This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: [PATCH] Fix boehm-gc crash on s390x


I tried a quick -Wall build, and things don't look that bad.
And it would be good to look at the warnings that did get generated.

Were you talking about that or -Werror?  I don't see any way to
use -Werror unless there's a way to override a few specific and
not easily avoidable warnings.  In particular:

1. The collector needs to get a bound on the stack pointer by
calling a function which returns the address of a local.  Gcc
correctly warns about that.  (The GC is probably about the only
program in the world that should be doing this.)

2. When I last looked into this, many years ago, there were
some bogus warnings about possibly uninitialized variables.
Fixing those would have involved a performance cost.  This
may no longer an issue on modern processors, but I would
have to look at it again.  (I generally have a low threshold
for adding instructions to get rid of warnings, since I have
to deal with more than one compiler, and this can quickly
get out of hand if they generate warnings in different
places.)

I'm mostly working on GC version 7.0alpha1, which tries to clean up
a bunch of things in the GC.  I'll start building that with
-Wall and get rid of as many of the warnings as possible.

If someone wants to do the same for 6.3, I certainly wouldn't
object.

Hans



> -----Original Message-----
> From: java-patches-owner@gcc.gnu.org
> [mailto:java-patches-owner@gcc.gnu.org]On Behalf Of Tom Tromey
> Sent: Tuesday, October 19, 2004 10:23 AM
> To: Ulrich Weigand
> Cc: gcc-patches@gcc.gnu.org; java-patches@gcc.gnu.org
> Subject: Re: [PATCH] Fix boehm-gc crash on s390x
> 
> 
> >>>>> "Ulrich" == Ulrich Weigand 
> <weigand@i1.informatik.uni-erlangen.de> writes:
> 
> Ulrich> (This would have been easier to find if boehm-gc was 
> built with
> Ulrich> warnings enabled ...)
> 
> I'm for enabling warnings.  I like warning-free code.  Before
> starting, though, it would make sense to see what Hans thinks about
> it, since we want to remain in sync with his releases.
> 
> Ulrich> Bootstrapped/regtested on s390-ibm-linux and s390x-ibm-linux.
> Ulrich> OK for mainline?
> 
> Yes, thanks.
> 
> Tom
> 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]