Separate avhv_foo() key handling into avhv_keys(). Slightly tweaked

[p5sagit/p5-mst-13.2.git] / README.threads

Building

If you want to build with multi-threading support and you are
running Linux 2.x (with the LinuxThreads library installed:
that's the linuxthreads and linuxthreads-devel RPMs for RedHat)
or Digital UNIX 4.x or Solaris 2.x for recentish x (2.5 is OK)
then you should be able to use
    ./Configure -Dusethreads -Doptimize=-g -ders
    make
and ignore the rest of this "Building" section. If it doesn't
work or you are using another platform which you believe supports
POSIX.1c threads then read on.

Omit the -e from your ./Configure arguments. For example, use
    ./Configure -drs
When it offers to let you change config.sh, do so. If you already
have a config.sh then you can edit it and do
    ./Configure -S
to propagate the required changes.
In ccflags, insert -DUSE_THREADS (and probably -DDEBUGGING since
that's what I've been building with). Also insert any other
arguments in there that your compiler needs to use POSIX threads.
Change optimize to -g to give you better debugging information.
Include any necessary explicit libraries in libs and change
ldflags if you need any linker flags instead or as well.

More explicitly, for Linux (when using the standard kernel-threads
based LinuxThreads library):
    Add -DUSE_THREADS -D_REENTRANT -DDEBUGGING to ccflags and cppflags
    Add -lpthread to libs
    Change optimize to -g
For Digital Unix 4.x:
    Add -pthread -DUSE_THREADS -DDEBUGGING to ccflags
    Add -DUSE_THREADS -DDEBUGGING to cppflags
    Add -pthread to ldflags
    Change optimize to -g
    Add -lpthread -lc_r to lddlflags
    For some reason, the extra includes for pthreads make Digital UNIX
    complain fatally about the sbrk() delcaration in perl's malloc.c
    so use the native malloc as follows:
    Change usemymalloc to n
    Zap mallocobj and mallocsrc (foo='')
    Change d_mymalloc to undef
For Solaris, do the same as for Linux above.

Now you can do a
    make


O/S specific bugs

Solaris qsort uses a hidden mutex for synchronisation. If you die()
while doing a sort() then the resulting longjmp() leaves the mutex
locked so you get a deadlock the next time you try to sort().

LinuxThreads 0.5 has a bug which can cause file descriptor 0 to be
closed after a fork() leading to many strange symptoms. The
development version of LinuxThreads has this fixed but the following
patch can be applied to 0.5 for now:

----------------------------- cut here -----------------------------
--- linuxthreads-0.5/pthread.c.ORI	Mon Oct  6 13:55:50 1997
+++ linuxthreads-0.5/pthread.c	Mon Oct  6 13:57:24 1997
@@ -312,8 +312,10 @@
   free(pthread_manager_thread_bos);
   pthread_manager_thread_bos = pthread_manager_thread_tos = NULL;
   /* Close the two ends of the pipe */
-  close(pthread_manager_request);
-  close(pthread_manager_reader);
+  if (pthread_manager_request >= 0) {
+    close(pthread_manager_request);
+    close(pthread_manager_reader);
+  }
   pthread_manager_request = pthread_manager_reader = -1;
   /* Update the pid of the main thread */
   self->p_pid = getpid();
----------------------------- cut here -----------------------------


Building the Thread extension

The Thread extension is now part of the main perl distribution tree.
If you did Configure -Dusethreads then it will have been added to
the list of extensions automatically.

You can try some of the tests with
    cd ext/Thread
    perl -Mblib create.t
    perl -Mblib join.t
    perl -Mblib lock.t
    perl -Mblib unsync.t
    perl -Mblib unsync2.t
    perl -Mblib unsync3.t
    perl -Mblib io.t
    perl -Mblib queue.t
The io one leaves a thread reading from the keyboard on stdin so
as the ping messages appear you can type lines and see them echoed.

Try running the main perl test suite too. There are known
failures for op/misc test 45 (tries to do local(@_) but @_ is
now lexical) and for some of the DBM/DB extensions (if there
underlying libraries were not compiled to be thread-aware).
may or may not work.


Bugs

* cond.t hasn't been redone since condition variable changed.

* FAKE_THREADS should produce a working perl but the Thread
extension won't build with it yet.

* There's a known memory leak (curstack isn't freed at the end
of each thread because it causes refcount problems that I
haven't tracked down yet) and there are very probably others too.

