Re: bug report

["Martin v. Loewis" <martin@mira.isdn.cs.tu-berlin.de>]

> Is there a better egcs snapshot for us to try?

Just to give you an answer: I don't think so.

If you *really* want to get your bug fixed, please have a look
at http://egcs.cygnus.com/faq.html#support .

Regards,
Martin
>From bauhaus@uni-duisburg.de Mon Jul 19 14:17:00 1999
From: bauhaus@uni-duisburg.de (Georg Bauhaus)
To: egcs-bugs@egcs.cygnus.com
Subject: internal compiler error 109
Date: Mon, 19 Jul 1999 14:17:00 -0000
Message-id: <199907161633.SAA02909@sonnenregen.uni-duisburg.de>
X-SW-Source: 1999-07/msg00651.html
Content-length: 740

//  Hello,
//  again, including <string> "makes" c++ report an internal compiler
//  error 109.  The following code *is* erroneous, I think. Yet, ...

#include <string>
typedef basic_string<unsigned char> stringT;
class Fatal {
  static const stringT message = "Unspezifischer Fehler";  // not o.k.,
							   // not integral
};

/*
Linux sonnenregen 2.2.1 #3 SMP Fri Apr 23 06:04:04 CEST 1999 i586 unknown
Reading specs from /usr/lib/gcc-lib/i486-linux/egcs-2.91.66/specs
gcc version egcs-2.91.66 Debian GNU/Linux (egcs-1.1.2 release)

bug.hh:8: Internal compiler error 109.
bug.hh:8: Please submit a full bug report to `egcs-bugs@egcs.cygnus.com'.
bug.hh:8: See <URL: http://egcs.cygnus.com/faq.html#bugreport > for details.

-# georg
*/
>From dave@hiauly1.hia.nrc.ca Mon Jul 19 14:28:00 1999
From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: dave@hiauly1.hia.nrc.ca (John David Anglin)
Cc: law@cygnus.com, egcs-bugs@egcs.cygnus.com
Subject: Re: shared libstdc++ is broken on hppa1.1-hp-hpux9.01
Date: Mon, 19 Jul 1999 14:28:00 -0000
Message-id: <199907192128.RAA18533@hiauly1.hia.nrc.ca>
References: <no.id>
X-SW-Source: 1999-07/msg00652.html
Content-length: 1888

> 
> >   In message <199907171614.MAA09812@hiauly1.hia.nrc.ca>you write:
> >   > When gcc-2.95 is configured with --enable-shared, the number of unexpected
> >   > failures in check-g++ increases from 2 to 69:
> >   > 
> >   > 		=== g++ Summary ===
> >   > 
> >   > # of expected passes            5098
> >   > # of unexpected failures        69
> >   > # of unexpected successes       3
> >   > # of expected failures          85
> >   > # of untested testcases         11
> >   > /xxx/gnu/gcc-2.95/objdir/gcc/testsuite/../xgcc version 2.95 19990714 (prere
> >   > lease)
> >   > 
> >   > I have examined one of the above failures in some detail.  In
> >   > g++-robertl-eb88-C, the code goes gets lost when it tries to call
> >   > __eh_alloc.  In the shared library situation, the call is made through
> >   > the following stub:
> > This is a known problem and is discussed on the web pages -- EH with
> > shared libraries does not work on several platforms, including the PA.
> 
> Since EH works with static libraries, is there any chance that linking
> g++ programs with immediate binding would resolve this problem?

Doesn't work.

However, immdiate binding allowed debugging the problem and it becomes
clearer what the problem is.  The "longjmp" from __sjthrow occurs from the
data section in the second quadrant.  It branches to normal text space in the
first quadrant.  The branch is taken with a bv instruction which doesn't
update IASQ.  The problem may be the indirect_jump insn defined in
pa.md which can't handle interspace branches.  Could the problem be
fixed by computing the space register for the jump from the upper two
bits of the branch address and using a branch which supports interspace
transfers (e.g., be)?

-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)
>From phdm@macqel.be Mon Jul 19 15:00:00 1999
From: "Philippe De Muyter" <phdm@macqel.be>
To: gcc-bugs@egcs.cygnus.com
Subject: gcc-2.95 `switch' handling / bootstrap failure
Date: Mon, 19 Jul 1999 15:00:00 -0000
Message-id: <199907192200.AAA29506@mail.macqel.be>
X-SW-Source: 1999-07/msg00653.html
Content-length: 1107

Since years, and up to egcs-1.1.2 included, switches were compiled on
m68k-motorola-sysv as follows.

        mov.w 12(%pc,%d0.l*2),%d0       ;# 1102 movhi+1
        jmp 8(%pc,%d0.w)        ;# 1104 tablejump+2
        swbeg &72
L%L244:
        short L%L191-L%L244
	...

Note that there is no instruction between the table fetch and the jump,
and that the offsets are given as constants in the `mov' and `jmp' instructions.

Now with gcc-2.95, there are sometimes intervening instructions as follows :

        mov.w 12(%pc,%d0.l*2),%d0       ;# 1102 movhi+1
        mov.l %d3,%d5   ;# 1363 movsi+1/1			<<<<<<<
        addq.l &1,%d5   ;# 1288 *addsi3_internal/4		<<<<<<<
        jmp 8(%pc,%d0.w)        ;# 1104 tablejump+2
        even
        swbeg &72
L%L244:
        short L%L191-L%L244
	...

Of course, this prevents gcc-2.95 to bootstrap on m68k-motorola-sysv, because
the native assembler can't cope with the solutions involving the
table address label in the `mov' instruction.

