------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca 2005-04-11 02:41 ------- Subject: Re: [4.1 Regression] ld segmentation fault linking libgfortran.sl.0.0
> I really doubt that this is a libfortran problem but a middle-end one or even > a target one. I only said libgfortran as that's what broke. > It might also just be a bug in HP's ld. Yes. I believe that it has been reported by Codesourcery. We had the same problem with Joseph's original patch to only emit .import directives for symbols that were actually referenced. If you incorrectly type a function symbol, the linker dies. The problem is finding the symbol. The last successful build was updated on Fri Apr 8 02:22:26 UTC 2005. I would guess that it's a varasm.c change that did it. Dave -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20939
