]> sourceware.org Git - glibc.git/blame - linuxthreads/Changes
(__pthread_initialize_manager): Remove one last p_multiple_threads call.
[glibc.git] / linuxthreads / Changes
CommitLineData
c9326ed5
UD
1Release 0.9:
2- more ports (SH, IA-64, s390)
3- many bug fixes
4- timed sync object wait functions
5- barrier implementation
6- spinlocks implementation
7- thread register on x86
8- variable stack size and position on some platforms
9
10Release 0.8:
11(ehmm, forgot to update, don't know anymore)
12
5afdca00
UD
13Release 0.7:
14- Destructors for thread-specific data now conform to the POSIX semantics
15 (call destructors again if non-NULL TSD remains after a round of
16 destruction).
17- Implemented thread-specific data as a sparse array, allows more TSD keys
18 and smaller thread descriptors (Ulrich Drepper).
19- Added "error checking" mutexes.
20- Protect against multiple sigwait() on the same signals.
c9326ed5 21- Simplified implementation of semaphores when compare_and_swap is
5afdca00
UD
22 not available.
23- Fixed bug in fork() where stdin was closed if fork() was called before
24 the first pthread_create().
25- Fixed bug in the gethostby*_r functions (bad result if null bytes
26 in addresses).
27- Typos in manual pages corrected.
28- First cut at a PowerPC port (not working yet, runs into problems
29 with gcc and with the C library).
30
31Release 0.6:
32- Validation of thread identifiers: no more crashes when operating on
33 a thread that has exited (based on Pavel Krauz's ideas).
34- Added fallback implementation of semaphores for the 386 and the
c9326ed5 35 Sparc.
5afdca00
UD
36- Fixed a bug in signal handling causing false restarts of suspended
37 threads.
38- Fixed a bug in realtime scheduling causing all threads to have
39 default scheduling on Ix86 with libc5.
40- With realtime scheduling, unlocking a mutex now restarts the
41 highest priority thread waiting on the mutex, not the
42 first-suspended thread (Richard Neitzel).
43- Timing a process now returns cumulative times for all threads, not
44 just times for the initial thread (suggested by Wolfram Gloger).
45- Cleaned up name space (internal defs prefixed by __, weak aliases
46 for non-portable extensions).
47- MIPS port (contributed by Ralf Baechle).
48
49Release 0.5:
50- Signal-safe semaphores a la POSIX 1003.1b added.
51- Locking bug in pthread_mutex_trylock over recursive mutexes fixed.
52- Race conditions in thread cancellation fixed.
53- Sparc port (contributed by Miguel de Icaza).
54- Support for getpwnam_r and getpwuid_r.
55- Added pthread_kill_other_threads_np to be used in conjunction with
56 exec*().
57
58Release 0.4:
59- Manual pages for all functions.
60- Synchronization bug causing accumulation of zombie processes fixed.
61- Race condition in pthread_cond_timedwait fixed.
62- Recursive mutexes are back by popular demand.
63- Partial support for realtime scheduling (initiated by Richard Neitzel).
c9326ed5 64- pthread.h cleaned up a lot: now C++ compatible, added missing "const"
5afdca00
UD
65 qualifiers, added short documentation, put to GNU libc standards
66 for name space pollution (Ulrich Drepper).
67- Motorola 68k port (contributed by Andreas Schwab).
68- Interaction with fork(2) cleaned up a lot.
69
70Release 0.3:
71- Thread creation and reclaimation now performed by a centralized
72 "thread manager" thread.
73- Removed recursive mutexes to make regular mutexes more efficient.
74- Now available as a shared library (contributed by Richard Henderson).
75- Alpha port (contributed by Richard Henderson).
76- Fixed many small discrepancies with Posix 1003.1c.
77- Put under the LGPL instead of the GPL.
78
79Release 0.2:
80- Reentrant libc functions (adapted from libc 5.3.9 by Peeter Joot)
81- pthread_cond_wait did not reacquire the mutex correctly on return
82- More efficient pthread_cond_broadcast
83
84Release 0.1:
85- First public release
This page took 0.194015 seconds and 5 git commands to generate.