Alex Rønne Petersen [Mon, 21 Oct 2013 18:21:06 +0000 (11:21 -0700)]
Merge pull request #268 from pcc/menudeactivate
Cause MenuDeactivate event to be delivered when appropriate
Alex Rønne Petersen [Mon, 21 Oct 2013 18:19:47 +0000 (11:19 -0700)]
Merge pull request #260 from pcc/topmost
X11: improve handling of WS_EX_TOPMOST
Michael Hutchinson [Mon, 21 Oct 2013 18:09:23 +0000 (11:09 -0700)]
Merge pull request #565 from rneatherway/master
This is a possible fix for finding xml documentation files
Alex Rønne Petersen [Mon, 21 Oct 2013 17:00:26 +0000 (10:00 -0700)]
Merge pull request #587 from madewokherd/gdipdllmap
Ignore the gdiplus dllmap on Windows.
Alex Rønne Petersen [Mon, 21 Oct 2013 16:17:10 +0000 (18:17 +0200)]
Remove commented-out irix path from configure.in.
Alex Rønne Petersen [Mon, 21 Oct 2013 15:53:28 +0000 (17:53 +0200)]
Add mono_gc_wbarrier_generic_store_atomic () function.
This is used in place of mono_gc_wbarrier_generic_store () when
we need the store to be atomic with release semantics.
Alex Rønne Petersen [Fri, 18 Oct 2013 00:29:02 +0000 (02:29 +0200)]
Add an extension point in atomic.c.
Alex Rønne Petersen [Fri, 18 Oct 2013 00:26:58 +0000 (02:26 +0200)]
Use slow 64-bit cmpxchg for 32-bit ARM, MIPS, and PowerPC.
GCC doesn't provide a 64-bit __sync_val_compare_and_swap intrinsic
for these architectures.
Alex Rønne Petersen [Fri, 6 Sep 2013 02:16:21 +0000 (04:16 +0200)]
Correct our implementation of Thread.VolatileRead ()/VolatileWrite ().
We have various issues:
* All of the functions assumed that doing volatile loads/stores was
enough, when load-acquire, store-release semantics were actually
needed. This may have worked before purely by chance. We now use
proper memory barriers so that we don't have to hope for the
compiler and CPU to do the right thing.
* Removes the hack for 64-bit quantities on 32-bit systems. This is
no longer needed now that we use explicit memory barriers in
these functions. Also, these functions are not supposed to do
atomic reads/writes in the first place - they're purely about the
memory barrier semantics.
* The VolatileWrite (object&, object) overload was not using the
volatile qualifier at all, thus not getting volatile semantics
as per: http://gcc.gnu.org/onlinedocs/gcc/Volatiles.html
Alex Rønne Petersen [Wed, 25 Sep 2013 18:31:18 +0000 (20:31 +0200)]
Use InterlockedAdd64 () for Interlocked.Add (long&, long).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:31:06 +0000 (20:31 +0200)]
Use InterlockedAdd () for Interlocked.Add (int&, int).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:23:36 +0000 (20:23 +0200)]
Use InterlockedExchange64 () for Interlocked.Exchange (double&, double).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:22:36 +0000 (20:22 +0200)]
Use InterlockedExchange64 () for Interlocked.Exchange (long&, long).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:20:53 +0000 (20:20 +0200)]
Use InterlockedDecrement64 () for Interlocked.Decrement (long&).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:20:20 +0000 (20:20 +0200)]
Use InterlockedIncrement64 () for Interlocked.Increment (long&).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:19:18 +0000 (20:19 +0200)]
Use InterlockedRead64 () for Interlocked.Read (long&).
Alex Rønne Petersen [Wed, 25 Sep 2013 18:13:37 +0000 (20:13 +0200)]
Use __sync_sub_and_fetch () for atomic decrements for clarity.
Alex Rønne Petersen [Wed, 25 Sep 2013 18:00:46 +0000 (20:00 +0200)]
Move definition of NEED_64BIT_CMPXCHG_FALLBACK to top of atomic.c.
Alex Rønne Petersen [Wed, 25 Sep 2013 17:59:37 +0000 (19:59 +0200)]
Add interlocked read/write functions to atomic.c/.h.
These are guaranteed to have sequential consistency regardless
of the bitness of the machine they run on.
