully this > will either work or give me some new information about what is > going on.

Thanks! Uploaded to Debian as libsys-cpuaffinity-perl_1.12~01-1. 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 of VIBE!AT & SPI, fellow of the Free Software Foundation Europe `- NP: Dire Straits: Romeo And Juliet

Download signature.asc
application/pgp-signature 963b

Message body not shown because it is not plain text.