https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117886
--- Comment #7 from David Malcolm <dmalcolm at gcc dot gnu.org> --- Created attachment 60257 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=60257&action=edit Cleaned up reproducer I cleaned up the reproducer enough to compile, and was able to use this to reproduce the crash in tree_type_to_jit_type on both aarch64 and x86_64, with a build of libgccjit prior to r15-7126-g27470f9a818538. On applying r15-7126-g27470f9a818538, the crash is fixed, so I believe the crash is fixed in trunk. Does trunk now work for you? (since r15-7126-g27470f9a818538) I tried adding a regression test for this (copying the attachment to gcc/testsuite/jit.dg), but the test fails to find -lasan when the compiled file runs. Looks like jit.exp needs to be smarter about populating LD_LIBRARY_PATH. A remaining issue here is that we're generating invalid .c reproducers. I have a work-in-progress fix for this.
