Closed h-2 closed 2 years ago
Hi,
I have tried your example with the current master and it seems to work fine (in both compilation cases). Can you try again and let me know? (I'm sure we can solve this one real quick).
I can still reproduce my problem with the main branch (commit d933a102c4eb4cc4d01c2a116a13f9969c531873 ).
I have tried with gcc-9 and gcc-11. I haven't been able to get this to build with clang, but I only tried for a couple of minutes.
Is this maybe related to the problems people have been reporting in #16 ?
I tried with a clean VM (Ubuntu 20), g++ 9.0, and I couldn't still reproduce the problem. Would you mind sharing more details of your environment/setup so I can reproduce it?
Sorry, I must have done something very weird with my local git branch. After getting a fresh clone, it now seems to work.
Thanks a lot and sorry for causing extra work!
No worries. I must admit that I feel relieved that there is no weird bug no more :-) Thanks
I have integrated WFA2-lib into an app, and I keep getting difficult to track alignment failures. After some successful alignments,
wavefront_align()
will just return-1
. The exact point where this happens, depends on whether I build in debug or release mode and whether I turn on sanitizers or not. Release mode produces many hundred correct alignments, debug fails after a couple and with ASAN, it fails immediately.I suspect there is undefined behaviour somewhere in WFA2-lib.
Here is a minimal example:
When building with these options:
g++ wfabug.cpp -I ~/devel/WFA2-lib ~/devel/WFA2-lib/lib/libwfa.a
it prints:When building with the address sanitizer:
g++ -fsanitize=address wfabug.cpp -I ~/devel/WFA2-lib ~/devel/WFA2-lib/lib/libwfa.a
it prints: