Closed Morwenn closed 3 years ago
The previous commit wasn't a silver bullet, but it did help. There are no segfaults anymore, and we're down to two failing tests:
175 - test most sorters with no_post_iterator - cppsort::merge_insertion_sorter (Timeout)
759 - merge_insertion_sorter tests with projections (Timeout)
The timeouts make me suspect an infinite loop, but that's an investigation for another day. Unfortunately, the PRNG seed wasn't logged. Maybe we should log it directly from CMake in order to make extra sure.
It looks like the commit 11601542aca486b26b058d915ad1c7a0ead2a5c0 made the timeout issues disappear. I don't know how they're linked, but it doesn't seem extremely surprising that they are.
I'm gonna keep the issue opened a bit longer, and close it if the problem doesn't reappear.
merge_insertion_sort
often segfaults with MSVC for an obscure reason. I managed to track the issue down to that specific line prior to the pairwise swaps:The line itself doesn't cause the segfault, but the destruction of
end
at the end of the algorithm scope does. I am so far unsure why, more tests are needed.