Re: Testsuite failure ?

[David Ronis <ronis at onsager dot chem dot mcgill dot ca>]

Hi Manfred,

Thanks for the reply.


	> 
	> I just uploaded a new set of testsuite results for gcc-ss-20000821
	> obtained after upgrading binutils from 2.10(release) to Binutils
	> (2.9.1.0.7) GNU ld version 2.10.90 (with BFD 2.10.90) [today's CVS].
	> 
	> The testsuites show essentially no improvements [well, perhaps I'm
	> being a bit harsh, the number of g++ unexpected failures is now 805
	> instead of 806].  The URL's for the two cases are:
	> 
	>      http://gcc.gnu.org/ml/gcc-testresults/2000-08/msg00205.html
	> 
	> and
	> 
	>      http://gcc.gnu.org/ml/gcc-testresults/2000-08/msg00181.html
	
	David, I'm afraid most of the failures are caused by the set of
	compilation flags you are using for running "make bootstrap"; I'm
	normally using something like
	
	  BOOT_CFLAGS="-O9 -march=i686 -funroll-loops -fomit-frame-pointer"
	
	i.e. omit all the -fforce-mem -fforce-addr -ffast-math -malign-*
	flags. However, the g++ compiler wasn't in too good shape during the
	fast few days. Maybe that caused your problems.
	
I know removing flags gets rid of errors.  I'm merely testing the
snapshots under agressive optimization--basically, I think the compiler
should at least be able to correctly compile itself and it's libraries
no matter flags are given.

	
	> 
	> The typical error message is now different though; e.g.,
	> 
	> PASS: g++.dg/special/conpr-1.C (test for excess errors)
	> spawn [open ...]
	> FAIL: g++.dg/special/conpr-1.C execution test
	> 
	> Is this real or does something else have to be upgraded?
	
	Which dejagnu version are you using? I have a recent checkout from
	sources.redhat.com for that as well....

Sigh...    I'm using the release version from ftp.gnu.org (1.3.1 with expect 
5.26).  I just checked out the version from redhat--it's much more recent.  I'll 
build it, rebuild/check gcc and report back.

Thanks for the pointer.


David