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: Enable -fomit-frame-pointer on i386


> Jan Hubicka writes:
>  > > Jan Hubicka writes:
>  > >  > Perhaps we can simply replace use of backtrace() in libjava by
>  > >  > the Jakub's unwind based implementation to break the dependency on
>  > >  > targets where unwind info shall be available (ia-64/x86-64/i386/s390 at
>  > >  > least)
>  > > 
>  > > That's a good idea.  Bryce is working on such a patch, I believe.
>  > > 
>  > > In the meantime, this is what we do with x86-64 
>  > > 
>  > > in configure.in:
>  > > 
>  > > BACKTRACESPEC=
>  > > case "${host}" in
>  > >  x86_64*-*-linux*)
>  > >     BACKTRACESPEC=-fno-omit-frame-pointer
>  > >     ;;
>  > > esac
>  > > 
>  > > And configure.host:
>  > > 
>  > >   x86_64-*)
>  > > 	CHECKREFSPEC="%{m32:-fcheck-references}"
>  > > 	sysdeps_dir=x86-64
>  > > 	libgcj_flags="${libgcj_flags} -ffloat-store -fno-omit-frame-pointer"
>  > 
>  > I this still important?  x86-64 glibc can do unwinding without frame
>  > pointer for some time already.
> 
> It's important until no-one uses an earlier glibc.  Leave it 'til 3.5.

When the glibc switched to dwarf2 unwinding?  I was under impressions
that all released glibcs does it.  At least glibc CVS apparently never
contained some other implementation:
RCS file: /cvs/glibc/libc/sysdeps/x86_64/backtrace.c,v
Working file: backtrace.c
head: 1.1
branch:
locks: strict
access list:
symbolic names:
keyword substitution: kv
total revisions: 1;     selected revisions: 1

Honza
> 
> Andrew.


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