Does someone know why this has changed and how this could be fixed (perhaps
with an appropriate target macro) ?

Philippe
>From martin@mira.isdn.cs.tu-berlin.de Mon Jul 19 15:11:00 1999
From: "Martin v. Loewis" <martin@mira.isdn.cs.tu-berlin.de>
To: bauhaus@uni-duisburg.de
Cc: egcs-bugs@egcs.cygnus.com
Subject: Re: internal compiler error 109
Date: Mon, 19 Jul 1999 15:11:00 -0000
Message-id: <199907192204.AAA00864@mira.isdn.cs.tu-berlin.de>
References: <199907161633.SAA02909@sonnenregen.uni-duisburg.de>
X-SW-Source: 1999-07/msg00654.html
Content-length: 395

> bug.hh:8: Internal compiler error 109.
> bug.hh:8: Please submit a full bug report to `egcs-bugs@egcs.cygnus.com'.
> bug.hh:8: See <URL: http://egcs.cygnus.com/faq.html#bugreport > for details.

Thanks for your bug report. gcc-2.95 19990701 writes

a.cc:8: in-class initialization of static data member of non-integral type `const stringT'

so it seems that the bug is fixed.

Regards,
Martin
>From doko@cs.tu-berlin.de Mon Jul 19 16:41:00 1999
From: Matthias Klose <doko@cs.tu-berlin.de>
To: gcc-bugs@gcc.gnu.org
Subject: gcc-testresults not archived
Date: Mon, 19 Jul 1999 16:41:00 -0000
Message-id: <14227.46683.323129.930858@bolero>
X-SW-Source: 1999-07/msg00655.html
Content-length: 90

gcc-bugs is archived again, the last message archived in
gcc-testresults is from 7/16/99.
>From cdunham@conwaycorp.net Mon Jul 19 16:48:00 1999
From: Curtis Dunham <cdunham@conwaycorp.net>
To: gcc-bugs@gcc.gnu.org
Subject: template code yields (almost) no asm output
Date: Mon, 19 Jul 1999 16:48:00 -0000
Message-id: <3793B9C5.FFD8F440@conwaycorp.net>
X-SW-Source: 1999-07/msg00656.html
Content-length: 988

potential bug in all recent egcs versions of g++ :

After a long time of trying to track down the problem in my code, I
found out that egcs (versions 1.1.1, 1.1.2, and the 7/16 and 7/18
pre-2.95's) was outputting almost no assembly for my template class
(linked list with fine-grained locking for thread safety) 


gcc --version : 2.95  (19990718)

i686-pc-linux-gnu system

g++ -g -march=pentiumpro -mpentiumpro -O6 ( -s, -E, -c etc )

(btw the same output occurs at all levels of optimization)

example error (several pages, an error for each instance of this
templated class in my code... btw this is during the link stage, as
***g++ outputs no errors during compilation***)

library_llist.o: In function `ExtractType(LList<SharedLibrary> *, char
*, int)':
/home/Zetta/War/War/new/library_llist.cpp:7: undefined reference to
`LList<SharedLibrary>::LList(void)'



Thanks so much, and keep up the terrific work on egcs.

Curtis Dunham
cdunham@conwaycorp.net
llist.ii.bz2
BZh91AY&SY{9w%
>From mrs@wrs.com Mon Jul 19 18:36:00 1999
From: mrs@wrs.com (Mike Stump)
To: egcs-bugs@egcs.cygnus.com, eric.freudenthal@nyu.edu
Subject: Re: bug report
Date: Mon, 19 Jul 1999 18:36:00 -0000
Message-id: <199907200136.SAA09274@kankakee.wrs.com>
X-SW-Source: 1999-07/msg00657.html
Content-length: 728

> From: "Eric Freudenthal" <eric.freudenthal@nyu.edu>
> Date: Mon, 19 Jul 1999 14:47:01 -0400

> I hate to be a pain, but I sent notes regarding a serious egcs bug
> several times over the last six weeks (see attached url) and have
> rec'd no response.

> Is there a better egcs snapshot for us to try?

You can always try the latest, see the web site for information on how
to find a spanshot.  If it works better for you, then you should be
unstuck.  If not, well, that's just a little wasted time.

We don't usually tell people that we don't yet have a solution for
them, and we sometimes miss it when we fix a person's problem (we fail
to tell them).  Because this is a volunteer effort, response time can
be unpredictable.
>From law@cygnus.com Mon Jul 19 19:41:00 1999
From: Jeffrey A Law <law@cygnus.com>
To: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
Cc: egcs-bugs@egcs.cygnus.com
Subject: Re: shared libstdc++ is broken on hppa1.1-hp-hpux9.01 
Date: Mon, 19 Jul 1999 19:41:00 -0000
Message-id: <1976.932438426@upchuck.cygnus.com>
References: <199907191542.LAA25834@hiauly1.hia.nrc.ca>
X-SW-Source: 1999-07/msg00658.html
Content-length: 444

  In message <199907191542.LAA25834@hiauly1.hia.nrc.ca>you write:
  > > This is a known problem and is discussed on the web pages -- EH with
  > > shared libraries does not work on several platforms, including the PA.
  > 
  > Since EH works with static libraries, is there any chance that linking
  > g++ programs with immediate binding would resolve this problem?
It might help some of the problems, but it will not solve all of them.


jeff
>From gthompsn@rap.ucar.EDU Mon Jul 19 21:42:00 1999
From: gthompsn@rap.ucar.EDU (Greg Thompson)
To: egcs-bugs@egcs.cygnus.com
Subject: FORTRAN bug?
Date: Mon, 19 Jul 1999 21:42:00 -0000
Message-id: <199907200441.WAA18420@binovc.rap.ucar.EDU>
X-SW-Source: 1999-07/msg00659.html
Content-length: 15994

Hi,

I'm an experienced FORTRAN programmer with 9 solid years of programming
on quite a few UNIX platforms (including IBM, Sun, SGI, HP, and CRAY).
I've been using this segment of code across these platforms for years.
Now, I'm trying to port to Linux (Dell PC running a Debian installation).
I'm getting a Segmentation Fault for what I think is valid code.

