Debian Bug report logs - #836599
shark: FTBFS on mips/mipsel: test errors

Display info messages

Message #5 received at [email protected] (full text, mbox, reply):

From: Emilio Pozuelo Monfort <[email protected]>

To: Debian Bug Tracking System <[email protected]>

Subject: shark: FTBFS on mips/mipsel: test errors

Date: Sun, 04 Sep 2016 13:59:35 +0200

Source: shark
Version: 3.1.3+ds1-1
Severity: serious

Your package failed to build on mips/mipsel:

95% tests passed, 8 tests failed out of 163

Logs at
https://buildd.debian.org/status/package.php?p=shark&suite=unstable

Emilio

Message #10 received at [email protected] (full text, mbox, reply):

From: Ghislain Vaillant <[email protected]>

To: [email protected]

Cc: [email protected]

Subject: forwarded upstream

Date: Sun, 4 Sep 2016 14:23:31 +0100

control: forwarded -1 https://github.com/Shark-ML/Shark/issues/112

Forwarded upstream. Not sure whether this package was intended to work
on these architectures, so upstream might just not care about this.

I'll wait for their reply. The options are:
- Upstream fixes it hopefully.
- Disable testing for mips / mipsel in d/rules.
- Drop shark from the archive for mips / mipsel.

Feel free to comment. Thanks again for the report Emilio.

Ghis

Message #17 received at [email protected] (full text, mbox, reply):

From: Ghislain Vaillant <[email protected]>

To: [email protected]

Subject: shark binaries to be removed from mips and mipsel

Date: Mon, 5 Sep 2016 12:31:30 +0100

Upstream failed to find the cause for the test issues.

I have requested the removal of the shark binaries currently sitting in
testing for mips and mipsel. Once done, I plan to downgrade this bug to
a non-RC severity to allow the package to transition.

I guess it will be up to mips porters to get it to work on these
architectures.

Cheers,
Ghis

Message #22 received at [email protected] (full text, mbox, reply):

From: Ghislain Vaillant <[email protected]>

To: [email protected]

Subject: downgrading severity to normal

Date: Thu, 15 Sep 2016 08:53:55 +0100

control: severity -1 normal

Previous builds on mips and mipsel have been removed [1]. Downgrading 
the severity of this bug to allow the updated package to transition.

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=836740

Cheers,
Ghis

Message #29 received at [email protected] (full text, mbox, reply):

From: Dejan Latinovic <[email protected]>

To: "[email protected]" <[email protected]>

Subject: RE: Bug#836599: shark: FTBFS on mips/mipsel: test errors

Date: Thu, 15 Sep 2016 13:28:14 +0000

[Message part 1 (text/plain, inline)]

Hello,

I took a look at this issue.
The reason of an illegal instruction is related to ldxc1 instruction.
> ldxc1	$f0,v0(a1)

> (gdb) p/x $a1
> $50 = 0xaaa8ae10
> (gdb) p/x $v0
> $51 = 0xaaaf0248

The address (a1+v0) seems to be valid and aligned to 8, but causes error anyway.
I will investigate this issue further.

This behavior is triggered by O2 optimization level. If O1 is used, this issue will not appear.

I have created patch that enables O1 flag for failing tests.

Also it was needed to mark ObjFunct_SvmLogisticInterpretation as slow as this test failed with timeout (1500) on my local build.

Both patches are attached.

Regards,
Dejan

[fix-build-on-MIPS.patch (text/x-patch, attachment)]

[mark-ObjFunct_SvmLogisticInterpretation-as-slow.patch (text/x-patch, attachment)]

Message #34 received at [email protected] (full text, mbox, reply):

From: Dejan Latinovic <[email protected]>

To: "[email protected]" <[email protected]>

Cc: Emilio Pozuelo Monfort <[email protected]>, Ghislain Vaillant <[email protected]>

Subject: RE: Bug#836599: shark: FTBFS on mips/mipsel: test errors

Date: Fri, 14 Oct 2016 13:18:42 +0000

Hi,

I after further investigation I have discovered that this address (a1+v0)  is valid in pure 32-bit system,
but if we have 32-bits Debian on 64-bits machine the address is not valid.
This happens because compiler for some reason counting on overflow in address calculation.
Kernel does not cover this situation, so it gives SIGBUS or SIGILL.

This issue is reported upstream:
https://dmz-portal.mips.com/bugz/show_bug.cgi?id=2408

I was not able to find better solution at this moment then the one that I had already proposed: disabling 02 as it is done in this patch fix-build-on-MIPS.patch.
Could you please consider including the patch?

Regards,
Dejan

Message #39 received at [email protected] (full text, mbox, reply):

From: Ghislain Vaillant <[email protected]>

To: [email protected]

Cc: Dejan Latinovic <[email protected]>

Subject: RE: Bug#836599: shark: FTBFS on mips/mipsel: test errors

Date: Fri, 14 Oct 2016 16:06:33 +0100

Hi Dejan,

Thanks for investigating the build issue of shark with mips and for
providing patches. Please consider updating these patches with an
appropriate DEP3 header [1], otherwise I can't accept them.

[1] http://dep.debian.net/deps/dep3/

One thing I also wonder is how I will know when to drop the MIPS patch?
Will I need to monitor the progress of the bug report you mentioned
myself? Please add as much information as possible in the Description
field of the DEP3 header to help future maintenance.

