Date: Sat, 24 Jan 1998 23:59:16 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: jfieber@indiana.edu (John Fieber) Cc: jdp@polstra.com, hackers@FreeBSD.ORG Subject: Re: egcs and exceptions Message-ID: <199801242359.QAA12224@usr04.primenet.com> In-Reply-To: <Pine.BSF.3.96.980124011506.9195D-100000@fallout.campusview.indiana.edu> from "John Fieber" at Jan 24, 98 01:19:41 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > > Do you also have a magic command line option to make some form of
> > > automatic template instantiation work? :)
> >
> > Have you tried "-frepo"? It is supposedly implemented in egcs, though
> > I haven't tried it.
>
> It sort of works, but some of the more complex STL classes don't
> end up with all their pieces. I have not completely eliminated
> the possibility of pilot error though...
Which STL implementation are you using?
I have just completed changes to the HP/SGI/Moscow version for
draft 4 and better pthreads support (draft 4 and later requires that
you not statically initialize nutexes anymore, a problem for the
ropeimpl.h, stl_alloc.h, stl_pthread_alloc.h, and stl_lock.h).
The correct way to do this would be to override new/delete, and do
the pthread_init() call to callback there. Unfortunately, being
a template, that would generate a garbage-scow load of these
functions -- one for each template class. I'm currently calling
the __THROW_BAD_ALLOC (if you don't have exceptions, it does a
"fprintf(stderr,...); exit(1);" instead, so it's portable).
I haven't sent them in because I'm not sure that everything is
happy. I'm trying to build the CMU acapd from the Cyrus project
sources, and I'm running into:
/var/tmp/cc011990.s: Assembler messages:
/var/tmp/cc011990.s:53679: Error: Local symbol L23795 never defined.
/var/tmp/cc011990.s:53679: Fatal error:1 error, 0 warnings, no object file generated.
... The .S file is:
L23970:
L23796:
leal -28(%ebp),%eax
pushl $2
pushl %eax
call __$_6String
addl $8,%esp
jmp L23968
.align 2,0x90
L23963:
nop
movl 4(%ebp),%edx
decl %edx
movl %edx,___eh_pc
movl $L23969,4(%ebp)
jmp L23968
.align 2,0x90
L23969:
call ___throw
movl $L23963,___eh_pc
call ___throw
L23968:
movl $L23795,___eh_pc
call ___throw
...I haven't been able to tell if this is a complier bug in FreeBSD
g++ 2.7.2.1, or something I introduced. 8-(.
Have you (or anyone else) ever seen this before? My suspicion is that
it's coming from a static member function for which g++ is failing
to generate a pseudo-symbol -- like when you do the following in gcc:
main()
{
char *p = "Hello World!\n" + 6;
printf( "p = '%s'\n", p);
printf( "p - 6 = '%s'\n", p - 6);
}
If you want to reproduce the acapd compilation, let me know, I have
a FreeBSD version of configure.in for autoconf, and a Makefile.in
that has been corrected for BSD's make.
Terry Lambert
terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199801242359.QAA12224>