Ticket #4504: 4504-1.bundle

1 patch for repository /home/adept/work/well-typed/parallel/kazu/ghc-src/ghc/libraries/unix:

Wed Dec  8 10:21:25 EET 2010  Dmitry Astapov <dastapov@gmail.com>
  * Fix for #4504: awaitSignal now works with threaded RTS

New patches:

[Fix for #4504: awaitSignal now works with threaded RTS
Dmitry Astapov <dastapov@gmail.com>**20101208082125
 Ignore-this: 34d8278d18f7776e211ac6e61f0e5bc5
] {
hunk ./System/Posix/Signals.hsc 109
 import System.Posix.Internals
 import System.Posix.Process
 import System.Posix.Process.Internals
+import Control.Concurrent (rtsSupportsBoundThreads)
 import Data.Dynamic
 
 #ifdef __GLASGOW_HASKELL__
hunk ./System/Posix/Signals.hsc 569
 -- | @awaitSignal iset@ suspends execution until an interrupt is received.
 -- If @iset@ is @Just s@, @awaitSignal@ calls @sigsuspend@, installing
 -- @s@ as the new signal mask before suspending execution; otherwise, it
--- calls @pause@.  @awaitSignal@ returns on receipt of a signal.  If you
--- have installed any signal handlers with @installHandler@, it may be
--- wise to call @yield@ directly after @awaitSignal@ to ensure that the
+-- calls @sigsuspend@ with current signal mask. With threaded RTS, 
+-- scheduler timer (either @virtualTimerExpired@ or @realTimeAlarm@) 
+-- is always added to the signal mask passed to @sigsuspend@ to
+-- prevent this call from waking up on context switches.
+--
+-- @awaitSignal@ returns when signal was received and processed by a
+-- signal handler, or if the signal could not be caught. If you have
+-- installed any signal handlers with @installHandler@, it may be wise
+-- to call @yield@ directly after @awaitSignal@ to ensure that the
 -- signal handler runs as promptly as possible.
 awaitSignal :: Maybe SignalSet -> IO ()
 awaitSignal maybe_sigset = do
hunk ./System/Posix/Signals.hsc 581
-  fp <- case maybe_sigset of
-         Nothing -> do SignalSet fp <- getSignalMask; return fp
-         Just (SignalSet fp) -> return fp
+  sigset <- case maybe_sigset of
+             Nothing -> getSignalMask
+             Just s -> return s
+  let SignalSet fp =
+        case rtsSupportsBoundThreads of
+          False -> sigset
+          True -> addSignal itimerSignal sigset -- Fix for #4504
   withForeignPtr fp $ \p -> do
   _ <- c_sigsuspend p
   return ()
hunk ./System/Posix/Signals.hsc 595
   -- (-1) with errno set to EINTR.
   -- XXX My manpage says it can also return EFAULT. And why is ignoring
   -- EINTR the right thing to do?
- 
+  where
+    itimerSignal =
+#if defined(USE_TIMER_CREATE)
+      virtualTimerExpired
+#elif defined(HAVE_SETITIMER)
+#  if !defined(HAVE_SETITIMER_VIRTUAL)
+      realTimeAlarm
+#  else
+      virtualTimerExpired
+#  endif
+#endif
+
 foreign import ccall unsafe "sigsuspend"
   c_sigsuspend :: Ptr CSigset -> IO CInt
 #endif
}

Context:

[Remove unnecessary fromIntegral calls
simonpj@microsoft.com**20101116172536
 Ignore-this: e77072c759445a64892738618ff03ea8
] 
[Bump version to 2.4.1.0
Ian Lynagh <igloo@earth.li>**20100917150114] 
[avoid Foreign.unsafePerformIO
Ross Paterson <ross@soi.city.ac.uk>**20100909125459
 Ignore-this: 890c8460db98d1ec18168d70291be9f1
] 
[Add bindings for setting effective user-id and group-id
Eric Mertens <emertens@galois.com>**20100517181945
 Ignore-this: 49f72b40befb133cdbf63742a389dad5
 