Alex Rønne Petersen [Wed, 25 Sep 2013 17:26:24 +0000 (19:26 +0200)]
Add implementations of various interlocked functions to atomic.c/.h.
Mainly 64-bit variants, but also InterlockedAdd ().
Alex Rønne Petersen [Wed, 25 Sep 2013 17:06:24 +0000 (19:06 +0200)]
Use mono_memory_barrier () for Thread.MemoryBarrier ().
This gets rid of the rather hacky lock-based barrier.
Alex Rønne Petersen [Wed, 25 Sep 2013 17:03:40 +0000 (19:03 +0200)]
Kill the Cygwin hack in mono-membar.h.
We only support GCC 4.x in Cygwin now.
Alex Rønne Petersen [Wed, 25 Sep 2013 17:00:57 +0000 (19:00 +0200)]
Remove MONO_ARCH_SAVE_REGS from threading icalls.
Alex Rønne Petersen [Tue, 24 Sep 2013 15:27:01 +0000 (17:27 +0200)]
Use GCC atomics on s390x.
Alex Rønne Petersen [Tue, 24 Sep 2013 15:20:13 +0000 (17:20 +0200)]
Remove unneeded ARM code from atomic.c.
Alex Rønne Petersen [Tue, 24 Sep 2013 15:16:17 +0000 (17:16 +0200)]
Use host preprocessor identifiers in atomic.c, not target.
Here we care about the system we run on, not the kind of
system that Mono is built to compile for.
Alex Rønne Petersen [Tue, 24 Sep 2013 15:14:55 +0000 (17:14 +0200)]
Simplify the ifdeffery in atomic.c/.h a bit.
Alex Rønne Petersen [Tue, 24 Sep 2013 14:59:11 +0000 (16:59 +0200)]
Remove unneeded NetBSD code path from atomic.h.
Marek Safar [Sat, 19 Oct 2013 15:34:26 +0000 (08:34 -0700)]
Merge pull request #783 from alistair/timezoneinfo_fix_for__convert_utc_to_local
Bug 15373: Return Local DateTime instead of Unspecified for >net_4_0
Alistair Bush [Sat, 19 Oct 2013 10:41:17 +0000 (23:41 +1300)]
Return Local DateTime instead of Unspecified for >net_4_0
Mark Probst [Fri, 18 Oct 2013 19:46:11 +0000 (12:46 -0700)]
[runtime] JIT debugging option to optimize only single methods.
The --single-method option run regression tests from an assembly, like
--regression. Instead of going through a list of optimizations,
however, it takes one optimization combination, runs all tests with
all optimizations turned off, while gathering a list of which methods
were jitted. Then, for each jitted method, it runs all tests again,
with the optimizations applied to only that one method, stopping as
soon as a test fails. Example:
mono --single-method=branch,peephole basic-float.exe
Zoltan Varga [Fri, 18 Oct 2013 20:23:54 +0000 (22:23 +0200)]
Disable a corlib test assertion which is no longer true.
Rodrigo Kumpera [Fri, 18 Oct 2013 18:09:11 +0000 (11:09 -0700)]
Merge pull request #768 from madewokherd/gcnotdll
Define GC_NOT_DLL for included boehm on win32.
Zoltan Varga [Fri, 18 Oct 2013 17:28:39 +0000 (19:28 +0200)]
[sgen] Set the TLS key offsets in one place in create_allocator ().
Zoltan Varga [Fri, 18 Oct 2013 17:25:09 +0000 (19:25 +0200)]
[aot] Add the JIT icall wrappers later, so it can pick up the JIT icalls registered by the managed allocators.
Zoltan Varga [Fri, 18 Oct 2013 16:32:28 +0000 (18:32 +0200)]
[runtime] Fix the build.
Zoltan Varga [Fri, 18 Oct 2013 16:14:25 +0000 (18:14 +0200)]
[runtime] Really fix the build.
Zoltan Varga [Fri, 18 Oct 2013 16:10:30 +0000 (18:10 +0200)]
[runtime] Fix the build.
Zoltan Varga [Fri, 18 Oct 2013 16:01:44 +0000 (18:01 +0200)]
[runtime] Add a missing file.
