As I understand it, the JIT had a bug where it would clobber registers on 32 bit ISAs with 64 bit stores when the offset was computed and the data was a constant. I discovered the bug while running the k_nucleotide test on ARM, then we debugged it with @zherczeg , plus I added a regression test for it and verified that it indeed catches the bug on x86 and ARM.
As I understand it, the JIT had a bug where it would clobber registers on 32 bit ISAs with 64 bit stores when the offset was computed and the data was a constant. I discovered the bug while running the
k_nucleotide
test on ARM, then we debugged it with @zherczeg , plus I added a regression test for it and verified that it indeed catches the bug on x86 and ARM.