https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118449
--- Comment #3 from Iain Buclaw <ibuclaw at gcc dot gnu.org> --- Speaking to the maintainer who's working on the new GC implementation for D. Initial suspicion is with the custom_gc.d test file itself, rather than changes within the runtime library. The custom_gc is just passing back a BlkInfo based on whatever malloc uses for alignment, and assumes the size requested is the size of the block. This never should have worked. I'm attempting to bisect this down to a specific commit, so at least we can confirm this.
