On Sun, 17 May 2009 19:52:50 +1000, Daniel Keep <daniel.keep.li...@gmail.com> wrote:
> > >Don wrote: >> Do you think it might be bug 1439? (note that the test case probably >> doesn't work any more, since Walter started hashing ultra-long mangled >> names, but the point remains that there's a (narrow) range of symbol >> lengths which will kill OPTLINK). Or yet another instance of bug 424? Or >> is likely to be something else? (IE, is your code template-heavy?) >> I can't make bug 2817 crash, and it's the only other plausible optlink >> segfault in bugzilla. > >I don't think it's 1439. The crash actually only shows up when I use -g >or -gc when I'm compiling. If I leave that off, it's fine. > >It's when I try to use tango.math.random.Random that it starts crashing. > I can use one method from that module, it's fine. If I try to use more >than one, it blows up. Note that it's quantity, not a specific symbol. > >Plus, if I compile the module with just a dummy main method, it's fine. > >It MIGHT be 424. I'm getting the crash at EIP=004244FB. I'm not >getting a fixups error message, if that's part of the symptoms, though. >My code isn't "template-heavy" (at least, not by my standards), but the >Random module does appear to be. > >I'm toying with a little OMF dumper program at the moment. I suppose >that's the easiest way to see if I'm hitting the fixup limit... > >Incidentally, if Walter or yourself would like the object files causing >the crash, I can provide those (probably via private email). I just >can't upload it generally since it's a private project. > > -- Daniel An object file was provided with this report http://d.puremagic.com/issues/show_bug.cgi?id=2436. But it was deemed too complex. This bug also shows up when debug-building QtD. FWIW, QtD uses lots of long identifiers but few templates.