[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] core dump in python bindings



SPELLCHECK got me again. That should be segfaults not seagulls.


Hi Mike,

Iâm not sure if you are referring  to a post by me about seagulls using htcondor python bindings. If so it seems that I had 2 problems (at least) that I have not been able to cleanly reproduce. My problems were in a custom conda environment. I donât believe OS packages were the problem


One problem was that the boost libraries had a mismatch between libboost-python and libboost the other was a constraint _expression_  bug in my code. 
It seems âOsUser==detcharâ is a No-No while OsUser==âdetcharââ works. I believe but canât prove that some combination of my bug and boost libraries can cause a SEGFAULT and others produce a timeout when trying to query history of a remote scheme.

By using my corrected python code and 

# Name                    Version                   Build  Channel
libboost                  1.86.0               h6c02f8c_3    conda-forge
libboost-python           1.86.0          py312hc39e661_3    conda-forge

Work fine for me

Best,
Joe


On Aug 25, 2025, at 08:22, Michael Thomas <wart@xxxxxxxxxxx> wrote:

I'm still trying to come up with a simple reproducible test case for this, but wanted to bring it up in case it's a known issue or sign that I'm doing something wrong.

I'm using the htcondor-24.10.2 python bindings to iterate over all user jobs and job slots to calculate the total requested memory for all jobs.  However, at some point the value returned by collector.query.eval('RequestMemory') seems to return an classad.ExprTree object, and when i try to do any debugging on that object (using repr() or __str__ to inspect it) I get a segfault.

This is code that I inherited from a now-defunct outside project, so my understanding of what it's doing may be slightly flawed.

Are there any known issues with printing out ExprTree objects?

I'll continue to work on a simplified test case, and try to capture a core dump if possible.

--Mike
_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe

The archives can be found at: https://www-auth.cs.wisc.edu/lists/htcondor-users/