Cheers,
Ghis

Message #44 received at [email protected] (full text, mbox, reply):

From: Dejan Latinovic <[email protected]>

To: Ghislain Vaillant <[email protected]>, "[email protected]" <[email protected]>

Subject: RE: Bug#836599: shark: FTBFS on mips/mipsel: test errors

Date: Tue, 18 Oct 2016 11:36:52 +0000

[Message part 1 (text/plain, inline)]

Control: tags -1 + patch
Control: user -1 [email protected]
Control: usertags -1 mips-patch


Hi Ghislain,

I have updated the patch that reduces optimization level with requested information, Fix-build-on-MIPS.patch‎.
Do you want me to add anything else?

Maybe it would be enough to apply only this patch, as the second one was needed for
one of my local machines that may have lower performance than Debian's build servers.

I will monitor the status of reported issues and ask for patch removal when the time comes.


Regards,
Dejan

[Fix-build-on-MIPS.patch (text/x-patch, attachment)]

Message #51 received at [email protected] (full text, mbox, reply):

From: Ghislain Vaillant <[email protected]>

To: [email protected]

Cc: Dejan Latinovic <[email protected]>

Subject: Re: RE: Bug#836599: shark: FTBFS on mips/mipsel: test errors

Date: Fri, 25 Nov 2016 19:21:54 +0000

On Tue, 18 Oct 2016 11:36:52 +0000 Dejan Latinovic 
<[email protected]> wrote:
> Control: tags -1 + patch
> Control: user -1 [email protected]
> Control: usertags -1 mips-patch
>
>
> Hi Ghislain,
>
> I have updated the patch that reduces optimization level with requested information, Fix-build-on-MIPS.patchý.
> Do you want me to add anything else?
>
> Maybe it would be enough to apply only this patch, as the second one was needed for
> one of my local machines that may have lower performance than Debian's build servers.
>
> I will monitor the status of reported issues and ask for patch removal when the time comes.
>
>
> Regards,
> Dejan
>

I have applied your patch and launched a test build on debomatic [1].

The good news is that it indeed fixes the build, but a few tests do 
timeout, including the one you mentioned:

The following tests FAILED:
	 60 - Trainers_LinearSvmTrainer (Timeout)
	139 - ObjFunct_NegativeLogLikelihood (Timeout)
	140 - ObjFunct_SvmLogisticInterpretation (Timeout)

I'll mark them as slow and push the update. Hopefully this should be 
enough to fix this FTBFS bug.

Thanks again for working on this.

[1] 
http://debomatic-mips.debian.net/distribution#unstable/shark/3.1.3+ds1-2/buildlog

Ghis

Message #56 received at [email protected] (full text, mbox, reply):

From: Ghislain Antony Vaillant <[email protected]>

To: [email protected]

Subject: Bug#836599: fixed in shark 3.1.3+ds1-2

Date: Fri, 02 Dec 2016 15:11:57 +0000

Source: shark
Source-Version: 3.1.3+ds1-2

We believe that the bug you reported is fixed in the latest version of
shark, which is due to be installed in the Debian FTP archive.

A summary of the changes between this version and the previous one is
attached.

Thank you for reporting the bug, which will now be closed.  If you
have further comments please address them to [email protected],
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Ghislain Antony Vaillant <[email protected]> (supplier of updated shark package)

(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing [email protected])


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Format: 1.8
Date: Thu, 01 Dec 2016 14:44:45 +0000
Source: shark
Binary: libshark0 libshark-dev shark-doc
Architecture: source
Version: 3.1.3+ds1-2
Distribution: unstable
Urgency: medium
Maintainer: Debian Science Maintainers <[email protected]>
Changed-By: Ghislain Antony Vaillant <[email protected]>
Description:
 libshark-dev - development files for Shark
 libshark0  - Shark machine learning library
 shark-doc  - documentation for Shark
Closes: 836599
Changes:
 shark (3.1.3+ds1-2) unstable; urgency=medium
 .
   * Fix FTBFS on mips and more architectures.
     - New patch Fix-build-on-MIPS.patch: override optimization level for
       selected sources.
     - New patch Update-list-of-slow-tests.patch: mark selected tests as
       slow to prevent testing from reaching a timeout.
     Thanks to Dejan Latinovic for investigating (Closes: #836599)
   * Drop superfluous Testsuite field.
   * Update description of binary packages.
   * Upgrade packaging to debhelper 10.
Checksums-Sha1:
 a9795d2d271acd0392d0744ef49d0b13d5253161 2361 shark_3.1.3+ds1-2.dsc
 b1310fa6f31cefdb3e120895d4fdb460da4da8ba 10572 shark_3.1.3+ds1-2.debian.tar.xz
Checksums-Sha256:
 925f00c653655f23db0d0c763bc1488da60655710186a957c24342cebb4c51e2 2361 shark_3.1.3+ds1-2.dsc
 baef807f69f56f9f081f935890b4fb52d7e73ed74a7fa013b741da744c5b47ff 10572 shark_3.1.3+ds1-2.debian.tar.xz
Files:
 b119b4513bf7738a3c4a0c4be5855aeb 2361 science optional shark_3.1.3+ds1-2.dsc
 485cddc468098f4e8de3fec621e1aec0 10572 science optional shark_3.1.3+ds1-2.debian.tar.xz

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=6seN
-----END PGP SIGNATURE-----

Send a report that this bug log contains spam.