Erik> On Sun, 7 Jan 1996, Mikael Nykvist wrote:
>> Any possibility of libg++? (I'm using Bob's old ones, but when
>> trying to compile the new Amiga emu, I got redefines betw your
>> libc.a and his libstdc++.a)
Erik> This is pretty high priority for the next release, but I don't
Erik> think it'll be right out of the box so I've been avoiding
Erik> it. David, you hard is this?
*Compiling* libg++ is pretty much trivial (I thought I had sent you a
patch or at least directions on the one or two things that I had to
change, if not, drop a note and I'll find out what I had to do...).
The real problem is that libg++ 2.7.1 comes with a big disclaimer that
it is not compatible with libc-4.6.27 (which is what the Alpha libc is
currently based on). I don't know what the incompatibilities are, but
I did notice multiply defined symbols (due to libio being both in libc
and libg++). If that's the only "incompatibility" it might be quite
easy to fix. Probably it would be best to talk to the libg++
developers.
Talking of libc: is there now a concensus of moving to libg? It
sounds like a fine idea to me, but on the other hand I think it would
be a bad idea to do a move before there is a reasonable chance that
the x86 world will move to that library as well.
Oh, also a word of warning: the new GNU rx-1.0 and sed-3.0 are *not*
64 bit clean just yet. The regex matcher seems to be reimplemented
from scratch and as a result my patches that make rx-0.5/sed-2.0 64
bit clean never made it into the GNU source tree. The good news is
that the new packages come with pretty extensive test scripts that
should make debugging much easier. If somebody wants to work on this,
be sure to get in touch with the package maintainers first.
>> Any possiblity of a 2.7.1(2) gcc rpm? 'case it looks like every
>> kernel above .52 doesnt work with your 2.7.0 rpm anymore.
Erik> That'll happen soon. I was hoping to go right to 2.7.2 but
Erik> it's buggy on the Alpha :-(
I appended the minimal test program that shows the 2.7.2 bug that I
stumbled across when compiling a kernel (the compiler segfaults, if I
remember correctly). The problem is somehow related to "reserved"
being the sign bit in a 32 bit integer, but that's as far as I tracked
it down. The 2.7.1->2.7.2 patch is very small so it might be possible
to find the bug by just looking at the patch. I submitted the bug to
Richard Kenner but don't know if it has been fixed already. If
somebody has a patch or can fix it, I'd love to hear about it.
Unfortunately, I don't have time right now to look into the matter.
Enjoy,
--david
--
typedef struct {
unsigned count:31,
reserved:1;
} mem_map_t;
mem_map_t * mem_map;
main(long a0)
{
while (a0--)
while (a0--)
mem_map[0].reserved = 0;
}