gnewsense-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnewsense-dev] binutils which works around loongson2f NOP bug by de


From: Karl Goetz
Subject: Re: [Gnewsense-dev] binutils which works around loongson2f NOP bug by default
Date: Sat, 10 Apr 2010 08:55:00 +0930

On Fri, 9 Apr 2010 16:13:44 -0400
Mark H Weaver <address@hidden> wrote:

> On Thu, Apr 08, 2010 at 08:06:59PM +0930, Karl Goetz wrote:
> > This issue has been discussed on debian-mips, and on this list
> > previously. Are you aware of those discussions?
> 
> I've found some discussions about the loongson2f NOP and JUMP
> workarounds, but I've been unable to find earlier suggestions that the
> NOP workaround should be applied by default.  Pointers welcome.

Perhaps the 'by default' discussion happened on irc. is there a reason
not build like this by default?

> > > * Added --enable-fix-loongson2f-nop option to gas/configure,
> > >   which makes -mfix-loongson2f-nop the default for MIPS targets
> > >   (unless overridden by -mnofix-loongson2f-nop).
> > > 
> > > * Pass --enable-fix-loongson2f-nop by default from debian/rules.
> > 
> > This enables it by default for all packages built with binutils?
> 
> Yes.  When this version of binutils is built using the
> --enable-fix-loongson2f-nop compile-time option, then all packages
> built using that binutils will include the NOP workaround (unless
> -mnofix-loongson2f-nop is passed to the assembler).

This sounds good.

> One minor caveat, however: some bare (unfixed) NOPs are still included
> in the C run-time library which is normally statically linked into
> executables.  I guess we need to recompile gcc using this version of
> binutils to take care of that issue.

Rebuilding gcc will only help for statically compiled packages from
this point though, so it seems like a suitably low priority issue there.

> > > I've also built binary packages for mipsel and placed them in
> > > fencepost.gnu.org:~mhw/binutils-metad/
> > 
> > Fencepost.gnu.org/~mhw/binutils-metad/ offers me a bin, is that
> > expected?
> 
> I'm not sure I understand the question, but if you're asking whether I
> intentionally included mipsel binary packages in that directory, the
> answer is yes.

If i visit your home via http icecat offers me the chance to download a
binary file. I have a login on fencepost, so i'll scp them out if i can.

> > * The changes are made directly to the source tree, rather then as
> > patches applied at package build time.
> 
> Ah yes, good point!  I've fixed this problem and put a new version of
> everything (the new patch relative to lenny-backports, and new source
> and mipsel binary packages) in:
> 
> fencepost.gnu.org:~mhw/binutils-metad1/

thanks,
kk

>    Thanks,
>      Mark


-- 
Karl Goetz, (Kamping_Kaiser / VK5FOSS)
Debian contributor / gNewSense Maintainer
http://www.kgoetz.id.au
No, I won't join your social networking group

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]