After much investigation, I tend to agree with him. I'll see if I can find the point where things broke and start from there (I keep all alpha gcc versions that I compile for upload, so this should be easy :-P).
I'll also start flipping through the code again in case it's easy to spot (which it might be). C ---------- Forwarded message ---------- Date: Tue, 17 Jul 2001 19:00:49 -0700 From: H . J . Lu <[EMAIL PROTECTED]> To: Christopher C. Chimelis <[EMAIL PROTECTED]> Subject: Re: Nevermind the nevermind (sorry!) On Tue, Jul 17, 2001 at 12:32:35PM -0400, Christopher C. Chimelis wrote: > > On Mon, 16 Jul 2001, H . J . Lu wrote: > > > Please send me the testcase in .c and .s files. > > I'm sending you a tarball that has .c, .s, .i, and an unstripped.o (just > for reference). I noticed that if this code gets linked, it's fine and > will strip ok without a problem. It seems that if it's not linked (ar > archive cases), then it won't strip properly. > > The compile command used was: > gcc -g -c hello.c > I believe it is a gcc bug since the asm output doesn't look right Basically, those debug line lables should be their own section, not in .text. Gcc 3.0 has fixed it. H.J.