Closed Augustin-Zidek closed 1 year ago
I've committed two fixes to HMMER and one to its underlying Easel library that fix this and other tsan failures. The fixes will appear in the next HMMER3 release.
Your fix also works but requires C11 compliance to get the
The problems are cosmetic only - I believe all results are still correct despite the races, and the fixes do not have any impact on HMMER results, only on tsan and valgrind/helgrind problem detection.
Thank you!
Thanks for fixing. Looking forward to the next release.
I ran Thread Sanitizer on our nhmmer test and it found this issue:
This is with hmmer v3.3 downloaded on 2020-04-07, the problematic line in nhmmer is the following:
This issue can be fixed by making
nseqs
atomic in theP7_PIPELINE
struct inhmmer.h
, i.e. changing:to
I am however unsure whether this is a proper fix, or whether it indicates there is a more serious underlying issue. Could you advise whether this fix is correct?