id	summary	reporter	owner	description	type	status	priority	milestone	component	version	resolution	keywords	cc
5184	Larger models of ScalableTestSuite fail on Hudson because of code generation timeout	Francesco Casella	Adrian Pop	"There are currently 4 large models of ScalableTestSuite

https://libraries.openmodelica.org/branches/newInst/ScalableTestSuite_noopt/files/ScalableTestSuite_noopt_ScalableTestSuite.Electrical.DistributionSystemAC.ScaledExperiments.DistributionSystemLinearIndividual_N_56_M_56.err
https://libraries.openmodelica.org/branches/newInst/ScalableTestSuite_noopt/files/ScalableTestSuite_noopt_ScalableTestSuite.Electrical.DistributionSystemAC.ScaledExperiments.DistributionSystemLinear_N_56_M_56.err
https://libraries.openmodelica.org/branches/newInst/ScalableTestSuite_noopt/files/ScalableTestSuite_noopt_ScalableTestSuite.Electrical.DistributionSystemDC.ScaledExperiments.DistributionSystemModelica_N_112_M_112.err
https://libraries.openmodelica.org/branches/newInst/ScalableTestSuite_noopt/files/ScalableTestSuite_noopt_ScalableTestSuite.Elementary.WhenEvents.ScaledExperiments.ManyEvents_N_8000_M_10.err
}}}
that fail with this error:
{{{
Failed to read output from testmodel.py, exit status != 0:
2018-10-24 16:38:03,733 - OMPython - INFO - OMC Server is up and running at file:////tmp/openmodelica.hudson.port.e9d7cbe3c6cd4f7ab47b5a4bec8f92ee pid=8924
2018-10-24 16:38:03,997 - OMPython - INFO - OMC Server is up and running at file:////tmp/openmodelica.hudson.port.53feb0cd9bb84d7584d5d5df086e3d3a pid=9007
}}}

The same models pass the building phase successfully when {{{--daeMode=new}}} is used (see [https://libraries.openmodelica.org/branches/newInst-daemode/ScalableTestSuite_noopt/ScalableTestSuite_noopt.html newInst-daemode report]).

Looking at the times of the various phases in the [https://libraries.openmodelica.org/branches/newInst/ScalableTestSuite_noopt/ScalableTestSuite_noopt.html newInst report] and how they scale, it seems to me that the reason of failure is that the backend or compile time exceeds 300 seconds in those cases and for some reason this problem is not reported correctly. 

The problem doesn't show up if {{{daeMode=new}}} is used because apparently the backend and compile times scale better, but on the other hand a) daeMode still has other issues and b) it is interesting to be able to compare the times of odeMode and daeMode also for larger models.

I would then suggest to increase the timeout of the code generation phases to 600 (or possibly even 1000) seconds. There are only four cases needing that extra time, so I guess this won't change the time to run the entire testsuite significantly."	defect	new	high		Testing Framework				Martin Sjölund