Anton Korobeynikov
Anton Korobeynikov
Does the issue reproduce with something less ancient? E.g. LLVM 18?
> In general, passes should preserve annotations on blocks, which inhibits some cases where blocks could be combined. Still, this looks a bit fragile. But I guess there is no...
Well, # of contigs frankly speaking does not mean anything. What is much more important is that the length of assembly increased drastically. You 50x assembly is twice as large...
As far as I can see from the log 30x case is definitely a coincidence. Because of the elevated levels of heterozygosity (and low coverage) SPAdes is having hard times...
@vlstill Looks like generic-struct.p4 is failing
Wow, very gross hack! :)
> but I don't see other clear way to create the instance on demand. We'd probably go with some value-based semantics here. But it seems pretty heavyweight for a task....
This: > At first, controls inlining renames a() to > @name("._xyz") action __xyz() Looks potentially problematic. Do we understand the logic behind the renaming rules? Looks like it might create...
Maybe the logic should be changed there? In order not to clash with internal identifiers? Tagging @ChrisDodd @vlstill @fruffy
I think at least, if the DEBUG build is expensive in terms of test runtime, we should have Release configuration with assertions enabled. Actually, I think all CI jobs must...