Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

Hi,

On Wed, Sep 09, 2015 at 11:22:10AM +0200, Julien Cristau wrote:
> On Wed, Sep  9, 2015 at 09:44:30 +0200, Andreas Tille wrote:
> 
> > Hi mips porters,
> > 
> > could anybody please find a more powerful machine to build the rdepends
> > libfreecontact-perl and python-freecontact.
> > 
> No, as I already said that is not acceptable.  Packages should be built
> on buildds.  Reliably.

I personally see two options to deal with this (as said before I do not
see any sense in simply skipping the test suite - why should we deliver
a package on an architecture where it can not even be tested?):

  1. Exclude mips explicitly in d/control
  2. Ask ftpmaster to remove the package for mips architecture

The latter would enable the package to migrate to testing and once there
might be more powerful autobuilders it will come back automatically.
Please let me know what you think.

Kind regards

  Andreas.

-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Aron Xu]

On Wed, Sep 9, 2015 at 3:44 PM, Andreas Tille  wrote:
> Hi mips porters,
>
> could anybody please find a more powerful machine to build the rdepends
> libfreecontact-perl and python-freecontact.
>
> Alternatively I would decide to exclude mips architecture for
> libfreecontact and its Perl and Python interfaces since I see no point
> in delivering a package that does not pass its build time check.
>

Would be nice if you can keep it as-is for a while, new machines are
being purchased with FPU (estimation is probably the end of year?),
that will solve the compilation stuck.

Thanks,
Aron

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

On Wed, Sep  9, 2015 at 09:44:30 +0200, Andreas Tille wrote:

> Hi mips porters,
> 
> could anybody please find a more powerful machine to build the rdepends
> libfreecontact-perl and python-freecontact.
> 
No, as I already said that is not acceptable.  Packages should be built
on buildds.  Reliably.

Cheers,
Julien


signature.asc
Description: Digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

Hi mips porters,

could anybody please find a more powerful machine to build the rdepends
libfreecontact-perl and python-freecontact.

Alternatively I would decide to exclude mips architecture for
libfreecontact and its Perl and Python interfaces since I see no point
in delivering a package that does not pass its build time check.

Thanks

   Andreas.

On Tue, Sep 08, 2015 at 10:26:58PM +0200, Julien Cristau wrote:
> Control: reopen -1
> 
> On Fri, Aug 28, 2015 at 17:42:10 +0200, Andreas Tille wrote:
> 
> > Hi Julien,
> > 
> > On Fri, Aug 28, 2015 at 05:20:55PM +0200, Julien Cristau wrote:
> > > On Wed, Aug 26, 2015 at 17:05:39 +0200, Andreas Tille wrote:
> > > 
> > > > Any chance to do a manual build on a more powerfull MIPS box which might
> > > > not run into this problem? 
> > > > 
> > > NAK.  We need packages to build reliably on our buildds.
> > 
> > The reliability of a build can be proven by passing the test suite.  So
> > in this case we need buildds that can reliably build the package.
> > Otherwise the package should probably be excluded from this architecture
> > at all.  BTW, the package is now available for mips since some kind soul
> > managed to build it.
> > 
> > Thanks to whomever the work did finally
> > 
> Nope, no sugar.  Now instead we have
> https://buildd.debian.org/status/fetch.php?pkg=libfreecontact-perl=mips=0.08-4=1441643884
> and
> https://buildd.debian.org/status/fetch.php?pkg=python-freecontact=mips=1.1-2=1441644922
> 
> That's not an improvement.
> 
> Cheers,
> Julien



-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

Control: reopen -1

On Fri, Aug 28, 2015 at 17:42:10 +0200, Andreas Tille wrote:

> Hi Julien,
> 
> On Fri, Aug 28, 2015 at 05:20:55PM +0200, Julien Cristau wrote:
> > On Wed, Aug 26, 2015 at 17:05:39 +0200, Andreas Tille wrote:
> > 
> > > Any chance to do a manual build on a more powerfull MIPS box which might
> > > not run into this problem? 
> > > 
> > NAK.  We need packages to build reliably on our buildds.
> 
> The reliability of a build can be proven by passing the test suite.  So
> in this case we need buildds that can reliably build the package.
> Otherwise the package should probably be excluded from this architecture
> at all.  BTW, the package is now available for mips since some kind soul
> managed to build it.
> 
> Thanks to whomever the work did finally
> 
Nope, no sugar.  Now instead we have
https://buildd.debian.org/status/fetch.php?pkg=libfreecontact-perl=mips=0.08-4=1441643884
and
https://buildd.debian.org/status/fetch.php?pkg=python-freecontact=mips=1.1-2=1441644922

That's not an improvement.

