On Sat, 15 Oct 2011 17:09:32 +0200, Marc Espie wrote: > On Sat, Oct 15, 2011 at 03:37:16PM +0200, Pascal Stumpf wrote: > > The -Wbounded warning is caused by somewhat strange and conflicting > > definitions for FILNMLEN/E_FILNMLEN in coff/*.h. Unless anyone cares > > about long NT filenames, I think we should just remove support for those > > until the gdb/bfd people sort this out and make the defines consistent. > > It probably wouldn't work anyway due to memcpy overflows. > > > > Unfortunately, loading core files still doesn't work (at least on > > amd64), but iirc that has never been the case with the new gdb. > > According to kettenis, gdb from cvs current compiles like a charm. > Can you try to filch the corresponding change instead ? > I asked Mark, but he doesn't appear to have the time...
Okay, this is _extremely_ strange: gdb from CVS obviously uses libbfd from the same repository, so I went there looking for peigen.c. And guess what: It has been removed *10 years ago*. That's right, gdb still includes and compiles that file in its own version of libbfd (claiming to be from mid 2010 according to the change log). Looks like they just forgot to remove it from the repo they create their tarballs from. Or maybe they did this on purpose, I have no idea. Anyway: Ouch. (Funny thing is that bfd/ChangeLog-0001 even has an entry for peigen.c as being deleted.) http://sourceware.org/cgi-bin/cvsweb.cgi/src/bfd/Attic/peigen.c?cvsroot=src&hideattic=0
