On Sun, Dec 28, 2025 at 11:09:04AM +0100, Rainer Orth wrote: > Several recent tests that use check-function-bodies on x86 FAIL on > Solaris: they all lack dg-add-options check_function_bodies which is > required to handle some Solaris differences. One test also needs > -fomit-frame-pointer to deal with a different Solaris/x86 default.
Hi Rainer, Sorry for breaking the test for Solaris. I was not aware of check-function-bodies. I appreciate CC'ing me. > > Tested on i386-pc-solaris2.11 and x86_64-pc-linux-gnu. > > Committed to trunk. > > > I'm increasingly disconcerted with HJ's constant disregard for the > requirements of non-Linux targets. This issue has been known for months > and the fix is well-known, but every new set of check-function-bodies > tests had the same issue. I've lost count of how many times I had to > fix them up subsequently. Even after I introduced dg-add-options > check_function_bodies to make this as easy as possible, it was again > ignored. > > Darwin is way worse off than Solaris here: currently all > check-function-bodies tests FAIL on Darwin/x86. For one, they need to > use -fdwarf2-cfi-asm to enforce the generation of the cfi directives. > However, I wonder if those have any bearing on what the tests check for > or are just present because they are generated by default on Linux. > Apart from that (easily handled with the dg-add-options above), there > are more differences: > > * Some are merely syntactical, like the use of L<N> for labels instead > of .L<N>. Those can/could be handled in scanasm.exp, but there are > several more, e.g. lC<N> instead of .LC<N> , so I gave up on an > initial attempt to fix this. > > * Unfortunately there are also some codegen differences that Iain is > currently trying to handle in some way. Labels and CFI statements can safely be removed/ignored for pr122675-1.c. For RISC-V this was done with r16-5941-g951f3ccefca097. Regards, Dimitar > > If this shouldn't work out, Darwin is faced with ca. 150 testsuite > failures due to this single issue. While one could xfail the dg-final > on darwin (which would likely be forgotten just like the Solaris issue > above), it would be better to introduce a new effective-target keyword > to handle the problem if more platforms with similar issues come up. > > Rainer
