Re: Bootstrap failure of gcc cvs trunk (3.3) on sparc64-{solaris*,linux}

Re: Bootstrap failure of gcc cvs trunk (3.3) on sparc64-{solaris*,linux}

[c.christian.joensson]

On Fri, Dec 06, 2002 at 09:12:31AM +0100, Robert Schiele wrote:
> On Fri, Dec 06, 2002 at 08:19:54AM +0100, Christian Jönsson wrote:
> > Hej.
> > 
> > I was just wondering if you have any clue on what the bootstrap
> > failure might depend on?
> 
> I tracked that down to a really simple case, which is c++/8183.
> Current mainline fails when trying to compile a diamond object
> structure when the inheritance to the common base class is declared
> virtual.
> 
> > I was really amazed that Dave M got it through... I still don't... and
> 
> Note that Dave is building the compiler for Linux, where my failing
> case is on Solaris, although both are sparc64 systems.  Because of
> that I assume the problem is somewhat related to a different data type
> definition in the C library headers of those two platforms and some
> code in gcc cannot handle the Solaris version of that data type
> correctly.
> 
> Did you try on Linux or on Solaris?
> 
> > I don't know how to debug it either...
> 
> I hope, the simplified test case in c++/8183 gives you some hints.
> 
> > It would be neat to have it ironed out before the branching 3.3...
> 
> Sure.  I'm still waiting for a generic bug removing tool.  ;-)

Anyhow, the bootstrap/build succeeded on sparc64-linux for 

./contrib/gcc_update -D "Sep 13 10:00 UTC 2002"

but failed in complex_io.cc (ICE) for

./contrib/gcc_update -D "Sep 17 10:00 UTC 2002"

and I'd guess, from Robert's testsuite results report, that that's
true for sparc64-solaris* also.

The cvs diff for that time period is

Re: Bootstrap failure of gcc cvs trunk (3.3) on sparc64-{solaris*,linux}

[Christian Jönsson]

On Fri, Jan 03, 2003 at 07:12:27PM +0100, c.christian.joensson@telia.com wrote:
> On Fri, Dec 06, 2002 at 09:12:31AM +0100, Robert Schiele wrote:
> > On Fri, Dec 06, 2002 at 08:19:54AM +0100, Christian Jönsson wrote:
> > > Hej.
> > > 
> > > I was just wondering if you have any clue on what the bootstrap
> > > failure might depend on?
> > 
> > I tracked that down to a really simple case, which is c++/8183.
> > Current mainline fails when trying to compile a diamond object
> > structure when the inheritance to the common base class is declared
> > virtual.
> > 
> > > I was really amazed that Dave M got it through... I still don't... and
> > 
> > Note that Dave is building the compiler for Linux, where my failing
> > case is on Solaris, although both are sparc64 systems.  Because of
> > that I assume the problem is somewhat related to a different data type
> > definition in the C library headers of those two platforms and some
> > code in gcc cannot handle the Solaris version of that data type
> > correctly.
> > 
> > Did you try on Linux or on Solaris?
> > 
> > > I don't know how to debug it either...
> > 
> > I hope, the simplified test case in c++/8183 gives you some hints.
> > 
> > > It would be neat to have it ironed out before the branching 3.3...
> > 
> > Sure.  I'm still waiting for a generic bug removing tool.  ;-)
> 
> Anyhow, the bootstrap/build succeeded on sparc64-linux for 
> 
> ./contrib/gcc_update -D "Sep 13 10:00 UTC 2002"
> 
> but failed in complex_io.cc (ICE) for
> 
> ./contrib/gcc_update -D "Sep 17 10:00 UTC 2002"
> 
> and I'd guess, from Robert's testsuite results report, that that's
> true for sparc64-solaris* also.
> 
> The cvs diff for that time period is

sorry, that was a blur, the last successful bootstrap of 3.3 I had was
the posting to gcc-testresults,
http://gcc.gnu.org/ml/gcc-testresults/2002-09/msg00563.html, that
happened with LAST_UPDATED: Mon Sep 16 08:16:13 UTC 2002.

And, the first one to bootstrap fail due to ICE in complex_io that I
now have done using

./contrib/gcc_update -D "Sep 20 10:00 UTC 2002"

and nothing else.

The cvs diff between these dates are way to big to be helpful to me...

The boostrap failures I've had in between these dates are due to
comparison failures as I have seen trying a few now. I'll try to just
build the compiler and see what I get. 

Cheers,

/ChJ

Re: Bootstrap failure of gcc cvs trunk (3.3) on sparc64-{solaris*,linux}

[Robert Schiele]

[-- Attachment #1: Type: text/plain, Size: 735 bytes --]

On Mon, Jan 06, 2003 at 09:59:16PM +0100, Christian Jönsson wrote:
> comparison failures as I have seen trying a few now. I'll try to just
> build the compiler and see what I get. 

You will get a compiler that compiles stuff just fine, but fails to
compile stuff like:

# /export/gccbuild/gcc/cc1plus
class a {};
class b : virtual a {};
class c : virtual a {};
class d : b, c {};
<stdin>:4: internal compiler error: in lookup_base_r, at cp/search.c:284
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
#

Robert

-- 
Robert Schiele			Tel.: +49-621-181-2517
Dipl.-Wirtsch.informatiker	mailto:rschiele@uni-mannheim.de

[-- Attachment #2: Type: application/pgp-signature, Size: 481 bytes --]