g++ -mabi=64 testsuite failures on irix6.5

Billinghurst, David (CRTS) David.Billinghurst@riotinto.com
Mon May 21 16:33:00 GMT 2001


Yes.  I have a work around in my local trees.  I had forgotten about it.
(Submitted to gcc-patches some time last year)

The failures I am seeing are not -n32/-64 linker errors.  The testsuite
failures core dump on execution.  I have had a few attempts to debug these,
but:
*	I am out of my depth,
*	gdb doesn't work (for me at least) with c++ code
*	dbx sort of works
*	the crashes seem to be in startup code

> -----Original Message-----
> From:	Jeffrey Oldham [SMTP:oldham@codesourcery.com]
> Sent:	Tuesday, 22 May 2001 7:46
> To:	Billinghurst, David (CRTS)
> Cc:	gcc-bugs@gcc.gnu.org; oldham@codesourcery.com
> Subject:	Re: g++ -mabi=64 testsuite failures on irix6.5
> 
> David, you complained about numerous C++ execution failures for
> mips-sgi-irix6.5 with -mabi=64.  Apparently, "-mabi=64" was not passed
> to gcc/testsuite/lib/g++.exp:g++_include_flags and g++_link_flags,
> which construct -I and -L compiler and linker options.  Thus, the
> "mabi=64" was omitted from directory paths.
> 
> This patch fixes that so the linker no longer complains that n32
> libraries are being provided when 64-bit libraries are desired.
> Unfortunately, I now experience numerically more errors, some
> execution and some excessive.  I'll look into these problems to see if
> I am causing them or they were previously present.
> 
> Mon May 21 15:29:41 2001  Jeffrey D. Oldham  <joldham@n02.acl.lanl.gov>
> 
> 	* lib/g++.exp (g++_include_flags): Use args to compute
>  	get_multilibs.
> 	(g++_link_flags): Likewise.
> 	(g++_init): Pass TOOL_OPTIONS as arguments to two above functions.
> 
> Under development.
> 
> Thanks,
> Jeffrey D. Oldham
> oldham@codesourcery.com << File: g++.exp.patch >> 



More information about the Gcc-bugs mailing list