[gentoo-science] Sage: Random exit failures on amd64

[gentoo-science] Sage: Random exit failures on amd64

[Schwan, Christopher]

Hi,

I just backported the fixes from sage-core-4.4.1 in src_prepare() to sage-core-4.3.5 and noticed now the same failures on amd64 as reported by Steven with sage-4.4 - so I have done something wrong in src_prepare(). I dont see the reason right now and will investigate on that later - maybe some of you see it.

Christopher

PS: Commited from PC with wrong time - dont be confused when looking at the history. I hope this doesnt break git

Re: [gentoo-science] Sage: Random exit failures on amd64

[Steven Trogdon]

[-- Attachment #1: Type: text/plain, Size: 1414 bytes --]

On   05/13/10 - 15:42:56, Schwan, Christopher wrote:
> Hi,
> 
> I just backported the fixes from sage-core-4.4.1 in src_prepare() to  
> sage-core-4.3.5 and noticed now the same failures on amd64 as  
> reported by Steven with sage-4.4 - so I have done something wrong in  
> src_prepare(). I dont see the reason right now and will investigate  
> on that later - maybe some of you see it.
> 
> Christopher
> 
> PS: Commited from PC with wrong time - dont be confused when looking  
> at the history. I hope this doesnt break git
> 
Christopher,

I backported to sage-core-4.3.5 here and I get none of the random exit  
failures. I didn't run the entire test suite but enough of it where  
random exit failures should appear. If I include the 4.3.5 networkx  
patch I get the same first four failures I had previously with 4.3.5  
and no random exit failures to that point. If I backport to 4.4.alpha0  
then random exit failures do appear. I didn't backport everything, only  
sage-core. I did a diff on the c_lib directory after unpacking  
backports for 4.3.5 and 4.4.alpha0. The only differences appear in  
c_lib/src/interrupt.c and are associated with message printouts when  
there are errors. So I believe that libcsage.so and its headers are the  
same for 4.4.alpha0 as for 4.3.5.

I just realized, and it's too late to try now, that I didn't check the  
notebook.

Steve

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-science] Sage: Random exit failures on amd64

[Christopher Schwan]

Hi Steven,

maybe I cheered too soon - could you please try to install sage-core-4.3.5-r1? 
That is the version which failed for me if I did not do something wrong. I 
also noticed that it took longer to trigger SIGABRT with this version.

In every case sage-4.3.5-r1 should be ready to get hg bisect working to find 
out which commit could have caused this issue.

Christopher

Re: [gentoo-science] Sage: Random exit failures on amd64

[Steven Trogdon]

[-- Attachment #1: Type: text/plain, Size: 721 bytes --]

On 05/14/10 01:55:21, Christopher Schwan wrote:
> Hi Steven,
> 
> maybe I cheered too soon - could you please try to install  
> sage-core-4.3.5-r1?
> That is the version which failed for me if I did not do something wrong. I
> also noticed that it took longer to trigger SIGABRT with this version.
> 
> In every case sage-4.3.5-r1 should be ready to get hg bisect working to find
> out which commit could have caused this issue.
> 
> Christopher
> 
Christopher,

It was sage-core-4.3.5-r1 that I was using. I ran the tests overnight on a  
laptop I have and there are no random exit failures. I don't see a newer  
commit than yesterday. I will run the complete test on another machine today.

Steve

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-science] Sage: Random exit failures on amd64

[Steven Trogdon]

[-- Attachment #1: Type: text/plain, Size: 677 bytes --]

On   05/14/10 - 01:55:21, Christopher Schwan wrote:
> Hi Steven,
> 
> maybe I cheered too soon - could you please try to install  
> sage-core-4.3.5-r1?
> That is the version which failed for me if I did not do something  
> wrong. I
> also noticed that it took longer to trigger SIGABRT with this version.
> 

OK, the test suite has completed on a second machine using  
sage-core-4.3.5-r1 backported to 4.3.5. I get _no_ SIGABRT failures  
here. I did notice that the order in which the tests are run depends on  
the machine. I'm not sure how the ordering is decided but, perhaps, a  
particular order is more inclined to produce SIGABRT failures?

Steve


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-science] Sage: Random exit failures on amd64

[Christopher Schwan]

Hi Steven,

On Friday 14 May 2010 19:44:06 Steven Trogdon wrote:
> On   05/14/10 - 01:55:21, Christopher Schwan wrote:
> > Hi Steven,
> > 
> > maybe I cheered too soon - could you please try to install
> > sage-core-4.3.5-r1?
> > That is the version which failed for me if I did not do something
> > wrong. I
> > also noticed that it took longer to trigger SIGABRT with this version.
> 
> OK, the test suite has completed on a second machine using
> sage-core-4.3.5-r1 backported to 4.3.5. I get _no_ SIGABRT failures
> here. I did notice that the order in which the tests are run depends on
> the machine. I'm not sure how the ordering is decided but, perhaps, a
> particular order is more inclined to produce SIGABRT failures?
> 
> Steve

On my amd64 it is still the same situation: I reinstalled sage-4.3.5-r1 and I 
still get SIGABRT failures. Even worse, downgrading to sage-4.3.5 showed the 
same failures, so I think the problem (additionally ?) lies somewhere in 
Sage's dependencies ...

I am preparing a script which outputs a list of Sage's dependencies and their 
versions so we can compare it.

Christopher

Re: [gentoo-science] Sage: Random exit failures on amd64

[Christopher Schwan]

Hi,

On Saturday 15 May 2010 12:34:30 Christopher Schwan wrote:
> [..]
>
> I am preparing a script which outputs a list of Sage's dependencies and
> their versions so we can compare it.
> 
> Christopher

please checkout the recent overlay and run:

./list-compare-deps sage sage-clib sage-core sage-data sage-extcode sage-
notebook sage-singular

in the Tools subdirectory of our repository. This should list the dependencies 
of Sage compared with the version that are installed on your pc.

Christopher

Re: [gentoo-science] Sage: Random exit failures on amd64

[Steven Trogdon]

[-- Attachment #1: Type: text/plain, Size: 1010 bytes --]

On 05/15/10 07:37:53, Christopher Schwan wrote:
> Hi,
> 
> On Saturday 15 May 2010 12:34:30 Christopher Schwan wrote:
>> [..]
>> 
>> I am preparing a script which outputs a list of Sage's dependencies and
>> their versions so we can compare it.
>> 
>> Christopher
> 
> please checkout the recent overlay and run:
> 
> ./list-compare-deps sage sage-clib sage-core sage-data sage-extcode sage-
> notebook sage-singular
> 
> in the Tools subdirectory of our repository. This should list the  
> dependencies
> of Sage compared with the version that are installed on your pc.
> 
> Christopher
> 
Christopher,

I hadn't masked things properly in backporting when using sage-core-4.3.5-r1  
on my laptop. After fixing things and running the tests again I now have one  
SIGABRT failure. I'm not sure if the fixes resulted in the failure or  
something else triggered it. The only library change was ghmm. I've been  
unable to get the test to fail when running it individually.

Steve

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]