MantisBT - Rosetta
View Issue Details
0000094Rosetta[All Projects] Crashpublic2012-09-05 09:252012-09-06 08:15
smlewis 
smlewis 
normalmajorsometimes
newopen 
Trunk 
 
oop_create, oop_dock_design
oop_create, oop_dock_design -s any_old_pdb.pdb
Confirmed As Bug
0000094: oop_* integration tests did run on server, but not on some local machines - code suggest server should have crashed them
As of SVN 50960, the oop_create and oop_dock_design integration tests do not run, but only on some copies of Rosetta. If they fail, they fail immediately; integration.py reports its usual

> *** Test oop_dock_design did not run on host local_host! Check your --mode flag and paths.

If you run the test manually, you see

ERROR: PDBJobInputter is not known to the JobInputterFactory. Was it
registered via a JobInputterRegistrator in one of the init.cc files
(devel/init.cc or protocols/init.cc)?
ERROR:: Exit from: src/protocols/jd2/JobInputterFactory.cc line: 81

This indicates a failure in factory load time registration. Indeed,
the OOP executables call core::init instead of devel::init as they are
supposed to. protocols::init is bypassed, so the job distributor is
never set up, and it promptly crashes.

Fixing core::init->devel::init fixes the bug (SVN 50961), but that doesn't explain why it wasn't failing on the server. This bug ticket is for figuring out why the server, etc. allowed the test to run when it was pretty clearly (by reading the code) destined to crash.
integration tests, oop, oops, server
Issue History
2012-09-05 09:25smlewisNew Issue
2012-09-05 09:25smlewisStatusnew => assigned
2012-09-05 09:25smlewisAssigned To => smlewis
2012-09-05 09:25smlewisAssigned Tosmlewis =>
2012-09-05 09:25smlewisTag Attached: integration tests
2012-09-05 09:25smlewisTag Attached: oop
2012-09-05 09:25smlewisTag Attached: oops
2012-09-05 09:25smlewisTag Attached: server
2012-09-06 08:15smlewisAssigned To => smlewis
2012-09-06 08:15smlewisStatusassigned => new

There are no notes attached to this issue.