Joseph S. Myers wrote: > On Thu, 28 Apr 2005, Richard Guenther wrote: > > >>The patch was bootstrapped and tested on i686-pc-linux-gnu for >>the C language with the only remaining regression being c99-init-4.c >>(I didn't manage to find the place to fix). > > > You don't say how it regresses. What diagnostic is it generating, what > code is generating it (for diagnostics GCC can generate in more than one > place), what are the relevant trees or other variables that caused the > diagnostic to be reached and what were they without the patch to cause it > not to be reached?
Yeah, sorry. The failure is FAIL: gcc.dg/c99-init-4.c (test for excess errors) Excess errors: /net/alwazn/home/rguenth/src/gcc/cvs/gcc-4.1/gcc/testsuite/gcc.dg/c99-init-4.c:8: error: initializer element is not constant /net/alwazn/home/rguenth/src/gcc/cvs/gcc-4.1/gcc/testsuite/gcc.dg/c99-init-4.c:8: error: (near initialization for 'a[0]') I guess the checking code is maybe confused by seeing &a[0] instead of &a, like in the places I fixed elsewhere. But I wasn't able to follow with the debugger, and we don't have a tree-dump available at this state. I didn't compare with an unpatched run, also the point that fails (c-typeck.c:5755) seems to be reached or not reached with the same tree (though than can't be). Well, I got bored staring at gdb and didn't debug this further. I'll try again, if the change is considered a good idea. But note that I guess there will be fallout that is not excercised by the testsuite maybe. Richard.