I've spent 4 straight days isolating the problem and have boiled it
down to what I think are the bare minimums.  Can you please view it and
see if you agree that I've found a compiler bug?  I've attached a "shar"
file after my .signature but I've never tried sending a shar in an e-mail
so I hope it works.  If not, I've also placed the file on our anonymous
ftp site -   ftp://ftp.rap.ucar.edu/incoming/fprog.shar

Thanks in advance for any help you might have.

-- 
+--------------------------------------------------------------+
| Greg Thompson       http://www.rap.ucar.edu/staff/gthompsn/  |
|                     Research Applications Program            |
| (303) 497-2805      National Center for Atmospheric Research |
|    (fax) -8401      P.O. Box 3000  Boulder, CO 80307-3000    |
+--------------------------------------------------------------+


#!/bin/sh
# This is a shell archive (produced by GNU sharutils 4.2).
# To extract the files from this archive, save it to some FILE, remove
# everything before the `!/bin/sh' line above, then type `sh FILE'.
#
# Made on 1999-07-20 04:27 GMT by <adds@cessna>.
# Source directory was `/home/adds/test'.
#
# Existing files will *not* be overwritten unless `-c' is specified.
#
# This shar contains:
# length mode       name
# ------ ---------- ------------------------------------------
#   3407 -rw-rw-r-- README
#    643 -rw-r--r-- cmem_utils.c
#   1943 -rw-r--r-- test.f
#   2042 -rw-r--r-- test_seg_fault.f
#
save_IFS="${IFS}"
IFS="${IFS}:"
gettext_dir=FAILED
locale_dir=FAILED
first_param="$1"
for dir in $PATH
do
  if test "$gettext_dir" = FAILED && test -f $dir/gettext \
     && ($dir/gettext --version >/dev/null 2>&1)
  then
    set `$dir/gettext --version 2>&1`
    if test "$3" = GNU
    then
      gettext_dir=$dir
    fi
  fi
  if test "$locale_dir" = FAILED && test -f $dir/shar \
     && ($dir/shar --print-text-domain-dir >/dev/null 2>&1)
  then
    locale_dir=`$dir/shar --print-text-domain-dir`
  fi
done
IFS="$save_IFS"
if test "$locale_dir" = FAILED || test "$gettext_dir" = FAILED
then
  echo=echo
else
  TEXTDOMAINDIR=$locale_dir
  export TEXTDOMAINDIR
  TEXTDOMAIN=sharutils
  export TEXTDOMAIN
  echo="$gettext_dir/gettext -s"
fi
touch -am 1231235999 $$.touch >/dev/null 2>&1
if test ! -f 1231235999 && test -f $$.touch; then
  shar_touch=touch
else
  shar_touch=:
  echo
  $echo 'WARNING: not restoring timestamps.  Consider getting and'
  $echo "installing GNU \`touch', distributed in GNU File Utilities..."
  echo
fi
rm -f 1231235999 $$.touch
#
if mkdir _sh02795; then
  $echo 'x -' 'creating lock directory'
else
  $echo 'failed to create lock directory'
  exit 1
fi
# ============= README ==============
if test -f 'README' && test "$first_param" != -c; then
  $echo 'x -' SKIPPING 'README' '(file already exists)'
else
  $echo 'x -' extracting 'README' '(binary)'
  sed 's/^X//' << 'SHAR_EOF' | uudecode &&
begin 600 README
M"E1H:7,@9FEL92!C;VYT86EN<R!S96-T:6]N<R!L86)E;&QE9"`B1FEL97,@
M9F]U;F0@:6X@<VAA<B(L(")'-S<@5F5R<VEO;B(L(")#;VUP:6QE9"!5<VEN
M9R(L"B)-86-H:6YE($1E<V-R:7!T:6]N(BP@(D1E<V-R:7!T:6]N(&]F(%!R
M;V)L96TB"@H*1FEL97,@9F]U;F0@:6X@<VAA<CH*/3T]/3T]/3T]/3T]/3T]
M/3T]/3T*("`@4D5!1$U%("`@("`@("`@("`M(%1H:7,@9FEL92X*("`@8VUE
