From: khan AT xraylith DOT wisc DOT edu (Mumit Khan) Subject: Re: serious C++ problem, and need for egcs release 13 Mar 1998 17:23:14 -0800 Message-ID: References: <199803140058 DOT QAA11086 AT skaro DOT cygnus DOT com> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII To: Geoffrey Noer Cc: cygwin32-developers AT cygnus DOT com On Fri, 13 Mar 1998, Geoffrey Noer wrote: > Yeah, I know. :-( I made the decision to go ahead with releasing b19 > on the basis that there would probably be a fixed egcs released for it > at some point relatively soon afterwards. Right decision IMO. > > Where is your snapshot area? ftp://ftp.xraylith.wisc.edu/pub/khan/egcs/cygwin32/snapshots/1.0.2/ for 1.0.2 pre-release + patches + pre-built binaries. I did extract your changes to b19 gcc (the pathname related code). The testuite results are there as well. > Probably best to overwrite the existing b19 C++-related stuff??? That > would be a prefix of /cygnus/b19 and exec-prefix of > /cygnus/b19/H-i386-cygwin32. Sounds good to me. I'll rebuild everything over the weekend and put that up for ftp after running the tests one more time. > > An area we're having trouble with is collect2 being broken. Benjamin > Kosnik (bkoz) is planning on looking into it but hasn't had time yet. > If you have patches, please let him know? > Jeff Law and I had decided to remove collect2 for the release branch and put it back in the mainline sources. The only problem I'm aware of is the (now fixed by cgf's patch) bug in freopen, and I did submit a workaround to egcs-bugs list. What trouble have you had with collect2? Given that the linker now does .linkonce editing and Win32 backend doesn't support DWARF2 unwinding based EH, the only reason for collect2 is -frepo support, and that's a dubious advantage. I would prefer to leave it out for performance reasons. Letting GCC and ld do the template instantiation and discarding duplicates produces better code than -frepo, much like on ELF. Mumit