Status of the Hurd parts in glibc

[Thomas Schwinge]

Hi!

The Hurd parts of glibc need maintenance!


We currently (have to) build glibc ``--without-tls'' and forcibly set
``libc_cv_z_relro=no'', to work around bugs in the otherwise enabled
parts of the code.


Now patches like the following are being applied to glibc:

#v+
CVSROOT:        /cvs/glibc
Module name:    libc
Changes by:     drepper@sources.redhat.com      2006-04-09 05:50:08

Modified files:
        inet           : inet_ntoa.c 

Log message:
        Rewrite to use TLS instead of TSD.

Patches:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/inet/inet_ntoa.c.diff?cvsroot=glibc&r1=1.12&r2=1.13
#v-

``Rewrite to use TLS [...]'' and we don't support TLS.  The outcome of
that should be obvious.  And to be clear: it's _us_ who have to act.


The maintenance needed for now mainly involves building glibc from the
source code, running the test suite and fixing whatever errors occur.
This ``fixing'' unfortunately might be as deep going as fixing GNU Mach,
which Jeroen Dekkers already contributed to last summer.


What also could / should (Roland?) be done is to back-port the following
from GNU Mach's trunk (a.k.a. OSKit-Mach) to the gnumach-1-branch:

#v+
2002-10-03  Roland McGrath  <roland@frob.com>

        * i386/include/mach/i386/mach_i386.defs (i386_set_gdt, i386_get_gdt):
        New routines.
        * i386/i386/user_ldt.c (i386_set_gdt, i386_get_gdt): New functions.
        * i386/i386/gdt.h (USER_GDT, USER_GDT_SLOTS): New macros.
        * i386/i386/thread.h (struct i386_machine_state): New member user_gdt.
        * i386/i386/pcb.c (switch_ktss): Copy those slots into the GDT.
[...]
#v-


Please coordinate on bug-hurd@gnu.org if you want to help with these
issues or have further questions.


Regards,
 Thomas