M;5]U=&EL<RYC("`@("`M($,@<V]U<F-E(&-O9&4@8V]N=&%I;FEN9R!A('=R
M87!P97(@9G5N8W1I;VX@*&ER86QL;V,I"B`@("`@("`@("`@("`@("`@("`@
M("!F;W(@)VUA;&QO8R<@8V%L;&5D(&)Y($9/4E1204XN"B`@('1E<W0N9B`@
M("`@("`@("`@+2!&3U)44D%.('-O=7)C92!C;V1E('1H870@87!P96%R<R!T
M;R!W;W)K(&9I;F4N"B`@('1E<W1?<V5G7V9A=6QT+F8@+2!&3U)44D%.('-O
M=7)C92!C;V1E('1H870@8V%U<V5S(%-E9VUE;G1A=&EO;@H@("`@("`@("`@
M("`@("`@("`@("`@1F%U;'0@*&]N;'D@9&EF9F5R96YC92!B971W965N('1H
M:7,@9FEL92!A;F0*("`@("`@("`@("`@("`@("`@("`@(")T97-T+F8B(&ES
M('1H92!A9&1I=&EO;B!O9B!O<&5N:6YG+"!W<FET:6YG('1O"B`@("`@("`@
M("`@("`@("`@("`@("!A;F0@8VQO<VEN9R!A(&9I;&4@:6X@=&AE(&UA:6X@
M<')O9W)A;2X*"@I'-S<@5F5R<VEO;CH*/3T]/3T]/3T]/3T]"@IG-S<@=F5R
M<VEO;B!E9V-S+3(N.3$N-C`@1&5B:6%N(#(N,2`H96=C<RTQ+C$N,2!R96QE
M87-E*2`H9G)O;2!&4T8M9S<W('9E<G-I;VX@,"XU+C(T+3$Y.3@P.#`T*0I$
M<FEV:6YG.B!G-S<@+78@+6,@+7AF-S<M=F5R<VEO;B`O9&5V+VYU;&P@+7AN
M;VYE"E)E861I;F<@<W!E8W,@9G)O;2`O=7-R+VQI8B]G8V,M;&EB+VDT.#8M
M;&EN=7@O96=C<RTR+CDQ+C8P+W-P96-S"F=C8R!V97)S:6]N(&5G8W,M,BXY
M,2XV,"!$96)I86X@,BXQ("AE9V-S+3$N,2XQ(')E;&5A<V4I"B`O=7-R+VQI
M8B]G8V,M;&EB+VDT.#8M;&EN=7@O96=C<RTR+CDQ+C8P+V-P<"`M;&%N9RUC
M("UV("UU;F1E9B`M1%]?1TY50U]?/3(@+41?7T=.54-?34E.3U)?7STY,2`M
M1%]?14Q&7U\@+41?7W5N:7A?7R`M1%]?:3,X-E]?("U$7U]I,S@V7U\@+41?
M7VQI;G5X7U\@+41?7W5N:7@@+41?7VDS.#8@+41?7VQI;G5X("U!<WES=&5M
M*'!O<VEX*2`M1%],04Y'54%'15]&3U)44D%.("UT<F%D:71I;VYA;"`M07-Y
M<W1E;2AU;FEX*2`M06-P=2AI,S@V*2`M06UA8VAI;F4H:3,X-BD@+41I,S@V
M("U$7U]I,S@V("U$7U]I,S@V7U\@+41I-#@V("U$7U]I-#@V("U$7U]I-#@V
M7U\@+V1E=B]N=6QL("]D978O;G5L;`I'3E4@0U!0('9E<G-I;VX@96=C<RTR
M+CDQ+C8P($1E8FEA;B`R+C$@*&5G8W,M,2XQ+C$@<F5L96%S92D@*&DS.#8@
M3&EN=7@O14Q&*0HC:6YC;'5D92`B+BXN(B!S96%R8V@@<W1A<G1S(&AE<F4Z
M"B-I;F-L=61E(#PN+BX^('-E87)C:"!S=&%R=',@:&5R93H*("]U<W(O;&]C
M86PO:6YC;'5D90H@+W5S<B]L:6(O9V-C+6QI8B]I-#@V+6QI;G5X+V5G8W,M
M,BXY,2XV,"]I;F-L=61E"B`O=7-R+VEN8VQU9&4*16YD(&]F('-E87)C:"!L
M:7-T+@H@+W5S<B]L:6(O9V-C+6QI8B]I-#@V+6QI;G5X+V5G8W,M,BXY,2XV
M,"]F-S<Q("UF;G5L;"UV97)S:6]N("UQ=6EE="`M9'5M<&)A<V4@9S<W+79E
M<G-I;VXN9B`M=F5R<VEO;B`M9G9E<G-I;VX@+6\@+W1M<"]C8V)S;4)W2BYS
M("]D978O;G5L;`I'3E4@1C<W('9E<G-I;VX@96=C<RTR+CDQ+C8P($1E8FEA
M;B`R+C$@*&5G8W,M,2XQ+C$@<F5L96%S92D@*&DT.#8M;&EN=7@I(&-O;7!I
M;&5D(&)Y($=.52!#('9E<G-I;VX@96=C<RTR+CDQ+C8P($1E8FEA;B`R+C$@
M*&5G8W,M,2XQ+C$@<F5L96%S92DN"D=.52!&;W)T<F%N($9R;VYT($5N9"!V
M97)S:6]N(#`N-2XR-"TQ.3DX,#@P-`H@87,@+58@+5%Y("UO("]T;7`O8V-S
M45IM;6LN;R`O=&UP+V-C8G-M0G=*+G,*1TY5(&%S<V5M8FQE<B!V97)S:6]N
M(#(N.2XQ("AI-#@V+6QI;G5X*2P@=7-I;F<@0D9$('9E<G-I;VX@,BXY+C$N
M,"XQ.0H@;&0@+6T@96QF7VDS.#8@+61Y;F%M:6,M;&EN:V5R("]L:6(O;&0M
M;&EN=7@N<V\N,B`M;R`O=&UP+V-C>G)K<7%7("]T;7`O8V-S45IM;6LN;R`O
M=7-R+VQI8B]C<G0Q+F\@+W5S<B]L:6(O8W)T:2YO("]U<W(O;&EB+V=C8RUL
M:6(O:30X-BUL:6YU>"]E9V-S+3(N.3$N-C`O8W)T8F5G:6XN;R`M3"]U<W(O
M;&EB+V=C8RUL:6(O:30X-BUL:6YU>"]E9V-S+3(N.3$N-C`@+6QG,F,@+6QM
M("UL9V-C("UL8R`M;&=C8R`O=7-R+VQI8B]G8V,M;&EB+VDT.#8M;&EN=7@O
M96=C<RTR+CDQ+C8P+V-R=&5N9"YO("]U<W(O;&EB+V-R=&XN;PH@+W1M<"]C
M8WIR:W%Q5PI?7T<W-U],24)&-S=?5D524TE/3E]?.B`P+C4N,C0*0"@C*4Q)
M0D8W-R!615)324].(#$Y.3<P.3$Y"E]?1S<W7TQ)0DDW-U]615)324].7U\Z
M(#`N-2XR-"TQ.3DX,#DR,`I`*",I($Q)0DDW-R!615)324].('!J=RQD;6<M
M;6]D<R`Q.3DX,#8Q-PI?7T<W-U],24)5-S=?5D524TE/3E]?.B`P+C4N,C0*
M0"@C*2!,24)5-S<@5D524TE/3B`Q.3DX,#<P.0H*"D-O;7!I;&5D(%5S:6YG
