Jonathan Wakely writes: > I'm afraid I missed your reply today and went ahead and opened > http://gcc.gnu.org/PR40027 - and blamed the linker, feel free to > correct my ignorant guesses in that report :-) it turned out that your guesses were partially right: the Sun ld is indeed to blame in this particular configuration, although the bug doesn't trigger on i386-pc-solaris2.10. In recent Solaris Express versions (I've tried snv_112), bootstrapping for i686-pc-solaris2.11 works. Further investigation revealed that __i686.get_pc_thunk.bx isn't generated for i386 at all (I've still have to find out why this is the case), while for i686 it is generated in a .gnu.linkonce section: .section .gnu.linkonce.t.__i686.get_pc_thunk.bx,"ax",@progbits .globl __i686.get_pc_thunk.bx .hidden __i686.get_pc_thunk.bx .type __i686.get_pc_thunk.bx, @function __i686.get_pc_thunk.bx: movl (%esp), %ebx ret The problem is that Sun ld had a couple of bugs wrt to those sections, which were only recently fixed in Solaris Express in an effort to improve compatiblity between Sun ld and GNU ld. I'll have to check if they have been (or could be) backported to Solaris 10 and how to handle this situation in gcc/configure.ac. > No problem, I'm not in a hurry for an answer, I was just updating my > build scripts for 4.4, but wasn't planning to use it until 4.4.1 at > least. > > Thanks for looking into it, Thanks for your report. I've assigned it to myself and will work towards resolving it. Rainer ----------------------------------------------------------------------------- Rainer Orth, Faculty of Technology, Bielefeld University
