On Mon, 13 May 2002, Marc Aurele La France wrote: >> New versions of the gcc compiler have an optimization called >> "string merging" which is enabled by default. The XFree86 module >> loader chokes on the ELF sections that this optimization adds to >> the ELF objects. > >> To work around this XFree86 module loader limitation, you need >> to pass -fno-merge-constants to the linker when modules are >> being built. This can be done from host.def with: > >> ModuleLdFlags -fno-merge-constants > >That should have a "#define " in front of it.
Right, I assumed one would be familiar with host.def et al. when adding it. >> This was automatically detected in later 4.1.0 CVS, and also in >> 4.2.0 CVS however it looks like someone removed the automatic >> checks in head CVS. > >No, it wasn't removed. But, it's possible that > > gcc -fmerge-constants -xc /dev/null -S -o /dev/null > >... no longer generates a zero return (i.e. is no longer a valid test to >determine whether or not the compiler has merge-constants support). Hmm, never thought of that. Could be failing I suppose. Just tested locally with: bash-2.05a$ gcc -fno-merge-constants -xc /dev/null -S -o /dev/null && echo yes yes bash-2.05a$ gcc3 -fno-merge-constants -xc /dev/null -S -o /dev/null && echo yes yes bash-2.05a$ gcc3 -fno-merge-constantsXXX -xc /dev/null -S -o /dev/null && echo yes cc1: Unrecognized option `-fno-merge-constantsXXX' bash-2.05a$ gcc -fno-merge-constantsXXX -xc /dev/null -S -o /dev/null && echo yes cc1: Unrecognized option `-fno-merge-constantsXXX' Can someone who is having this failure, run the same test and report what they get? -- Mike A. Harris Shipping/mailing address: OS Systems Engineer 190 Pittsburgh Ave., Sault Ste. Marie, XFree86 maintainer Ontario, Canada, P6C 5B3 Red Hat Inc. http://www.redhat.com ftp://people.redhat.com/mharris _______________________________________________ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