M.@H]/3T]/3T]/3T]/3T]/3T*("`@(#X@9V-C("UG("UC("U$1E]53D1%4E-#
M3U)%(&-M96U?=71I;',N8PH@("`@/B!G-S<@+6<@+6,@=&5S="YF"B`@("`^
M(&<W-R`M9R`M;R!T97-T('1E<W0N;R!C;65M7W5T:6QS+F\*"@I-86-H:6YE
M($1E<V-R:7!T:6]N.@H]/3T]/3T]/3T]/3T]/3T]/3T]/0H@("`@1&5L;"!D
M=6%L+5!E;G1I=6T@24D@=VET:"`U,3(@34(@4D%-(')U;FYI;F<@3&EN=7@@
M*$1E8FEA;B`R+C$I"B`@("!)(&%S<W5M92!A('-T86YD87)D(&EN<W1A;&P@
M;V8@8V]M<&EL97)S('=A<R!D;VYE("T@22!D:60@;F]T(&1O(&ET+"!A(&-O
M+7=O<FME<BP*("`@('-Y<RUA9&UI;B!D:60N"@H*1&5S8W)I<'1I;VX@;V8@
M4')O8FQE;3H*/3T]/3T]/3T]/3T]/3T]/3T]/3T]/3T*("`@($-O;7!I;&EN
M9R!T:&4@8V]D92!F;W(@(G1E<W0N9B(@<')O9'5C97,@=&AE(&5X86-T(')E
M<W5L=',@22!E>'!E8W0@86YD"B`@("!I<R!T:&4@<V%M92!O;B!A;&P@8F]X
M97,@22!T97-T960@*$E"32P@2%`L(%-U;BP@86YD(%-'22!R=6YN:6YG(%5.
M25@I+@H@("`@2&]W979E<BP@8V]M<&EL:6YG('1H92!C;V1E(&9O<B`B=&5S
M=%]S96=?9F%U;'0N9B(@<')O9'5C97,@82!S96=M96YT871I;VX*("`@(&9A
M=6QT('5P;VX@<F5A8VAI;F<@=&AE(&9O<G1R86X@)V-L;W-E("AU;FET/3$Q
M*2<@<W1A=&5M96YT+B`@270@9&]E<PH@("`@<')O9'5C92!T:&4@<V%M92!R
M97-U;'1S(&IU<W0@<')I;W(@=&\@<V5G(&9A=6QT+6EN9R!O;B!T:&4@1DE2
M4U0@871T96UP=`H@("`@8G5T('1H96X@:6UM961I871E;'D@<V5G(&9A=6QT
@<R!A9G1E<B!T:&%T($9)4E-4(&%T=&5M<'0N"@H*"@IE
`
end
SHAR_EOF
  $shar_touch -am 072004261999 'README' &&
  chmod 0664 'README' ||
  $echo 'restore of' 'README' 'failed'
  if ( md5sum --help 2>&1 | grep 'sage: md5sum \[' ) >/dev/null 2>&1 \
  && ( md5sum --version 2>&1 | grep -v 'textutils 1.12' ) >/dev/null; then
    md5sum -c << SHAR_EOF >/dev/null 2>&1 \
    || $echo 'README:' 'MD5 check failed'
22148f1124b93cc31e6e42ac43fc0c32  README
SHAR_EOF
  else
    shar_count="`LC_ALL= LC_CTYPE= LANG= wc -c < 'README'`"
    test 3407 -eq "$shar_count" ||
    $echo 'README:' 'original size' '3407,' 'current size' "$shar_count!"
  fi
fi
# ============= cmem_utils.c ==============
if test -f 'cmem_utils.c' && test "$first_param" != -c; then
  $echo 'x -' SKIPPING 'cmem_utils.c' '(file already exists)'
else
  $echo 'x -' extracting 'cmem_utils.c' '(text)'
  sed 's/^X//' << 'SHAR_EOF' > 'cmem_utils.c' &&
#if defined(F_UNDERSCORE)
#define iralloc iralloc_
#define irfree irfree_
#endif
X
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <g2c.h>
X
/*---+----------------------------------------------------------------*/
int iralloc(int *memtot, int ia, int *ioff)
X
{
X   int iaddr;
X   iaddr = (int) malloc((*memtot)*4);
X   *ioff = (iaddr-ia)/4;
X      /*   printf("in ralloc- %i %i %i \n",*memtot,iaddr,*ioff); */
X   return(iaddr);
}
/*---+----------------------------------------------------------------*/
X
void irfree(int *iaddr)
{
X   free((void *) *iaddr);
}
/*---+----------------------------------------------------------------*/
SHAR_EOF
  $shar_touch -am 072003521999 'cmem_utils.c' &&
  chmod 0644 'cmem_utils.c' ||
  $echo 'restore of' 'cmem_utils.c' 'failed'
  if ( md5sum --help 2>&1 | grep 'sage: md5sum \[' ) >/dev/null 2>&1 \
  && ( md5sum --version 2>&1 | grep -v 'textutils 1.12' ) >/dev/null; then
    md5sum -c << SHAR_EOF >/dev/null 2>&1 \
    || $echo 'cmem_utils.c:' 'MD5 check failed'
4792773fb08728707a92b6dff1cbbc17  cmem_utils.c
SHAR_EOF
  else
    shar_count="`LC_ALL= LC_CTYPE= LANG= wc -c < 'cmem_utils.c'`"
    test 643 -eq "$shar_count" ||