Zoltan Varga [Fri, 18 Oct 2013 15:54:01 +0000 (17:54 +0200)]
[runtime] Clean up the usage of the CEE_MONO_TLS opcode in wrappers. Instead of passing it the tls offset, pass one of the TLS_KEY_ constants defined in mono-tls.h, and resolve the offset in the JIT. This allows these wrappers to work with AOT.
Zoltan Varga [Fri, 18 Oct 2013 09:22:32 +0000 (02:22 -0700)]
Merge pull request #781 from directhex/master
SPARC build fixes
Jo Shields [Fri, 18 Oct 2013 08:36:00 +0000 (09:36 +0100)]
Define REG_SP on SPARC in sgen-archdep.h, if needed.
This definition is stolen from exceptions-sparc.c
Jonathan Lima [Thu, 26 Sep 2013 16:56:16 +0000 (13:56 -0300)]
Fix MSVC build.
Jo Shields [Fri, 18 Oct 2013 08:07:52 +0000 (09:07 +0100)]
Add missing ; characters to mono-hwcap-sparc.c. Fixes build failure.
Atsushi Eno [Fri, 18 Oct 2013 07:52:08 +0000 (00:52 -0700)]
Merge pull request #762 from echampet/wsdl
[FIX] Bug 267: wsdl fails with "Argument is out of range."
Atsushi Eno [Fri, 18 Oct 2013 07:49:51 +0000 (00:49 -0700)]
Merge pull request #766 from remobjects/resx-delete-on-clean
When generating resources, it intends to add the .resources file to the ...
Atsushi Eno [Fri, 18 Oct 2013 07:48:26 +0000 (00:48 -0700)]
Merge pull request #767 from ovatsus/Bug14922
Fix for #14922
Zoltan Varga [Thu, 17 Oct 2013 09:53:17 +0000 (11:53 +0200)]
Fill out AssemblyName.ProcessorArchitecture for assemblies. Fixes #15347.
Marek Safar [Wed, 16 Oct 2013 18:21:18 +0000 (20:21 +0200)]
Better check for open generic types in probing expression. Fixes #15422
Marek Safar [Wed, 16 Oct 2013 17:17:25 +0000 (19:17 +0200)]
Fixes too early optimization of binary &&/& bool constants. Fixes part of #15392
Martin Baulig [Wed, 16 Oct 2013 09:24:57 +0000 (11:24 +0200)]
Dummy test commit.
Mark Probst [Tue, 15 Oct 2013 23:47:18 +0000 (16:47 -0700)]
[sgen] Don't crash on nurseries of 512M or larger.
Martin Baulig [Tue, 15 Oct 2013 18:30:29 +0000 (20:30 +0200)]
[PCL]: Update the PCL Facade Assemblies.
Zoltan Varga [Tue, 15 Oct 2013 15:01:24 +0000 (17:01 +0200)]
[aot] Only use an AOT shortcut in the trampoline code if its safe to do so. Fixes #15345.
Martin Baulig [Tue, 15 Oct 2013 12:49:49 +0000 (14:49 +0200)]
Makefile magic ...
Marek Safar [Mon, 14 Oct 2013 17:17:23 +0000 (19:17 +0200)]
Marek Safar [Mon, 14 Oct 2013 07:32:13 +0000 (00:32 -0700)]
Merge pull request #779 from LogosBible/UTF8Encoding
Fixing bugs in UTF8Encoding's GetMaxByteCount and GetMaxCharCount.
Martin Potter [Sun, 13 Oct 2013 00:19:52 +0000 (17:19 -0700)]
Fixing bugs in UTF8Encoding's GetMaxByteCount and GetMaxCharCount.
The previous implementation of GetMaxByteCount has two bugs: it incorrectly took into account the byte order mark and it was taking the max of a byte count (3) and a character count (EncoderFallback.MaxCharCount). This could result in returning the max number of bytes required to convert a string that was less than the number bytes used when converting the string.
GetMaxCharCount did not previously take into account DecoderFallback.MaxCharCount or the fact that there may still be part of a surrogate pair left on the buffer from a previous call to convert.
I removed the #if NET_2_0 conditionals and NotWorking category since these methods now return results consistent with Microsoft's implementation. I also added two simple tests for custom EncoderFallback and DecoderFallback cases.
This commit is licensed under MIT/X11.