Cheers,
Julien


signature.asc
Description: Digital signature

Processed: Re: Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Debian Bug Tracking System]

Processing control commands:

> reopen -1
Bug #796731 {Done: Debian FTP Masters } 
[src:freecontact] freecontact: FTBFS on mips (test suite failure)
Bug reopened
Ignoring request to alter fixed versions of bug #796731 to the same values 
previously set

-- 
796731: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=796731
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

Hi Julien,

On Fri, Aug 28, 2015 at 05:20:55PM +0200, Julien Cristau wrote:
 On Wed, Aug 26, 2015 at 17:05:39 +0200, Andreas Tille wrote:
 
  Any chance to do a manual build on a more powerfull MIPS box which might
  not run into this problem? 
  
 NAK.  We need packages to build reliably on our buildds.

The reliability of a build can be proven by passing the test suite.  So
in this case we need buildds that can reliably build the package.
Otherwise the package should probably be excluded from this architecture
at all.  BTW, the package is now available for mips since some kind soul
managed to build it.

Thanks to whomever the work did finally

 Andreas.

-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

On Fri, Aug 28, 2015 at 17:42:10 +0200, Andreas Tille wrote:

 Hi Julien,
 
 On Fri, Aug 28, 2015 at 05:20:55PM +0200, Julien Cristau wrote:
  On Wed, Aug 26, 2015 at 17:05:39 +0200, Andreas Tille wrote:
  
   Any chance to do a manual build on a more powerfull MIPS box which might
   not run into this problem? 
   
  NAK.  We need packages to build reliably on our buildds.
 
 The reliability of a build can be proven by passing the test suite.  So
 in this case we need buildds that can reliably build the package.
 Otherwise the package should probably be excluded from this architecture
 at all.  BTW, the package is now available for mips since some kind soul
 managed to build it.
 
Having the package available because it's been built on a non-buildd is
strictly worse than not having the package available.

Cheers,
Julien


signature.asc
Description: Digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

On Wed, Aug 26, 2015 at 17:05:39 +0200, Andreas Tille wrote:

 Any chance to do a manual build on a more powerfull MIPS box which might
 not run into this problem? 
 
NAK.  We need packages to build reliably on our buildds.

Cheers,
Julien


signature.asc
Description: Digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Dejan Latinovic]

Hi,

this is obviously a consequence of mips buildd's limitations.
In past all three successfully builds had happened on ball.
I also was able to built freecontact successfully for mips on machine with FPU, 
localy.
On machine without FPU, if icme_timeout is significantly increased
tests executes successfully but need a lot of time
so the TERM signal breaks the build.

I was not able to find ball specification, so I am not sure if
it have FPU.
Are there any mips buildd's that have FPU?
If yes, maybe we could blacklist freecontact on other machines?

If no, should we disable tests for mips?
I could provide patches and test it if needed?

Regards,
Dejan

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Aníbal Monsalve Salazar]

On Fri, 2015-08-28 18:08:33 +, Dejan Latinovic wrote:
 Hi,

 this is obviously a consequence of mips buildd's limitations.
 In past all three successfully builds had happened on ball.
 I also was able to built freecontact successfully for mips on machine
 with FPU, locally.
 On machine without FPU, if icme_timeout is significantly increased
 tests executes successfully but need a lot of time so the TERM signal
 breaks the build.

 I was not able to find ball specification, so I am not sure if it have
 FPU.
 Are there any mips buildd's that have FPU?
 If yes, maybe we could blacklist freecontact on other machines?

 If no, should we disable tests for mips?
 I could provide patches and test it if needed?