 Having the ability to set the effective user-id and group-id
 enables the application to switch between the real user-id
 and set-user-id multiple times.
] 
[Allow base-4.3
Simon Marlow <marlowsd@gmail.com>**20100602082509
 Ignore-this: 5d6dea210a68d3f290f69ee1feb1488a
] 
[add a comment about 3816 failing due to Samba
Simon Marlow <marlowsd@gmail.com>**20100518084422
 Ignore-this: daf0c2087064925e67b1a5a09ce7632c
] 
[Provide a dummy input file for queryfdoption01
Matthias Kilian <kili@outback.escape.de>**20100418172619
 Ignore-this: 4d7bfb388193ff6090f4a5fcfe798bbe
] 
[mention that forkProcess is not supported with +RTS -Nn where n > 1
Simon Marlow <marlowsd@gmail.com>**20100430082841
 Ignore-this: c7bc7e8262bc72244061a9aba2dd468d
] 
[make getAllGroupEntries work when called again; fixes #3816
Simon Marlow <marlowsd@gmail.com>**20100329112713
 Ignore-this: 8b5e3c1d8648d556b9f032a198a8475d
] 
[add test for #3816
Simon Marlow <marlowsd@gmail.com>**20100329112646
 Ignore-this: 4d07e5956a5bc6dfb7be59473ccf6b98
] 
[fix warnings
Simon Marlow <marlowsd@gmail.com>**20100322131615
 Ignore-this: 9fd97646b47d5d8d2957c0927885041e
] 
[handleToFd: close both sides of a DuplexHandle (#3914)
Simon Marlow <marlowsd@gmail.com>**20100319210802
 Ignore-this: d1c17df79644a7ad140099d0c1474c81
] 
[check for EINTR in openFd
Simon Marlow <marlowsd@gmail.com>**20100127114600
 Ignore-this: 28859b6a74c6d47d5c92db0eb688fa4c
] 
[accept --with-cc to set the path to gcc (#2966)
Simon Marlow <marlowsd@gmail.com>**20100127114329
 Ignore-this: 127eccfccbfc849ce05aed5904d16b75
] 
[System.Posix.Temp: rm whitespace
gwern0@gmail.com**20100119003437
 Ignore-this: 33b2bcf99f7b6b21461ea0aee7c12ea8
] 
[System.Posix.Temp: pad input filenames with 6 Xs
gwern0@gmail.com**20100119003022
 Ignore-this: 1e5a45074881f75d58e3a0f02525b264
 If the argument doesn't terminate in capital Xs, the C mkstemp will simply bomb out
 with zero warning. This was not documented.
 By arbitrarily sticking a bunch of Xes at the end of all arguments, we guarantee that
 this exception will not be thrown, the type signature will not change, and no existing
 code can break (since if it was manually avoiding the exception by adding "XXX" itself,
 the temp files will now be simply 3 random characters longer, nothing worse).
] 
[System.Posix.Temp: improve haddocks for mkstemp to explain what that returned String actually is
gwern0@gmail.com**20100119002755
 Ignore-this: 5f5ec7871a687a024cda18b1b0d1e044
] 
[fix base dependency: should be >= 4.2 (#3780), and bump verison to 2.4.0.1
Simon Marlow <marlowsd@gmail.com>**20100113113803
 Ignore-this: 121b61a9ea1ce1dcbe499285d1910d25
] 
[no good reason these tests should be expect_fail, as far as I can tell
Simon Marlow <marlowsd@gmail.com>**20091231145739
 Ignore-this: a8f0dcb753f8520fb6953b7f9acd50c6
] 
[Support for DragonFly
Simon Marlow <marlowsd@gmail.com>**20091221110709
 Ignore-this: cafb4888bc25aa8ed97ce947acabfaca
 Not sure where this patch came from, I found it in my validate tree.
] 
[Don't use absolute paths to headers (#3728)
Simon Marlow <marlowsd@gmail.com>**20091221110634
 Ignore-this: 64b17ec4d274fdc5bd9ee021d412c997
] 
[Add Haiku (#3727)
Simon Marlow <marlowsd@gmail.com>**20091221110554
 Ignore-this: a76b1450c341d80312f1354efa95c8a0
] 
[#include <fcntl.h>, not <sys/fcntl.h> (#3723)
Simon Marlow <marlowsd@gmail.com>**20091216104154
 Ignore-this: 9f11594ce66b2e2537b9a20a33912e0d
] 
[forking works in GHCi too
Simon Marlow <marlowsd@gmail.com>**20091111142055
 Ignore-this: 90ead989e64e3cbffbae15e0c74b5f89
 Or at least, it works well enough to run this test.  The main GHCi
 thread is gone after forking, but the current evaluation continues to
 run.
] 
[add a test for #1185
Simon Marlow <marlowsd@gmail.com>**20091021090741
 Ignore-this: 581e367a446bc4a3996b09d39fbb102b
] 
[run fdReadBuf only the threaded ways (fixed unregisterised failure)
Simon Marlow <marlowsd@gmail.com>**20091008110126
 Ignore-this: 7893c3ab18b25aa3c67b71633a1e7d35
] 
[Add a test from trac #2969
Ian Lynagh <igloo@earth.li>**20090923194837] 
[Bump version to 2.4.0.0
Ian Lynagh <igloo@earth.li>**20090920142001] 
[Add comments about why user001 might fail on Linux.
Simon Marlow <marlowsd@gmail.com>**20090917134046
 See  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=466647
] 
[Allow building with base 4.2
Ian Lynagh <igloo@earth.li>**20090911160736] 
[remove dead code
Simon Marlow <marlowsd@gmail.com>**20090908144627
 Ignore-this: f7e3696e27307be289f9fbf85750fab8
] 
[Use Foreign.Concurrent for Haskell finalizers (#3473)
Simon Marlow <marlowsd@gmail.com>**20090908142536
 Ignore-this: 72ce9a134562b451b0237e124d03c839
] 
[Update to follow RTS tidyp changes
Simon Marlow <marlowsd@gmail.com>**20090801193459
 Ignore-this: 825b19edb4cee29945b179f21a8538e4
] 
[NetBSD does not have support for symbol versioning, so updated systen
Simon Marlow <marlowsd@gmail.com>**20090723075725
 Ignore-this: 19a0cc5a913ba591074aab6c560ff5d8
 functions need to be given a new name, and the header files contain
 some __asm hackery in order to let the program call the correct function.
 
