Commit Graph

18 Commits

Author SHA1 Message Date
hpk 941d06e509 [svn r63040] try harder to record and auto-exit gateways after test runs
--HG--
branch : trunk
2009-03-18 16:51:55 +01:00
hpk 9b81b15b74 [svn r57544] still notify callbacks if we can't send to the other side anymore.
--HG--
branch : trunk
2008-08-21 14:00:08 +02:00
hpk 2371fb16fa [svn r45549] fix a bug that servemain() wouldn't actually
try to join threads at exit because of a
(hidden) attribute error.

--HG--
branch : trunk
2007-08-08 14:02:55 +02:00
hpk 62634136b9 [svn r45548] - refactoring cleanup mechanics into its own class
- have setDaemon(1) for the receiverThread as
  otherwise on python2.5 atexit will not be invoked
  (the receiver thread apparently blocks it)

--HG--
branch : trunk
2007-08-08 13:45:04 +02:00
hpk 3cb7d3579a [svn r45547] remove global
--HG--
branch : trunk
2007-08-08 12:38:22 +02:00
hpk a94866a20b [svn r45541] don't use globals when we might be called from destructors
--HG--
branch : trunk
2007-08-07 19:54:09 +02:00
hpk a5e69d2035 [svn r45539] merge the execnet lessthreads branch (using the branch'es history):
* now by default Gateways DO NOT SPAWN execution threads
  you can call "remote_init_threads(NUM)" on an already instantiated
  gateway, which will install a loop on the other side which will
  dispatch each execution task to its own thread.

* execution is dissallowed on the side which initiates a gateway
  (rarely used, anyway)

* some cleanups (hopefully)

--HG--
branch : trunk
2007-08-07 19:34:59 +02:00
hpk 8bf738614e [svn r39982] introduce gateway._send and have all places route
their sending of Messages (or None's) through that method.

--HG--
branch : trunk
2007-03-06 13:51:18 +01:00
hpk ed2759b262 [svn r38579] library code should not use magic timeouts,
testing code may use a global TESTTIMEOUT

--HG--
branch : trunk
2007-02-12 16:55:48 +01:00
hpk 8e33cf6772 [svn r38427] significantly reducing the trailing "Exception in Thread"
messages by having threads die more silently.  Most
of these threads and tracebacks are anyway on some
remote machine, and we need to take care that errors
are routed to the place of invocation (e.g. the
remote_exec() invoker should get remote exceptions,
if the remote thread goes out of scope, the resulting
exception will not be seen likely, anyway)

--HG--
branch : trunk
2007-02-10 23:16:47 +01:00
hpk 59264d57ae [svn r37822] added lots of docstrings, general cleanup
--HG--
branch : trunk
2007-02-02 20:57:47 +01:00
hpk 272812fd76 [svn r37787] more precision at shutdown: first remove
from the sendqueue, then trigger the queue
to shutdown. (This is an attempt to get
rid of the still appearing "exception in thread"
messages)

--HG--
branch : trunk
2007-02-02 02:07:37 +01:00
hpk 0c55062e74 [svn r37786] improve docstring
--HG--
branch : trunk
2007-02-02 02:05:31 +01:00
hpk 7a718ca2e7 [svn r37785] privatizing some more attributes
--HG--
branch : trunk
2007-02-02 02:02:55 +01:00
hpk b4919c8102 [svn r37784] privatizing some attributes
--HG--
branch : trunk
2007-02-02 01:34:40 +01:00
hpk 3634701598 [svn r37780] simplifying code a bit (but test_confusion* for Ssh still fails)
--HG--
branch : trunk
2007-02-02 00:57:55 +01:00
hpk dad2da4583 [svn r37456] some small adjustments/cleanups regarding intiialisation
--HG--
branch : trunk
2007-01-28 09:48:59 +01:00
hpk 5992a8ef21 [svn r37264] create the new development trunk
--HG--
branch : trunk
2007-01-24 15:24:01 +01:00