You are here

Error with the mpi version of simple_cycpep_predict

6 posts / 0 new
Last post
Error with the mpi version of simple_cycpep_predict
#1

Dear users and developers,

 

I recently compiled rosetta with MPI enabled and tried to use the simple_cycpep_predict application. It produced the following error:

ERROR: Assertion `seqpos != 0` failed.
ERROR:: Exit from: src/core/pose/variant_util.cc line: 167
protocols.cyclic_peptide_predict.SimpleCycpepPredictApplication_MPI: (2) {0} [ ERROR ] Exception in SimpleCycpepPredictApplication caught:
protocols.cyclic_peptide_predict.SimpleCycpepPredictApplication_MPI: (2) {0} [ ERROR ] 

File: src/core/pose/variant_util.cc:167
[ ERROR ] UtilityExitException
ERROR: Assertion `seqpos != 0` failed.

 

At first, it looked like an error in the input file but the input files seem valid as the non-mpi compiled version of the application works just fine...

Any hint on what's wrong here (I'm using Rosetta 2019.35, but there is no change to this part of the code in the newest versions, according to the commit logs)?

 

Best,

 

Florent

Post Situation: 
Wed, 2020-02-05 05:16
Florent Langenfeld

Can you post the full backtrace, or contents of the ROSETTA_CRASH.log file?  Without that, it's hard to know what's going wrong.

Also, try it with the latest Rosetta.  Whether or not the variant_util.cc file has changed, the code that called that function could have.  I do fix a lot of bugs over time, and this one might already have been fixed.

Wed, 2020-02-05 22:36
vmulligan

The job didn't produced any ROSETTA_CRASH.log file. Attached is the output of the command, generated with the latest (2020.03.61102) Rosetta. I uploaded my input files as well.

I'm recompiling rosetta with different `extra` flags to see whether it impacts the function behavior or not.

Thanks for your help

File attachments: 
Thu, 2020-02-06 06:21
Florent Langenfeld

Reading through the code, it looks like you might get that error if you are:

  • cyclizing through a sidechain-sidechain isopeptide bond, AND
  • not specifying isopeptide bond indices

 

If specifying the indices with -cyclic_peptide:sidechain_isopeptide_indices <int> <int> solves the problem, let me know.  I should add a better, more informative error message to catch that.

Wed, 2020-02-05 22:49
vmulligan

The error occurs for a `terminal_disulfide` cyclization. No isopeptide around.

Thu, 2020-02-06 05:46
Florent Langenfeld

>The error occurs for a `terminal_disulfide` cyclization. No isopeptide around.

Ah!  That's a clue.

Could you:
- Compile in debug mode (./scons.py -j8 mode=debug extras=mpi,serialization simple_cycpep_predict) and rerun?  That should spit out a backtrace when it crashes.
- Send me a reproducible test case at vmulligan@flatironinstitute.org?  Then I can fix the bug and be sure that it is fixed.

Mon, 2020-02-10 11:07
vmulligan