SIGSEGV (when using older version: munmap SIGABRT) about hh-suite HOT 6 CLOSED

Hi spock,

Thank you for your bug report. There is something wrong with the Uniclust. We are recomputing the broken part and will reupload it once that is done in about 1-2 days.

Best regards,

Milot

from hh-suite.

Thank you for a quick response.
Indeed, the same input file and binary against uniprot20_2016_02 work well.

I'm also getting quite a lot of random/unreproducible segfaults/sigaborts when hhblitsing a number of proteins one-by-one; by "unreproducible" I mean that re-trying is usually successful. Have not yet figured out why that is happening. Any advice on resolving/debugging this would also be appreciated.

from hh-suite.

Random crashes on the Uniclust? Please don't use that database right now. About 1/3 of the necessary calculations for a fixed release or done by now. You should have a working Uniclust database for hh-suite by tomorrow.

from hh-suite.

Could you try the new Uniclust30 and report back if you see any issues with it?

http://wwwuser.gwdg.de/~compbiol/uniclust/current_release/uniclust30_2016_09_hhsuite.tar.gz

Best regards,
Milot

from hh-suite.

Thank you for a quick response and DB rebuild 👍
Got the new database, running a test with an older binary: went past this input protein 245, will see tomorrow if there were any new problematic ones. Looks good so far :)

Question: for the recalculated uniclust30, do I also need to re-download the _consensus.fasta file? (I'm only using the 5 protein descriptions from there, not the actual consensus.)

Random crashes on the Uniclust?

Not only, also on uniprot20_2016_02.
However,

• restarting such failed searches usually finishes successfully;
• these random failures are of 3 types: mostly -11, rarely -6, and rarely also extra-long calculations which I just have to interrupt (multi-hour, multi-threaded searches which normally last less than a minute)
• I can't pinpoint what exactly is the problem with these random failures, as they are not reproducible.

from hh-suite.

The new uniclust worked well for me, the problem is apparently resolved.

from hh-suite.