On Thu, 2003-11-06 at 18:59, Stefan Jones wrote:
> Hi,
> 
> I am just wondering why TLS is not enabled by default in glibc with
> linuxthreads. Both Debian and Redhat enable it by default.
> 
> Also the Glibc INSTALL file states:
> 
> 
> `--without-tls'
>      By default the C library is built with support for thread-local
>      storage if the used tools support it.  By using `--without-tls'
>      this can be prevented though there generally is no reason since it
>      creates compatibility problems.
>  
> 
> I realise it used to be the case that nvidia-glx would not work with it
> but that is no longer the case.
> 

Possibly.

> It could be that you don't trust the build tools to work (binutils,
> gcc); but I don't think that is any longer the case.
> 

Not really.

> It is just with this multiple version glibc thing, I am trying to select
> suitable flavours of glibc:
> 
> So far I am thinking of:
> /lib - linuxthreads, kernel-version=2.2.5 no-tls
> /lib/i686 - linuxthreads tls kernel-version=2.4.1
> /lib/tls - nptl tls kernel-version=2.6.0
> 
> with the ld-linux.so in /lib being the one from 
> "linuxthreads tls kernel-version=2.4.1" (if any of the libraries have
> TLS then the main ld-linux.so needs TLS)
> 

This is one of the reasons I did not want to do this.  Remember, all
that have to be built ...

> Debian unstable uses:
> 
> -march=i686 -mcpu=i686 -g0 -O3 -fomit-frame-pointer
> -D__USE_STRING_INLINES
> 
> for i686 optimised binaries; I would like to see the same possibility in
> gentoo. Debian after all has always been seem to err on the side of
> caution.
> 

We just use -O2, which I still think is the safer one.  I have
no idea on -D__USE_STRING_INLINES, but -fomit-frame-pointer
will at least create issues if anybody wants to build a glibc
with -ggdb, and I am not sure on if it have any effect on
stuff linking runtime against it when run in a debugger.


Thanks,

-- 

Martin Schlemmer
Gentoo Linux Developer, Desktop/System Team Developer
Cape Town, South Africa


Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to