CVS log for dietlibc/libpthread/pthread_internal.c

Up to [DVB] / dietlibc / libpthread

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

we don't need errno_location in libpthread if WANT_TLS is defined (in
fact it's harmful)

  aarch64 port (Christian Seiler)

  various TLS and pthread fixes (Christian Seiler)

remove compiler warnings

  fix fork+pthread closing fd 0 (Tim München)

use TCB (from TLS) on i386 if available

  first stab at getting TLS to work in actual threads
  use fs: to find current thread on x86_64

move <sys/poll.h> to <poll.h>

add missing _GNU_SOURCE

fix pthread_cond_timedwait

the regparm attribute in typedefinition generats error with gcc-2.95 on
non-x86 arch... (reported: Johannes Stezenbach)

ARGH!  gcc 3.4 breaks regparm attributes on i386!

fixing the pthread_join function
minor of the thread_specific data handling

add the semaphore support.

removed problematic longjmp/setjmp for timedwait

- checkin of the new libpthread
- add confstr

- parisc clone & atomic testandset
- modifications to libpthread to respect the UPgrowing stack of rarisc
  and inverse lock logic
- added compiler option to i386 (this saves 9 byte per C-function)

removed "some" warnings ...

major change:
 - fixed a misunderstanding bug / errors are returned not set to errno
 - new pthread_join struct (now it passes the glib check)
 - changed (and probably broke) the pthread_key* struct (but now it passes the glib test)

missed to commit ... (a.u.)

lots of errno cleanups,
lots of errno related thread-safeness cleanups in librpc (don't ask).

change return 42 to return (void*)42 (Olaf said it's OK, honest!)

fixed a bug...
if the thread-life was shorter then one time-slice it had slipped the
creator...

added internal call __get_errno

and fixed shadow passwd entry (spent) function NOW sshd works
(tested on my machine @ uni)

set_errno again ...

set_errno ...

finalize THREAD_SAVE renaming, include libpthread in standard targets
for each platform.