 Regards,
 Dejan

Hello,

At Imagination (http://imgtec.com/mips), we are aware of this problem
and would like to donate MIPS bi-endian machines with FPUs to replace
the old MIPS hardware used by Debian as buildd servers. We have plans
to order a couple of the new machines for evaluation and testing, and
depending on that, to decide to order more for Debian.

Cheers,

Anibal
-- 
anibal.monsalvesala...@imgtec.com


signature.asc
Description: Digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Laszlo Kajan]

Hello Andreas!

On Wed, 26 Aug 2015 17:05:39 +0200 Andreas Tille ti...@debian.org wrote:
 Any chance to do a manual build on a more powerfull MIPS box which might
 not run into this problem? 

I think we should disable the test suite (override_dh_auto_test).

What do you think?

Best regards,
Laszlo



signature.asc
Description: OpenPGP digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

On Tue, Aug 25, 2015 at 11:09:27PM +0200, Julien Cristau wrote:
 On Mon, Aug 24, 2015 at 09:32:25 +0200, Andreas Tille wrote:
 
  I wonder whether this might be connected to some transition issue.  At
  least I have no idea how to track this down.  Its no upstream version
  and before I did the gcc transition of libfreecontact everything went
  fine.
 
 did you try to reproduce the failure on the porterbox and/or provide a
 useful log of the error (the current summary is not exactly useful,
 running make check with VERBOSE=1 would at least improve things)?

I tried on gabrielli.d.o but there is no pdebuild nor sbuild.

Could you please more verbose where I can find a changeroot for
building on a porter box?

Kind regards

  Andreas.

-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

https://dsa.debian.org/doc/schroot

On August 26, 2015 1:36:45 PM CEST, Andreas Tille ti...@debian.org wrote:
On Tue, Aug 25, 2015 at 11:09:27PM +0200, Julien Cristau wrote:
 On Mon, Aug 24, 2015 at 09:32:25 +0200, Andreas Tille wrote:
 
  I wonder whether this might be connected to some transition issue. 
At
  least I have no idea how to track this down.  Its no upstream
version
  and before I did the gcc transition of libfreecontact everything
went
  fine.
 
 did you try to reproduce the failure on the porterbox and/or provide
a
 useful log of the error (the current summary is not exactly useful,
 running make check with VERBOSE=1 would at least improve things)?

I tried on gabrielli.d.o but there is no pdebuild nor sbuild.

Could you please more verbose where I can find a changeroot for
building on a porter box?

Kind regards

  Andreas.

-- 
http://fam-tille.de

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Dejan Latinovic]

Hi,

I had tried to build freecontact locally on two different cavium boards.
On both board build failed with TERM signal error on testing.

It seems that this is the command that  causes TERM signal:
 /build/freecontact-hkgMtc/freecontact-1.0.21/src/.libs/lt-freecontact 
 --parprof=psicov --density=0 -f ./demo_1000.aln --threads 0 --debug


The results of running this command manually:


 time /build/freecontact-hkgMtc/freecontact-1.0.21/src/.libs/lt-freecontact 
 --parprof=psicov --density=0 -f ./demo_1000.aln --threads 0 --debug

 Effective parameters: --apply-gapth=0 --clustpc=62 --cov20=0 
 --estimate-ivcov=1 --gapth=0.9 --mincontsep=5 --pseudocnt=1 
 --pscount-weight=0 --rho=-1 --density=0 [shrink_lambda=0.1] 
 --output-format=pfrmat_rr
will use 2 OMP threads
 total weight (variation) of alignment = 561.595172234625
 
