> Thanks for this reduced test case, that's saving me a lot of work! You're welcome.
> Can you please try and see if the following C test case also fails? Yup, with the same compilation options, there are the same dreadful lines ld [%g0+0], %fn in the assembly file, which translates into (mem:SF (const_int 0 [0])) in the RTL dumps. And the executable does segfault at runtime. -- Eric Botcazou