pdbsplit stalls with ncore > 1
Issue #397
resolved
In the example from the eNMA vignette (part II) pdbsplit hangs after reaching 100% with ncore > 1. Aborting with Ctrl-C and then re-running gives following warnings:
Warning: closing unused connection 4 (/tmp/RtmpbJWOy2/file531e30db9119)
Warning: closing unused connection 5 (/tmp/RtmpbJWOy2/file531e2bb649c4)
Warning: closing unused connection 4 (/tmp/RtmpbJWOy2/file531e30db9119)
Warning: closing unused connection 5 (/tmp/RtmpbJWOy2/file531e2bb649c4)
Have you seen this before? Probably linked to the new progress bar? I can not reproduce it on my office computer. hmm..
Comments (3)
-
-
Found the reason of hanging up. It is due to the decimal increase of the progress bar when #chains >1 in some pdb files. The thread checking the progress bar stops only when progress >=max. The numerical round up may lead to progress < max even if all chains are processed.
The problem is fixed by this commit. I just add one more "update" of the progress bar (no any effect on output) to account for the possible round up problem.
Let me know if it works for you!
-
- changed status to resolved
- Log in to comment
Yes, Ctrl-C will break some TEMP file connection unexpectedly and so the warning message. One solution is to catch SIGNALS and close files properly before exit. Will have a look of this shortly.
Btw, I've tried the vignette and see the hangs of progress bar at 100% also. Not sure why this happens...