 seq weight loop for 1000 seqs took 0.126313 secs
 will use 2 OMP threads
 calculated column aa frequencies, gap cols = 4
 calculated pair frequency table in 9.83136 secs
 collected apc_mean[MI] = 0.165528914074767
 aa freq sum (cell) = 0.93945475299, pairfreq sum (cell) = 
 0.88196197513
 formed covariance matrix (126/126,4)
 cov matrix Cholesky... not positive definite, dmean dev: 0.0457782
 cov matrix Cholesky... not positive definite, dmean dev: 0.0412004
 cov matrix Cholesky... not positive definite, dmean dev: 0.0370804
 cov matrix Cholesky... not positive definite, dmean dev: 0.0333723
 cov matrix Cholesky... not positive definite, dmean dev: 0.0300351
 cov matrix Cholesky... not positive definite, dmean dev: 0.0270316
 cov matrix Cholesky... not positive definite, dmean dev: 0.0243284
 cov matrix Cholesky... not positive definite, dmean dev: 0.0218956
 cov matrix Cholesky... not positive definite, dmean dev: 0.019706
 cov matrix Cholesky... not positive definite, dmean dev: 0.0177354
 cov matrix Cholesky... not positive definite, dmean dev: 0.0159619
 cov matrix Cholesky... not positive definite, dmean dev: 0.0143657
 cov matrix Cholesky... not positive definite, dmean dev: 0.0129291
 cov matrix Cholesky...
 shrank covariance matrix in 15300.1 secs
 will try rho 0.00178064, exact solution
  iteration =   1Error: inverse covariance matrix estimation exceeded time 
 limit (1800 sec).
 
 real  290m35.185s
 user  37m40.530s
 sys   253m0.660s


Regards,
Dejan

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

Hi Dejan,

On Wed, Aug 26, 2015 at 02:44:50PM +, Dejan Latinovic wrote:
 I had tried to build freecontact locally on two different cavium boards.

Thanks for this.  (I'm running the build on gabrielli currently ... and yes,
it takes time).

 On both board build failed with TERM signal error on testing.
  cov matrix Cholesky... not positive definite, dmean dev: 0.019706
  cov matrix Cholesky... not positive definite, dmean dev: 0.0177354
  cov matrix Cholesky... not positive definite, dmean dev: 0.0159619
  cov matrix Cholesky... not positive definite, dmean dev: 0.0143657
  cov matrix Cholesky... not positive definite, dmean dev: 0.0129291
  cov matrix Cholesky...
  shrank covariance matrix in 15300.1 secs
  will try rho 0.00178064, exact solution
   iteration =   1Error: inverse covariance matrix estimation exceeded time 
  limit (1800 sec).

Any chance to do a manual build on a more powerfull MIPS box which might
not run into this problem? 

Kind regards

  Andreas.

-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

On Mon, Aug 24, 2015 at 09:32:25 +0200, Andreas Tille wrote:

 Hi Mips porters,
 
 I wonder whether this might be connected to some transition issue.  At
 least I have no idea how to track this down.  Its no upstream version
 and before I did the gcc transition of libfreecontact everything went
 fine.
 
Andreas,

did you try to reproduce the failure on the porterbox and/or provide a
useful log of the error (the current summary is not exactly useful,
running make check with VERBOSE=1 would at least improve things)?

Thanks,
Julien


signature.asc
Description: Digital signature

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Andreas Tille]

Hi Mips porters,

I wonder whether this might be connected to some transition issue.  At
least I have no idea how to track this down.  Its no upstream version
and before I did the gcc transition of libfreecontact everything went
fine.

Any help would be welcome

  Andreas.

On Sun, Aug 23, 2015 at 08:06:24PM +0200, Julien Cristau wrote:
 Source: freecontact
 Version: 1.0.21-4
 Severity: serious
 Justification: fails to build from source (but built successfully in the past)
 
 Hi,
 
 your package failed to build on the mips buildd, see the log at
 https://buildd.debian.org/status/fetch.php?pkg=freecontactarch=mipsver=1.0.21-4stamp=1440138640
 
 Cheers,
 Julien



 ___
 Debian-med-packaging mailing list
 debian-med-packag...@lists.alioth.debian.org
 http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-med-packaging


-- 
http://fam-tille.de

Bug#796731: freecontact: FTBFS on mips (test suite failure)

[Julien Cristau]

Source: freecontact
Version: 1.0.21-4
Severity: serious
Justification: fails to build from source (but built successfully in the past)

Hi,

your package failed to build on the mips buildd, see the log at
https://buildd.debian.org/status/fetch.php?pkg=freecontactarch=mipsver=1.0.21-4stamp=1440138640

Cheers,
Julien


signature.asc
Description: Digital signature