* There may still be races where bugs show up under contention.

* Need to document "lock", Thread.pm, Queue.pm, ...

* Plenty of others


Debugging

Use the -DL command-line option to turn on debugging of the
multi-threading code. Under Linux, that also turns on a quick
hack I did to grab a bit of extra information from segfaults.
If you have a fancier gdb/threads setup than I do then you'll
have to delete the lines in perl.c which say
    #if defined(DEBUGGING) && defined(USE_THREADS) && defined(__linux__)
        DEBUG_L(signal(SIGSEGV, (void(*)(int))catch_sigsegv););
    #endif


Background

Some old globals (e.g. stack_sp, op) and some old per-interpreter
variables (e.g. tmps_stack, cxstack) move into struct thread.
All fields of struct thread which derived from original perl
variables have names of the form Tfoo. For example, stack_sp becomes
the field Tstack_sp of struct thread. For those fields which moved
from original perl, thread.h does
    #define foo (thr->Tfoo)
This means that all functions in perl which need to use one of these
fields need an (automatic) variable thr which points at the current
thread's struct thread. For pp_foo functions, it is passed around as
an argument, for other functions they do
    dTHR;
which declares and initialises thr from thread-specific data
via pthread_getspecific. If a function fails to compile with an
error about "no such variable thr", it probably just needs a dTHR
at the top.


Fake threads

For FAKE_THREADS, thr is a global variable and perl schedules threads
by altering thr in between appropriate ops. The next and prev fields
of struct thread keep all fake threads on a doubly linked list and
the next_run and prev_run fields keep all runnable threads on a
doubly linked list. Mutexes are stubs for FAKE_THREADS. Condition
variables are implemented as a list of waiting threads.


Mutexes and condition variables

The API is via macros MUTEX_{INIT,LOCK,UNLOCK,DESTROY} and
COND_{INIT,WAIT,SIGNAL,BROADCAST,DESTROY}.

A mutex is only required to be a simple, fast mutex (e.g. it does not
have to be recursive). It is only ever held across very short pieces
of code. Condition variables are only ever signalled/broadcast while
their associated mutex is held. (This constraint simplifies the
implementation of condition variables in certain porting situations.)
For POSIX threads, perl mutexes and condition variables correspond to
POSIX ones.  For FAKE_THREADS, mutexes are stubs and condition variables
are implmented as lists of waiting threads. For FAKE_THREADS, a thread
waits on a condition variable by removing itself from the runnable
list, calling SCHEDULE to change thr to the next appropriate
runnable thread and returning op (i.e. the new threads next op).
This means that fake threads can only block while in PP code.
A PP function which contains a COND_WAIT must be prepared to
handle such restarts and can use the field "private" of struct
thread to record its state. For fake threads, COND_SIGNAL and
COND_BROADCAST work by putting back all the threads on the
condition variables list into the run queue. Note that a mutex
must *not* be held while returning from a PP function.

Perl locks and condition variables are both implemented as a
condpair_t structure, containing a mutex, an "owner" condition
variable, an owner thread field and another condition variable).
The structure is attached by 'm' magic to any SV. pp_lock locks
such an object by waiting on the ownercond condition variable until
the owner field is zero and then setting the owner field to its own
thread pointer. The lock is semantically recursive so if the owner
field already matches the current thread then pp_lock returns
straight away. If the owner field has to be filled in then
unlock_condpair is queued as an end-of-block destructor and
that function zeroes out the owner field and signals the ownercond
condition variable, thus waking up any other thread that wants to
lock it. When used as a condition variable, the condpair is locked
(involving the above wait-for-ownership and setting the owner field)
and the spare condition variable field is used for waiting on.


Thread states


              $t->join
R_JOINABLE ---------------------> R_JOINED >----\
    |      \  pthread_join(t)         |  ^      |
    |       \                         |  | join | pthread_join
    |        \                        |  |      |
    |         \                       |  \------/
    |          \                      |
    |           \                     |
    |  $t->detach\ pthread_detach     |
    |            _\|                  |
ends|             R_DETACHED     ends | unlink
    |                       \         |
    |                   ends \ unlink |
    |                         \       |
    |                          \      |
    |                           \     |
    |                            \    |
    |                             \   |
    V    join          detach     _\| V
ZOMBIE ----------------------------> DEAD
       pthread_join   pthread_detach
       and unlink     and unlink



Malcolm Beattie
mbeattie@sable.ox.ac.uk
6 November 1997