    $echo 'cmem_utils.c:' 'original size' '643,' 'current size' "$shar_count!"
  fi
fi
# ============= test.f ==============
if test -f 'test.f' && test "$first_param" != -c; then
  $echo 'x -' SKIPPING 'test.f' '(file already exists)'
else
  $echo 'x -' extracting 'test.f' '(text)'
  sed 's/^X//' << 'SHAR_EOF' > 'test.f' &&
C
C23456789012345678901234567890123456789012345678901234567890123456789012
C----+-----------------------------------------------------------------+
C
X      program test
X
X      implicit none
X      integer nx, ny, nz, ia1, ia2, ia3
X      real a1(1), a2(1), a3(1)
X
X      nx = 151
X      ny = 113
X      nz = 37
X      call mem_alloc(nx*ny*nz, a1, ia1)
X      call mem_alloc(nx*ny*nz, a2, ia2)
X      call mem_alloc(nx*ny, a3, ia3)
X      print*, '  Allocated arrays with nx,ny,nz = ', nx,ny,nz
X      print*, '  Offsets: ', ia1, ia2, ia3
X      call assign_vals(a1(ia1), a2(ia2), a3(ia3), nx, ny, nz)
X      call print_vals(a1(ia1), a2(ia2), a3(ia3), nx, ny, nz)
X
X      end
C----+-----------------------------------------------------------------+
X      subroutine mem_alloc(npts, arr, iarr)
X      implicit none
X      integer npts, iarr, iaddr, iralloc
X      real arr(*)
X
X      iaddr = iralloc(npts, arr, iarr)
X
X      return
X      end
C----+-----------------------------------------------------------------+
X      subroutine assign_vals(a1, a2, a3, n1, n2, n3)
X      implicit none
X      integer n1, n2, n3, i, j, k
X      real a1(n1,n2,n3), a2(n1,n2,n3), a3(n1,n2)
X
X      do k = 1, n3
X         do j = 1, n2
X            do i = 1, n1
X               a1(i,j,k) = float(i)*0.6 + float(j)*50 - float(k)
X               a2(i,j,k) = float(i)/100. - float(j) + float(k)*10.
X            enddo
X         enddo
X      enddo
X      do j = 1, n2
X         do i = 1, n1
X            a3(i,j) = float(i) + (3.14159/180.0)*sin(float(j)/n2)
X         enddo
X      enddo
X
X      return
X      end
C----+-----------------------------------------------------------------+
X      subroutine print_vals(a1, a2, a3, n1, n2, n3)
X      implicit none
X      integer n1, n2, n3
X      real a1(n1,n2,n3), a2(n1,n2,n3), a3(n1,n2)
X
X      print*, '  A1 at (5,5,5) = ', a1(5,5,5)
X      print*, '  A2 at (50,50,5) = ', a2(50,50,5)
X      print*, '  A3 at (100,100) = ', a3(100,100)
X
X      return
X      end
SHAR_EOF
  $shar_touch -am 072003521999 'test.f' &&
  chmod 0644 'test.f' ||
  $echo 'restore of' 'test.f' 'failed'
  if ( md5sum --help 2>&1 | grep 'sage: md5sum \[' ) >/dev/null 2>&1 \
  && ( md5sum --version 2>&1 | grep -v 'textutils 1.12' ) >/dev/null; then
    md5sum -c << SHAR_EOF >/dev/null 2>&1 \
    || $echo 'test.f:' 'MD5 check failed'
56fcf5e4e4a8e29715947f37d8598782  test.f
SHAR_EOF
  else
    shar_count="`LC_ALL= LC_CTYPE= LANG= wc -c < 'test.f'`"
    test 1943 -eq "$shar_count" ||
    $echo 'test.f:' 'original size' '1943,' 'current size' "$shar_count!"
  fi
fi
# ============= test_seg_fault.f ==============
if test -f 'test_seg_fault.f' && test "$first_param" != -c; then
  $echo 'x -' SKIPPING 'test_seg_fault.f' '(file already exists)'
else
  $echo 'x -' extracting 'test_seg_fault.f' '(text)'
  sed 's/^X//' << 'SHAR_EOF' > 'test_seg_fault.f' &&
C
C23456789012345678901234567890123456789012345678901234567890123456789012
C----+-----------------------------------------------------------------+
C
X      program test
X
X      implicit none
X      integer nx, ny, nz, ia1, ia2, ia3
X      real a1(1), a2(1), a3(1)
X
X      open (unit=11, file='junk.output')
X      write (11,*) 'this is a test'
X      nx = 151
X      ny = 113
X      nz = 37
X      call mem_alloc(nx*ny*nz, a1, ia1)
X      call mem_alloc(nx*ny*nz, a2, ia2)
X      call mem_alloc(nx*ny, a3, ia3)
X      print*, '  Allocated arrays with nx,ny,nz = ', nx,ny,nz
X      print*, '  Offsets: ', ia1, ia2, ia3
X      call assign_vals(a1(ia1), a2(ia2), a3(ia3), nx, ny, nz)
X      call print_vals(a1(ia1), a2(ia2), a3(ia3), nx, ny, nz)
X      close (unit=11)
X
X      end
C----+-----------------------------------------------------------------+
X      subroutine mem_alloc(npts, arr, iarr)
X      implicit none
X      integer npts, iarr, iaddr, iralloc
X      real arr(*)
X
X      iaddr = iralloc(npts, arr, iarr)
X
X      return
X      end
C----+-----------------------------------------------------------------+
X      subroutine assign_vals(a1, a2, a3, n1, n2, n3)
X      implicit none
X      integer n1, n2, n3, i, j, k
X      real a1(n1,n2,n3), a2(n1,n2,n3), a3(n1,n2)
X
X      do k = 1, n3
X         do j = 1, n2
X            do i = 1, n1
X               a1(i,j,k) = float(i)*0.6 + float(j)*50 - float(k)
X               a2(i,j,k) = float(i)/100. - float(j) + float(k)*10.
X            enddo
X         enddo
X      enddo
X      do j = 1, n2
X         do i = 1, n1
X            a3(i,j) = float(i) + (3.14159/180.0)*sin(float(j)/n2)
X         enddo
X      enddo
X
X      return
X      end
C----+-----------------------------------------------------------------+
X      subroutine print_vals(a1, a2, a3, n1, n2, n3)
X      implicit none
X      integer n1, n2, n3
X      real a1(n1,n2,n3), a2(n1,n2,n3), a3(n1,n2)
X
X      print*, '  A1 at (5,5,5) = ', a1(5,5,5)
X      print*, '  A2 at (50,50,5) = ', a2(50,50,5)
X      print*, '  A3 at (100,100) = ', a3(100,100)
X
X      return
X      end
SHAR_EOF
  $shar_touch -am 072003581999 'test_seg_fault.f' &&
  chmod 0644 'test_seg_fault.f' ||
  $echo 'restore of' 'test_seg_fault.f' 'failed'
  if ( md5sum --help 2>&1 | grep 'sage: md5sum \[' ) >/dev/null 2>&1 \
  && ( md5sum --version 2>&1 | grep -v 'textutils 1.12' ) >/dev/null; then
    md5sum -c << SHAR_EOF >/dev/null 2>&1 \
    || $echo 'test_seg_fault.f:' 'MD5 check failed'
3c3c8e2794a736a0be0c5e79594d5824  test_seg_fault.f
SHAR_EOF
  else
    shar_count="`LC_ALL= LC_CTYPE= LANG= wc -c < 'test_seg_fault.f'`"
    test 2042 -eq "$shar_count" ||
    $echo 'test_seg_fault.f:' 'original size' '2042,' 'current size' "$shar_count!"
  fi
fi
rm -fr _sh02795
exit 0
>From rth@cygnus.com Mon Jul 19 21:46:00 1999
From: Richard Henderson <rth@cygnus.com>
To: Toshiyasu Morita <tm@netcom.com>
Cc: gcc-bugs@egcs.cygnus.com, gcc-patches@egcs.cygnus.com
Subject: Re: new_ia32_branch internal compiler errors
Date: Mon, 19 Jul 1999 21:46:00 -0000
Message-id: <19990719214606.A15488@cygnus.com>
References: <199907162319.QAA18297@netcom15.netcom.com>
X-SW-Source: 1999-07/msg00660.html
Content-length: 4375

