Bug 17266 - Libgfortran doesn't build
Summary: Libgfortran doesn't build
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: rtl-optimization (show other bugs)
Version: 4.0.0
: P2 normal
Target Milestone: 4.0.0
Assignee: Eric Botcazou
URL:
Keywords: build, rejects-valid, wrong-code
Depends on:
Blocks: 16991
  Show dependency tree / graph
 
Reported: 2004-09-01 18:51 UTC by Eric Botcazou
Modified: 2004-09-21 06:00 UTC (History)
3 users (show)

See Also:
Host: sparc64-sun-solaris2.9
Target: sparc64-sun-solaris2.9
Build: sparc64-sun-solaris2.9
Known to work:
Known to fail:
Last reconfirmed: 2004-09-06 15:17:07

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Eric Botcazou 2004-09-01 18:51:47 UTC 
gcc version 3.5.0 20040901 (experimental)

Libgfortran aborts with the following error message:
/opt/build/eric/gcc/gcc/gfortran -B/opt/build/eric/gcc/gcc/
-B/opt/build/eric/local/gcc-3.5.0/sparc64-sun-solaris2.9/bin/
-B/opt/build/eric/local/gcc-3.5.0/sparc64-sun-solaris2.9/lib/ -isystem
/opt/build/eric/local/gcc-3.5.0/sparc64-sun-solaris2.9/include -isystem
/opt/build/eric/local/gcc-3.5.0/sparc64-sun-solaris2.9/sys-include -g -O2 -Wall
-fno-repack-arrays -fno-underscoring -c
/home/eric/cvs/gcc/libgfortran/intrinsics/selected_real_kind.f90  -fPIC -DPIC -o
.libs/selected_real_kind.o
 In file /home/eric/cvs/gcc/libgfortran/intrinsics/selected_real_kind.f90:61

end function
           1
 Internal Error at (1):
 gfc_validate_kind(): Got bad kind
gmake[2]: *** [selected_real_kind.lo] Error 1
gmake[2]: Leaving directory `/opt/build/eric/gcc/sparc64-sun-solaris2.9/libgfortran'
gmake[1]: *** [all] Error 2
gmake[1]: Leaving directory `/opt/build/eric/gcc/sparc64-sun-solaris2.9/libgfortran'
gmake: *** [all-target-libgfortran] Error 2


The selected_real_kind.inc file reads:

  integer, parameter :: c = 3
  type (real_info), parameter :: real_infos(c) = (/ &
    real_info (4, precision(0.0_4), range(0.0_4)), &
    real_info (8, precision(0.0_8), range(0.0_8)), &
    real_info (16, precision(0.0_16), range(0.0_16)) /)

and is the same as for SPARC 32-bit, which doesn't ICE on it.
Comment 1 Richard Henderson 2004-09-01 21:38:37 UTC 
I can't reproduce this with a cross-compiler.
Comment 2 Eric Botcazou 2004-09-01 22:48:08 UTC 
> I can't reproduce this with a cross-compiler.

Ok, that's what I feared, thanks for trying.  I'll wait a couple of days for a
miracle :-) and then start digging.
Comment 3 Eric Botcazou 2004-09-06 15:17:06 UTC 
Confirmed on Linux by C.Joensson.
Comment 4 Christian Joensson 2004-09-20 05:39:10 UTC 
AuroraFedora Core release 1.91 (Wombat FC2) UP UltraSparc IIi (Sabre) sun4u:

binutils-2.15.90.0.3-6
bison-1.875-7.1
dejagnu-1.4.2-11
expect-5.39.0-96.1
gcc-3.3.3-3
glibc-2.3.3-26sparc3.sparc
glibc-2.3.3-26sparc3.sparc64
glibc-kernheaders-2.6-14sparc
kernel-2.6.8-1.571sp1
tcl-8.4.5-6
zlib-1.2.1.1-2.1.sparc
zlib-1.2.1.1-2.1.sparc64

LAST_UPDATED: Sun Sep 19 21:25:30 UTC 2004

configure: --enable-__cxa_atexit --enable-shared --enable-
languages=c,c++,f95,java,objc

/bin/sh ./libtool --mode=compile /usr/local/src/trunk/objdir/gcc/gfortran -B/us
r/local/src/trunk/objdir/gcc/ -B/usr/local/sparc64-unknown-linux-gnu/bin/ -B/us
r/local/sparc64-unknown-linux-gnu/lib/ -isystem /usr/local/sparc64-unknown-linu
x-gnu/include -isystem /usr/local/sparc64-unknown-linux-gnu/sys-include  -g -O2
 -Wall -fno-repack-arrays -fno-underscoring -c -o _exp_r4.lo `test -f 'generate
d/_exp_r4.f90' || echo '/usr/local/src/trunk/gcc/libgfortran/'`generated/_exp_r
4.f90
/usr/local/src/trunk/objdir/gcc/gfortran -B/usr/local/src/trunk/objdir/gcc/ -B\
usr/local/sparc64-unknown-linux-gnu/bin/ -B/usr/local/sparc64-unknown-linux-gnu
/lib/ -isystem /usr/local/sparc64-unknown-linux-gnu/include -isystem /usr/local
/sparc64-unknown-linux-gnu/sys-include -g -O2 -Wall -fno-repack-arrays -fno-und
erscoring -c /usr/local/src/trunk/gcc/libgfortran/generated/_exp_r4.f90  -fPIC 
-DPIC -o .libs/_exp_r4.o
 In file /usr/local/src/trunk/gcc/libgfortran/generated/_exp_r4.f90:28

   specific__exp_r4 = exp (parm)
                          1
Error: Type of argument 'x' in call to 'exp' at (1) should be REAL(6965448), no
t REAL(4)
 In file /usr/local/src/trunk/gcc/libgfortran/generated/_exp_r4.f90:28

   specific__exp_r4 = exp (parm)
                     1
Error: Function reference to '/usr/local/src/trunk/gcc/libgfortran/generated/_e
xp_r4.f90:0: internal compiler error: Segmentation fault                    
Please submit a full bug report,                   
with preprocessed source if appropriate.                       
See <URL:http://gcc.gnu.org/bugs.html> for instructions.                       
make[3]: *** [_exp_r4.lo] Error 1                                              
make[3]: Leaving directory `/usr/local/src/trunk/objdir/sparc64-unknown-linux-g
nu/libgfortran'
Comment 5 Andrew Pinski 2004-09-20 05:44:22 UTC 
So this is wrong-code, I think this is a regression too but I almost think this is an RTL-opt problem or a 
target problem.
Comment 6 Eric Botcazou 2004-09-20 06:32:22 UTC 
I've been on it since yesterday.  The miscompiled function on Solaris is
intrinsic.c:add_functions, but I suspect it's not the same on Linux because of
the different code model (the C compiler can't even be bootstrapped on Solaris
as of today).
Comment 7 GCC Commits 2004-09-21 05:58:14 UTC 
Subject: Bug 17266

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	ebotcazou@gcc.gnu.org	2004-09-21 05:57:58

Modified files:
	gcc            : ChangeLog regmove.c 

Log message:
	PR rtl-optimization/17266
	* regmove.c (optimize_reg_copy_3): Make a new SUBREG for each insn.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.5544&r2=2.5545
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/regmove.c.diff?cvsroot=gcc&r1=1.162&r2=1.163
Comment 8 Eric Botcazou 2004-09-21 06:00:45 UTC 
See http://gcc.gnu.org/ml/gcc-patches/2004-09/msg01966.html