[fix] Analysis shouldn't fail on non-existing directory #3943
Conversation
There was a problem hiding this comment.
When the compilation command database contains the build information of a file in a non-existing directory then the pre-analysis phase of CTU shouldn't fail.
Is that okay though? When can that naturally happen and not be a sign of a user error? I am not a fan on this patch.
bruntib
force-pushed
the
non_existing_dir_fail
branch
2 times, most recently
from
ae20e63 to
4850392
Compare
bruntib
force-pushed
the
non_existing_dir_fail
branch
from
4850392 to
97ceeb5
Compare
There was a problem hiding this comment.
Its not your fault, but I'm not thrilled about the reasons behind this patch. If someone runs all their configure and setup scripts inside CodeChecker log, and logs build commands for temporary files, thats a user error.
I understand that we need to bend the knee on this one. But I'd prefer to ease up on the hard error / exception escape (before proper logging is implemented) as little as possible.
When the compilation command database contains the build information of a file in a non-existing directory then the pre-analysis phase of CTU shouldn't fail.
bruntib
force-pushed
the
non_existing_dir_fail
branch
from
97ceeb5 to
1aeafb7
Compare
When the compilation command database contains the build information of a file in a non-existing directory then the pre-analysis phase of CTU shouldn't fail.