On Fri, Jul 16, 1999 at 04:19:30PM -0700, Toshiyasu Morita wrote:
> With the new_ia32_branch compiler from cvs today (gcc-2.96 1990621)
> using only the "-O2" compiler option:
> 
> ./g_phys.i: In function `ClipVelocity':
> ./g_phys.i:4067: internal error--unrecognizable insn:
> (insn 179 23 180 (set (reg:SI 82)
>         (and:SI (zero_extract:SI (reg:HI 34)
>                 (const_int 8 [0x8])
>                 (const_int 8 [0x8]))
>             (const_int 69 [0x45]))) -1 (insn_list 23 (nil))
>     (expr_list:REG_DEAD (reg:HI 34)
>         (nil)))

Puts the ug in ugly.  I'd fixed a similar problem with cmove by making
jump re-recognize the insns it emits and fail the optimization.  But for
setcc the call chain is much messier.

A more direct solution is to accept constructs we'd created before.
This also cures the loss of optimization seen in the cmove change.


r~


	* i386.c (cmpsi_operand): New predicate.
	* i386.h (PREDICATE_CODES): Add it.
	* i386.md (cmpsi): Use it.

Index: i386.c
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/config/i386/i386.c,v
retrieving revision 1.81.2.10
diff -c -p -d -r1.81.2.10 i386.c
*** i386.c	1999/07/08 20:01:39	1.81.2.10
--- i386.c	1999/07/20 04:39:28
*************** memory_displacement_operand (op, mode)
*** 1175,1180 ****
--- 1175,1207 ----
  
    return parts.disp != NULL_RTX;
  }
+ 
+ /* To avoid problems when jump re-emits comparisons like testqi_ext_0,
+    re-recognize the operand to avoid a copy_to_mode_reg that will fail.
+ 
+    ??? It seems likely that this will only work because cmpsi is an
+    expander, and no actual insns use this.  */
+ 
+ int
+ cmpsi_operand (op, mode)
+       rtx op;
+       enum machine_mode mode;
+ {
+   if (general_operand (op, mode))
+     return 1;
+ 
+   if (GET_CODE (op) == AND
+       && GET_MODE (op) == SImode
+       && GET_CODE (XEXP (op, 0)) == ZERO_EXTRACT
+       && GET_CODE (XEXP (XEXP (op, 0), 1)) == CONST_INT
+       && GET_CODE (XEXP (XEXP (op, 0), 2)) == CONST_INT
+       && INTVAL (XEXP (XEXP (op, 0), 1)) == 8
+       && INTVAL (XEXP (XEXP (op, 0), 2)) == 8
+       && GET_CODE (XEXP (op, 1)) == CONST_INT)
+     return 1;
+ 
+   return 0;
+ }
  \f
  /* Return true if the constant is something that can be loaded with
     a special instruction.  Only handle 0.0 and 1.0; others are less
Index: i386.h
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/config/i386/i386.h,v
retrieving revision 1.61.2.5
diff -c -p -d -r1.61.2.5 i386.h
*** i386.h	1999/07/01 17:04:18	1.61.2.5
--- i386.h	1999/07/20 04:39:28
*************** do { long l;						\
*** 2387,2393 ****
  				 UMIN, UMAX, COMPARE, MINUS, DIV, MOD,	\
  				 UDIV, UMOD, ASHIFT, ROTATE, ASHIFTRT,	\
  				 LSHIFTRT, ROTATERT}},			\
!   {"memory_displacement_operand", {MEM}},
  \f
  /* Functions in i386.c */
  
--- 2387,2395 ----
  				 UMIN, UMAX, COMPARE, MINUS, DIV, MOD,	\
  				 UDIV, UMOD, ASHIFT, ROTATE, ASHIFTRT,	\
  				 LSHIFTRT, ROTATERT}},			\
!   {"memory_displacement_operand", {MEM}},				\
!   {"cmpsi_operand", {CONST_INT, CONST_DOUBLE, CONST, SYMBOL_REF,	\
! 		     LABEL_REF, SUBREG, REG, MEM, AND}},
  \f
  /* Functions in i386.c */
  
*************** extern int mult_operator XPARAMS((xrtx, 
*** 2434,2439 ****
--- 2436,2442 ----
  extern int div_operator XPARAMS((xrtx, xmode));
  extern int arith_or_logical_operator XPARAMS((xrtx, xmode));
  extern int memory_displacement_operand XPARAMS((xrtx, xmode));
+ extern int cmpsi_operand XPARAMS((xrtx, xmode));
  
  extern int standard_80387_constant_p XPARAMS((xrtx));
  extern int symbolic_reference_mentioned_p XPARAMS((xrtx));
Index: i386.md
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/config/i386/i386.md,v
retrieving revision 1.78.2.8
diff -c -p -d -r1.78.2.8 i386.md
*** i386.md	1999/07/08 19:42:27	1.78.2.8
--- i386.md	1999/07/20 04:39:28
***************
*** 604,610 ****
  
  (define_expand "cmpsi"
    [(set (reg:CC 17)
! 	(compare:CC (match_operand:SI 0 "general_operand" "")
  		    (match_operand:SI 1 "general_operand" "")))]
    ""
    "
--- 604,610 ----
  
  (define_expand "cmpsi"
    [(set (reg:CC 17)
! 	(compare:CC (match_operand:SI 0 "cmpsi_operand" "")
  		    (match_operand:SI 1 "general_operand" "")))]
    ""
    "
>From martin@mira.isdn.cs.tu-berlin.de Mon Jul 19 23:00:00 1999
From: "Martin v. Loewis" <martin@mira.isdn.cs.tu-berlin.de>
To: cdunham@conwaycorp.net
Cc: gcc-bugs@gcc.gnu.org
Subject: Re: template code yields (almost) no asm output
Date: Mon, 19 Jul 1999 23:00:00 -0000
Message-id: <199907200553.HAA00615@mira.isdn.cs.tu-berlin.de>
References: <3793B9C5.FFD8F440@conwaycorp.net>
X-SW-Source: 1999-07/msg00661.html
Content-length: 1474

> example error (several pages, an error for each instance of this
> templated class in my code... btw this is during the link stage, as
> ***g++ outputs no errors during compilation***)
> 
> library_llist.o: In function `ExtractType(LList<SharedLibrary> *, char
> *, int)':
> /home/Zetta/War/War/new/library_llist.cpp:7: undefined reference to
> `LList<SharedLibrary>::LList(void)'

Thanks for your bug report. You did not provide the source for
library_llist, but I suspect the bug is in your code. 

There is a couple of things I noticed:

a) The declaration of template<class>class LList is apparently in an
   implementation file llist.cpp. How come that library_llist.cpp
   knows what this declaration is? Do you have a second copy of that
   declaration? That would be bad.

b) I'm not surprised that g++ does not generated any code in llist.s.
   You are not instantiating the template, and you get code only for
   template instantiations. So, in order to generate code for
   LList<SharedLibrary>, the compiler needs to see both LList and
   SharedLibrary at the same time. More formally, a template class or
   member of template class must be defined at the point of
   instantiation of the template. Most likely, this is not the case in
   your code; make sure the template code is defined in a header file.
   Perhaps you were using a compiler that accepted your code. This was
   an extension of the C++ standard in that compiler.

Hope this helps,
Martin