Jeffrey Stedfast [Sat, 12 Oct 2013 22:07:27 +0000 (15:07 -0700)]
Merge pull request #774 from DavidS/patch-1
Minor fix to avoid ArgumentOutOfRangeException
Zoltan Varga [Sat, 12 Oct 2013 13:17:59 +0000 (15:17 +0200)]
[runtime] Run the tests in mono/tests in sorted order.
Zoltan Varga [Sat, 12 Oct 2013 00:49:10 +0000 (02:49 +0200)]
[xdebug] Avoid using a MonoDebugMethodJitInfo structure after it has been freed. Fixes #15090.
Sebastien Pouliot [Fri, 11 Oct 2013 20:44:59 +0000 (16:44 -0400)]
[mono-api-html] Several indexer properties (Item) can have the name (with different parameters)
Sebastien Pouliot [Fri, 11 Oct 2013 15:05:02 +0000 (11:05 -0400)]
[mono-api-html] Output nested type correctly (was missing the declaring type name)
Marek Safar [Fri, 11 Oct 2013 13:28:54 +0000 (15:28 +0200)]
The first definite assignment cleanup
Mark Probst [Thu, 10 Oct 2013 23:54:26 +0000 (16:54 -0700)]
[regalloc2] Handle sreg3.
Mark Probst [Thu, 10 Oct 2013 23:47:33 +0000 (16:47 -0700)]
[regalloc2] Lazily allocate intervals.
Rodrigo Kumpera [Thu, 10 Oct 2013 18:44:45 +0000 (11:44 -0700)]
Merge pull request #777 from nealef/master
Remove debug message
Neale Ferguson [Thu, 10 Oct 2013 18:37:43 +0000 (14:37 -0400)]
Remove debug message
Rodrigo Kumpera [Thu, 10 Oct 2013 18:26:54 +0000 (11:26 -0700)]
Merge pull request #776 from nealef/master
Fix memory leaks in Oracle Client. The largest leaks come from the curso...
Neale Ferguson [Thu, 10 Oct 2013 17:17:17 +0000 (13:17 -0400)]
Fix memory leaks in Oracle Client. The largest leaks come from the cursors being created but never freed in OracleParameter.cs. The size of the memory used by these cursors varies from tens to thousands of bytes depending on the statement being executed. Several hours of transactions can lead to OOM errors. The other fixes are relatively minor but they do clean up other unmanaged memory not being returned.
Aaron Bockover [Thu, 10 Oct 2013 16:35:02 +0000 (12:35 -0400)]
bxc#13642: MacNetworkChange: implement using managed code
The previous version used SCNetworkReachability
in C inside libMonoPosixHelper. This was less
than ideal and made it more problematic for
supporting iOS as it doesn't actually link
MPH in.
Rewrite using P/Invokes directly into the
SCNetworkReachability APIs for Mac OS X and
iOS.
Fixes https://bugzilla.xamarin.com/show_bug.cgi?id=13642
for iOS and Mac OS X.
Marek Safar [Thu, 10 Oct 2013 14:05:59 +0000 (16:05 +0200)]
Compound assignment of optimized binary constant needs to decompose it first. Fixes #15315
Marek Safar [Thu, 10 Oct 2013 10:10:06 +0000 (12:10 +0200)]
Better debugger visualization for .net 1.1 style custom collections
Sebastien Pouliot [Thu, 10 Oct 2013 12:11:50 +0000 (08:11 -0400)]
[mono-api-html] Output C# 'char' instead of 'System.Char'
Sebastien Pouliot [Thu, 10 Oct 2013 12:10:44 +0000 (08:10 -0400)]
[mono-api-html] Do not produce (near-)empty files when no (public) changes are detected
Marek Safar [Thu, 10 Oct 2013 09:22:03 +0000 (11:22 +0200)]
Ensure that CodeDom generated temporary directory is always created. Fixes #12202
Sebastien Pouliot [Thu, 10 Oct 2013 00:59:21 +0000 (20:59 -0400)]
[mono-api-html] Show 'static' fields as such
Sebastien Pouliot [Thu, 10 Oct 2013 00:30:58 +0000 (20:30 -0400)]
Process/compare/output method's generic parameters [#15311]
Marek Safar [Wed, 9 Oct 2013 18:00:33 +0000 (20:00 +0200)]
Anonymous type are shared across module hence need module counters.