 This mean that you need to use the header files in order to call the
 correct system functions, which prevents things like "foreign import ccall" from working.
 
 Ghc solves this with wrapper functions for some of the renamed functions,
 but it has not been updated for newer versions of NetBSD that has recently
 versioned some more functions.
 
 The attached patches introduces wrapper functions for all currently
 NetBSD-versioned functions used in libraries/unix.  Solves ~20 testsuite
 failures.
 
 Contributed by: Krister Walfridsson <krister.walfridsson@gmail.com>
] 
[Fix warning
Ian Lynagh <igloo@earth.li>**20090711112701] 
[Fix some "warn-unused-do-bind" warnings where we want to ignore the value
Ian Lynagh <igloo@earth.li>**20090710200806] 
[Remove unused imports
Ian Lynagh <igloo@earth.li>**20090707115900] 
[Move directory stuff from base to here
Simon Marlow <marlowsd@gmail.com>**20090625093258
 Ignore-this: cd2ce9c12bac46285b5435f00c50a21e
 leaving out Windows-specific hacks
] 
[rename cache variables to keep recent autoconfs happy
Ross Paterson <ross@soi.city.ac.uk>**20090616140611
 Ignore-this: b522604509c77edeff37b3ec42f3741f
] 
[TAG 2009-06-25
Ian Lynagh <igloo@earth.li>**20090625160426] 
Patch bundle hash:
6907e120734ddfaeedf497472aba4a0722c7957a