Higher order regression test problems egcs-1.0

[Jeff Johnson <jbj at jbj dot org>]

Hi --

I'm trying to stabilize the following pile of bits on a sparc-sun-sunos4.1.3

	egcs-1.0	(with-gnu-as,with-gnu-ld,enable-shared,enable-haifa)
	egcs-1.0.diff
	libg++-2.8.0b6.5
	binutils-2.8.1.0.18	(and precursors)

I've been shoveling these bits on several platforms for the last month
while acquiring experience on what patches are needed, etc. I've managed to
pass "make -k check" on all platforms from a variety of starting points.

So I tried to get on a final approach pattern and essentially did the
following

	rebuilt	egcs		(same tree, 4th time)
	rebuilt binutils	(2.8.0.1.18 1st time)
	rebuilt egcs		(skipped make check, 1st mistake)
	make install

I say essentially because I can't remember exactly -- I've been doing
several builds a day for too long.

After noticing something was smelly, I ran "make -k check" and have
massive problems with g++. Since many of the problems were core dumps
in libg++, I started looking more closely at binutils, backing up to
2.8.1.0.17, 2.8.1.0.15, and 2.8.1.

Running make check on binutils-2.8.1.0.18, I get

                === ld tests ===

Running target unix
Using /usr/local/bin/../share/dejagnu/baseboards/unix.exp as board description f
ile for target.
Using /usr/local/bin/../share/dejagnu/config/unix.exp as generic interface file
for target.     
Using /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/config/default.exp as tool-a
nd-target-specific interface file.
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-bootstrap/bootstrap.exp
 ...
FAIL: bootstrap
FAIL: bootstrap with --static
FAIL: bootstrap with --traditional-format
FAIL: bootstrap with --no-keep-memory
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-cdtest/cdtest.exp ...
FAIL: cdtest
FAIL: cdtest with -Ur
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-elfvers/vers.exp ...
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-empic/empic.exp ...
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-scripts/crossref.exp ..
.
FAIL: NOCROSSREFS 1
FAIL: NOCROSSREFS 2 
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-scripts/defined.exp ...
FAIL: DEFINED (PRMS 5699)
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-scripts/phdrs.exp ...
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-scripts/script.exp ...
FAIL: script
FAIL: MRI script  
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-scripts/sizeof.exp ...
FAIL: SIZEOF
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-sh/sh.exp ...
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-shared/shared.exp ...
FAIL: shared
FAIL: shared (PIC main)
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-srec/srec.exp ...
FAIL: S-records
FAIL: S-records with constructors
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-undefined/undefined.exp
 ...
FAIL: undefined
FAIL: undefined function
FAIL: undefined line
Running /u/obpa5/jbj/binutils-2.8.1.0.18/ld/testsuite/ld-versados/versados.exp .
..

                === ld Summary ===
 
# of unexpected failures        19


Running make check on the other binutils has *only*

FAIL: S-records with constructors

but that's enough to booger most of the g++ tests in egcs-1.0

I see no way that I can compile my way out of this mess with my
currently installed egcs-1.0, so I'm gonna replay it real slow starting
with gcc-2.7.2.3.

Any hints or shortcuts appreciated. In particular, what binutils is
appropriate for sparc-sun-sunos4.1.3?

73 de Jeff
	
-- 
Jeff Johnson	ARS N3NPQ
jbj@jbj.org
Gaithersburg, MD