Rodrigo Kumpera [Wed, 9 Oct 2013 16:58:44 +0000 (12:58 -0400)]
Fix net_2_0 test.
Jonathan Pryor [Wed, 9 Oct 2013 15:07:05 +0000 (08:07 -0700)]
Merge pull request #775 from directhex/fix-gendarme-doc-generation
[monodoc] Ensure we don't try to use <Parameters/> when it might be null. Fixes Gendarme doc generation.
Jo Shields [Wed, 9 Oct 2013 14:02:35 +0000 (16:02 +0200)]
[monodoc] Support fields that match operator naming convention
Jo Shields [Wed, 9 Oct 2013 13:57:47 +0000 (15:57 +0200)]
Jo Shields [Wed, 9 Oct 2013 13:31:23 +0000 (15:31 +0200)]
[monodoc] Ensure we don't try to use <Parameters> when it might be null. Fixes Gendarme doc generation.
Gustavo Guerra [Wed, 9 Oct 2013 12:38:35 +0000 (13:38 +0100)]
Adjust defines to check for 4_0 instead of 3_0
Marek Safar [Wed, 9 Oct 2013 12:27:19 +0000 (14:27 +0200)]
anonymous: move counters to TypeContainer
With the anonymous counters on ModuleContainer,
anonymous method and type names will be generated
in a way that depends on the order of the sources
passed to mcs. This can cause undesirable API
and IL differences.
Moving the counters to TypeContainer instead
avoids this issue.
Marek Safar [Wed, 9 Oct 2013 12:04:47 +0000 (14:04 +0200)]
Enum IConvertible can convert to same enum or System.Enum. Fixes #15289
Marek Safar [Wed, 9 Oct 2013 11:52:10 +0000 (13:52 +0200)]
Cleanup redundant code
Marek Safar [Wed, 9 Oct 2013 10:51:16 +0000 (12:51 +0200)]
Simplify work item registration for internal work items
Marek Safar [Wed, 9 Oct 2013 09:39:14 +0000 (11:39 +0200)]
Remove redundant initialization
Marek Safar [Wed, 9 Oct 2013 09:26:34 +0000 (11:26 +0200)]
Allow to queue multicast delegates in thread-pool
Marek Safar [Wed, 9 Oct 2013 08:43:41 +0000 (10:43 +0200)]
Release continuation ancestor before running. Fixes #15036
David Schmitt [Wed, 9 Oct 2013 08:36:37 +0000 (10:36 +0200)]
Minor fix to avoid ArgumentOutOfRangeException
When the file is empty, data[0] would throw and swallow the original exception.
Marek Safar [Tue, 8 Oct 2013 17:30:32 +0000 (19:30 +0200)]
Adds async type inference of void return type. Fixes #15238
Marek Safar [Tue, 8 Oct 2013 08:15:21 +0000 (10:15 +0200)]
Make Task.Delay more reliable. Fixes #14585
Jeffrey Stedfast [Mon, 7 Oct 2013 21:51:17 +0000 (17:51 -0400)]
[System.ComponentModel.DataAnnotations] Implemented EmailAddressAttribute.IsValid()
Atsushi Eno [Mon, 7 Oct 2013 16:57:17 +0000 (01:57 +0900)]
Element() and Elements() could also reduce use of extraneous iterators.
Jonathan Pryor [Mon, 7 Oct 2013 16:27:30 +0000 (12:27 -0400)]
Fallback to $LANG if $LC_ALL isn't set.
Fixes: https://bugzilla.xamarin.com/show_bug.cgi?id=15214
This is a regression introduced in
be532638.
Xamarin.Android assumes that the $LANG environment variable can be
used to control System.Globalization.CultureInfo.CurrentCulture,
setting LANG based on the current Java locale.
This was broken in
be532638 because $LANG was no longer being checked
to determine the default locale, as $LC_ALL was not set.
Fix the fallback logic so that $LANG is checked when $LC_ALL is not.
Atsushi Eno [Mon, 7 Oct 2013 16:25:51 +0000 (01:25 +0900)]
reduce memory use on Attribute() caused by Attributes().
Marek Safar [Mon, 7 Oct 2013 15:04:13 +0000 (17:04 +0200)]
Inline more is instance checks