public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* Problem to build egcs-19981026 on i686
@ 1998-10-29  0:50 Philippe Elbaz-Vincent
  0 siblings, 0 replies; 2+ messages in thread
From: Philippe Elbaz-Vincent @ 1998-10-29  0:50 UTC (permalink / raw)
  To: egcs

Dear all,

During the build of the last snapshot I got the following error:

/tmp/egcs-19981026/gcc/xgcc -B/tmp/egcs-19981026/gcc/
-B/usr/local/egcs/i686-pc-
linux-gnulibc1/bin/ -c -g -O2 -fno-implicit-templates -I. -I.
-nostdinc++ -D_G_H
AVE_MMAP fstream.cc
iostream.h:49: sorry, not implemented: `function_decl' not supported by
dump_typ
e
iostream.h: In method `ostream::ostream()':
In file included from fstream.h:30,
                 from fstream.cc:34:
iostream.h:49: uninitialized const `vtable for '
iostream.h:122: sorry, not implemented: `function_decl' not supported by
dump_ty
pe
iostream.h: In method `istream::istream()':
iostream.h:122: uninitialized const `vtable for '
iostream.h:229: sorry, not implemented: `function_decl' not supported by
dump_ty
pe
iostream.h: In method `istream::~istream()':
iostream.h:229: uninitialized const `vtable for '
iostream.h:229: sorry, not implemented: `function_decl' not supported by
dump_ty
pe
...
.....etc..... (I can send you the complete stuff if it's useful)
...
fstream.cc: In method `ofstream::~ofstream()':
fstream.cc:110: uninitialized const `vtable for '
fstream.cc:110: sorry, not implemented: `function_decl' not supported by
dump_ty
pe
fstream.cc: In method `ifstream::~ifstream()':
fstream.cc:110: uninitialized const `vtable for '
make[1]: *** [fstream.o] Error 1
make[1]: Leaving directory
`/tmp/egcs-19981026/i686-pc-linux-gnulibc1/libio'
make: *** [all-target-libio] Error 2

What goes wrong here ?

(NB: I've tried to look  in the archive of the different lists, but
without success).


Notice that usually I never have problems to build the snapshot (except
the 19981012 !), but I also had a  problem with the 19981019, the build
seems fine, but attempting to build PARI/GP (a program for number
theory), I get 
 /usr/local/egcs/bin/gcc -c -O3 -fexpensive-optimizations    -I.
-I../src/headers -o mp.o ../src/kernel/none/mp.c
../src/kernel/none/mp.c: In function `mulrr':
../src/kernel/none/mp.c:791: Internal compiler error,
output_operand_lossage `invalid expression as operand'

it's why I have tried the 19981026 ...(well I usually try all the new
snapshot of pgcc, and thus the one of egcs too ...)

here are the specs of my system:
cpu: p2/233
libc: 5.4.44
binutils: 2.8.1.xx
compiler used to build egcs: gcc 2.7.2.1
kernel: 2.0.33 (distrib: SuSE 5.2)



Thanks for your help (and your nice work),
all the best, Ph.

-- 
Ph. Elbaz-Vincent, Marie-Curie Fellow,
DFG-Forschergruppe "Arithmetic and Geometry", FB 6, Mathematik,
Universitaet GH Essen, D-45117 Essen | Email:elbaz-vincent@uni-essen.de 
Phone: +49 201 183 2198 | Fax: +49 201 183 2426

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Problem to build egcs-19981026 on i686
@ 1998-10-30  2:40 N8TM
  0 siblings, 0 replies; 2+ messages in thread
From: N8TM @ 1998-10-30  2:40 UTC (permalink / raw)
  To: pev; +Cc: egcs

In a message dated 10/29/98 9:17:16 AM Pacific Standard Time,
pev@milnor.mathematik.uni-essen.de writes:

> ostream.h:49: sorry, not implemented: `function_decl' not supported by
>  dump_typ
This prevents building the g++ libraries, no matter what your target.  We've
been assured it was fixed the next day.  You can go ahead and build and test
everything which doesn't depend on those libraries.

19981019 had some optimization bugs which could be worked around by lowering
the optimization level; -Os was the highest working level for Ix86 and hppa,
and there were a few expressions which broke at -Os, maybe even at -O.  For my
cases (obviously, I don't care about g++), this was fixed at 1026.

The only problem I've seen discussed with 19981012 was the spurious dependence
on libc-lock.h, which could be worked around by creating a 0 size file of that
name in libio/ or by altering the Makefile.in so it doesn't look for that on
non-libc systems.

To add to that, 1012 and 1019 seemed to give poor performance on i686 for
certain loops where even -Os leads to excessive spills, but they performed
relatively well on hppa and mips.  1026 seems to perform more like the best
recent releases for i686.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~1998-10-30  2:40 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1998-10-29  0:50 Problem to build egcs-19981026 on i686 Philippe Elbaz-Vincent
1998-10-30  2:40 N8TM

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).