Re: Toolchain TLS support

[Thomas Schwinge]

[Cced to bug-hurd and hurd-devel.]

Hello!

For one part, this is essentally a repetition of
;
everything in there still applies.  For the other part, new issues are to
be worked on, see below.


On Sat, Aug 26, 2006 at 12:49:45AM +0200, Aurelien Jarno wrote:
> - hurd-i386
>   It has TLS support in binutils, gcc, and glibc, but it is currently
>   disabled in glibc (if I remember correctly, it is due to kernel
>   freeze with TLS activated).

The kernel panic issue has been resolved by Jeroen Dekkers.  What is left
to do is locating and fixing remaining bugs in glibc (and possibly GNU
Mach as well, who knows).


> However TLS is starting to be used in some packages. This include for
> example glibc 2.4

Indeed.  glibc 2.4 is completely unusable for GNU/Hurd at the moment,
also for reasons other than tls.  I have the beginnings of patches for
some of these issues, but lack both time and Roland's (or someone else's)
help to get these into an useful state.  If someone wants to work on
that, please speak up on.


> In short TLS support will probably be unavoidable and a requirement for 
> etch + 1. I think it is time for you porters (mainly for m68k and 
> hurd-i386) to start working on TLS support so that it will be ready for 
> etch + 1.

Nobody is working on that at the moment, as far as I know.


Also, a GCC 4.1-build glibc doesn't work correctly: it already fails
during the building process, as soon as the newly created libc.so is
being used (might also be ld.so, I don't remember exactly).


So, dear reader, if you want to help, please speak up: either directly to
me or -- preferred -- on the  mailing list.


Regards,
 Thomas


signature.asc
Description: Digital signature
___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd

Re: Toolchain TLS support

[Thomas Schwinge]

On Tue, Sep 05, 2006 at 11:05:12AM +0200, Alfred M. Szmidt wrote:
>glibc 2.4 is completely unusable for GNU/Hurd at the moment,
> 
> Not true.  It might be true for Debian GNU/Hurd though.

Wrong.  It's also true for the glibc as it is in upstream's cvs
repository.  Period.  If you have patches to fix these issues, then you
forgot to publish them.


>Also, a GCC 4.1-build glibc doesn't work correctly: it already
>fails during the building process, as soon as the newly created
>libc.so is being used (might also be ld.so, I don't remember
>exactly).
> 
> GCC 4.1 built glibc works very well.

Wrong.  The glibc, as it is in upstreams's cvs repository, does not work
(for GNU/Hurd) with GCC 4.1.  Period.  If you have patches to fix this
issue, then you forgot to publish them.


>So, dear reader, if you want to help, please speak up: either
>directly to me or -- preferred -- on the  mailing
>list.
> 
> Sorry, but I am not allowed to help.

Wrong.  Why shouldn't you be allowed to help?  You have only been banned
from the public bug-hurd mailing list, because of your intolerable
behavior.  But -- as you should know -- there are more ways to contribute
to the project.  The Savannah {bug,patch,task} trackers come to mind, for
example.  Apart from that, you can send email directly to the
maintainers.  Etc.


Regards,
 Thomas


signature.asc
Description: Digital signature
___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd

Re: Toolchain TLS support

[Samuel Thibault]

Alfred M. Szmidt, le Tue 05 Sep 2006 11:57:08 +0200, a écrit :
> As I said, these patches will become avaiable with time.  But as such,
> both GCC and GLIBC work fine on GNU; you are of course free to inisist
> on otherwise if that makes you happy.

By "as such", do you mean "upstream glibc 2.4 release and gcc 4.1", or
your patched versions?

Samuel


___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd

Re: Toolchain TLS support

[Samuel Thibault]

Alfred M. Szmidt, le Tue 05 Sep 2006 12:44:12 +0200, a écrit :
>> As I said, these patches will become avaiable with time.  But as
>> such, both GCC and GLIBC work fine on GNU; you are of course free
>> to inisist on otherwise if that makes you happy.
> 
>By "as such", do you mean "upstream glibc 2.4 release and gcc 4.1",
>or your patched versions?
> 
> That is a trick question,

There is no trick here.

> glibc hasn't worked out of the box for
> atleast a year since malloc/memusage.c uses __thread without #if's.

Ok so glibc 2.4 as such doesn't work (that was the original question).

Samuel


___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd