mono.git
8 years ago[tests] Xamarin.Android works the same way as the desktop Mono
Marek Habersack [Tue, 7 Jul 2015 09:12:37 +0000 (11:12 +0200)]
[tests] Xamarin.Android works the same way as the desktop Mono

8 years ago[tests] Adjust test to work in unsupported scenarios
Marek Habersack [Fri, 3 Jul 2015 10:27:45 +0000 (12:27 +0200)]
[tests] Adjust test to work in unsupported scenarios

On Android (possibly on other systems too) it is possible that no gateway address is available and its lack is NOT an error
Here is a sample of /proc/net/route from Nexus 9 running Android 5.1.1 (IPInterfaceProperties parses that file on Linux)

  Iface Destination Gateway  Flags RefCnt Use Metric Mask MTU Window IRTT
  wlan0 0001A8C0 00000000 0001 0 0 0 00FFFFFF 0 0 0

Gateway is set to any address and it is explicitly ignored by the route information parser

For comparison, here's route contents from an Android 4.4.4 device:

  Iface Destination Gateway  Flags RefCnt Use Metric Mask MTU Window IRTT
  wlan0 00000000 0101A8C0 0003 0 0 0 00000000 0 0 0
  wlan0 00000000 0101A8C0 0003 0 0 203 00000000 0 0 0
  wlan0 0001A8C0 00000000 0001 0 0 0 00FFFFFF 0 0 0
  wlan0 0001A8C0 00000000 0001 0 0 0 00FFFFFF 0 0 0
  wlan0 0001A8C0 00000000 0001 0 0 203 00FFFFFF 0 0 0
  wlan0 0101A8C0 00000000 0005 0 0 0 FFFFFFFF 0 0 0

Obviously, this test fails on the first device and succeeds on the second. For this reason the test is modified to succeed
in case of devices like the first one since it's not a real failure but a shortcoming of the .NET API

8 years ago[tests] Fix tests which rely on temp dir location being a real dir
Marek Habersack [Thu, 2 Jul 2015 15:59:06 +0000 (17:59 +0200)]
[tests] Fix tests which rely on temp dir location being a real dir

Android M (that's where I ran into it) makes the path returned by
Path.GetTempPath () a symlink to the real directory used for temporary
files. When the tests change the directory the symlink is followed and
subsequent reads of the current directory name will return the real dir
name which doesn't match what was constructed using the value returned by
GetTempPath ()

8 years ago[tests] Fails on Android ARM64
Marek Habersack [Thu, 2 Jul 2015 15:15:53 +0000 (17:15 +0200)]
[tests] Fails on Android ARM64

The test works on all the other supported architectures but segfaults on
ARM64 (Nexus 9 running Lollipop 5.1.1)

8 years ago[test] Disable on Android for now
Marek Habersack [Thu, 25 Jun 2015 18:40:37 +0000 (20:40 +0200)]
[test] Disable on Android for now

The test throws the following exception

  MonoTests.System.Security.Cryptography.AesCfbTests.Roundtrip=System.Security.Cryptography.CryptographicException : Bad PKCS7 padding. Invalid length 236.
  at Mono.Security.Cryptography.SymmetricTransform.ThrowBadPaddingException (PaddingMode padding, Int32 length, Int32 position) <0x3bf26858 + 0x000bb> in <filename unknown>:0
  at Mono.Security.Cryptography.SymmetricTransform.FinalDecrypt (System.Byte[] inputBuffer, Int32 inputOffset, Int32 inputCount) <0x3bf264f8 + 0x0022f> in <filename unknown>:0
  at Mono.Security.Cryptography.SymmetricTransform.TransformFinalBlock (System.Byte[] inputBuffer, Int32 inputOffset, Int32 inputCount) <0x3bf219b0 + 0x00077> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.CfbTests.Decryptor (System.Security.Cryptography.SymmetricAlgorithm algo, System.Byte[] encdata) <0x3bf261f8 + 0x00221> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.CfbTests.CFB (System.Security.Cryptography.SymmetricAlgorithm algo, Int32 feedbackSize) <0x3bf20000 + 0x0011b> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.AesCfbTests.CFB (System.Security.Cryptography.SymmetricAlgorithm algo) <0x3ca53ef0 + 0x00043> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.CfbTests.ProcessPadding (System.Security.Cryptography.SymmetricAlgorithm algo) <0x3ca53e80 + 0x0004f> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.CfbTests.ProcessKeySizes (System.Security.Cryptography.SymmetricAlgorithm algo) <0x3bf3fc20 + 0x000b7> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.CfbTests.ProcessBlockSizes (System.Security.Cryptography.SymmetricAlgorithm algo) <0x3bf3f910 + 0x0007b> in <filename unknown>:0
  at MonoTests.System.Security.Cryptography.AesCfbTests.Roundtrip () <0x3bf3cf28 + 0x00073> in <filename unknown>:0
  at (wrapper managed-to-native) System.Reflection.MonoMethod:InternalInvoke (System.Reflection.MonoMethod,object,object[],System.Exception&)
  at System.Reflection.MonoMethod.Invoke (System.Object obj, BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) <0x3bf3cc10 + 0x0008f> in <filename unknown>:0

8 years ago[tests] Don't use Assembly.Location on Android
Marek Habersack [Thu, 25 Jun 2015 18:32:10 +0000 (20:32 +0200)]
[tests] Don't use Assembly.Location on Android

Assembly.Location will contain just the assembly file name without any path, which
is not what the test(s) expect. The reason is that the assemblies on Android live
in the APK, are embedded in the native code or don't exist at all (AOT)

8 years ago[tests] Android behaves the same as desktop Mono
Marek Habersack [Thu, 25 Jun 2015 09:25:20 +0000 (11:25 +0200)]
[tests] Android behaves the same as desktop Mono

The same exceptions are thrown in Xamarin.Android, unlike in Xamarin.iOS which
behaves differently to desktop Mono

8 years ago[tests] Disable for Android
Marek Habersack [Thu, 25 Jun 2015 09:16:17 +0000 (11:16 +0200)]
[tests] Disable for Android

The test works when only its fixture is ran, fails when ran as part of the
entire BCL test suite.

8 years ago[tests] Disable because URL is no longer there
Marek Habersack [Wed, 24 Jun 2015 21:49:09 +0000 (23:49 +0200)]
[tests] Disable because URL is no longer there

The test relies on presence of a specific file that is now gone. Disabled
until a stable URL is found :P

8 years ago[tests] Disable for Android
Marek Habersack [Wed, 24 Jun 2015 21:43:25 +0000 (23:43 +0200)]
[tests] Disable for Android

The test works when only its fixture is ran, fails when ran as part of the
entire BCL test suite.

8 years ago[tests] Disable for Android
Marek Habersack [Wed, 24 Jun 2015 21:37:06 +0000 (23:37 +0200)]
[tests] Disable for Android

The test works when only its fixture is ran, fails when ran as part of the
entire BCL test suite.

8 years ago[tests] Android doesn't have access to Assembly.Location
Marek Habersack [Wed, 24 Jun 2015 21:21:53 +0000 (23:21 +0200)]
[tests] Android doesn't have access to Assembly.Location

Attempts to access the test dll which won't work on Android since the
dll is either in the apk or bundled in the native code or not there at
all (AOT)

8 years ago[tests] Disable for Android
Marek Habersack [Wed, 24 Jun 2015 21:18:09 +0000 (23:18 +0200)]
[tests] Disable for Android

The tests work fine if only this fixture is ran, they fail when ran as part
of the entire test suite.

8 years ago[tests] Disable for Android
Marek Habersack [Wed, 24 Jun 2015 20:10:31 +0000 (22:10 +0200)]
[tests] Disable for Android

The tests work fine if only this fixture is ran, they fail when ran as part
of the entire test suite.

8 years ago[tests] System zlib errors out for these on Android
Marek Habersack [Wed, 24 Jun 2015 17:14:34 +0000 (19:14 +0200)]
[tests] System zlib errors out for these on Android

8 years ago[tests] Large locks not supported on Android
Marek Habersack [Wed, 24 Jun 2015 16:37:52 +0000 (18:37 +0200)]
[tests] Large locks not supported on Android

While bionic reports it supports 64-bit offsets for locks and exposes the
necessary APIs, attempt to use offsets larger than 32 bits results in
an error.

8 years ago[tests] The test hangs on Android
Marek Habersack [Wed, 24 Jun 2015 16:35:53 +0000 (18:35 +0200)]
[tests] The test hangs on Android

8 years ago[tests] The test hangs on Android
Marek Habersack [Wed, 24 Jun 2015 16:35:39 +0000 (18:35 +0200)]
[tests] The test hangs on Android

8 years ago[tests] Use xamarin web server
Marek Habersack [Fri, 12 Jun 2015 16:55:36 +0000 (18:55 +0200)]
[tests] Use xamarin web server

http://elmundo.es/ started to send more than one cookie while
xamarin.com sends just one.

8 years ago[tests] Don't assume specific charset
Marek Habersack [Fri, 12 Jun 2015 15:33:24 +0000 (17:33 +0200)]
[tests] Don't assume specific charset

Don't assume a specific charset in HTTP response from a server. Google
changes response charset based on the country of the client's
IP. There's no point in checking for the specific charset, it's enough
to check whether the response header has correct format.

8 years ago[tests] Different CPUs have different precision
Marek Habersack [Fri, 12 Jun 2015 15:23:15 +0000 (17:23 +0200)]
[tests] Different CPUs have different precision

Fix the test to compare the subtraction result to value more
conservative than Double.Epsilon, to account for differences in
different CPU floating point units on Android.

8 years ago[tests] Use correct path in process tests
Marek Habersack [Fri, 12 Jun 2015 13:55:24 +0000 (15:55 +0200)]
[tests] Use correct path in process tests

/bin/ls doesn't exist on Android, it's /system/bin/ls

8 years ago[test] There is no runtime directory on Android
Marek Habersack [Thu, 11 Jun 2015 18:15:50 +0000 (20:15 +0200)]
[test] There is no runtime directory on Android

We return an empty string which is considered to be a valid value in
Android. Adjust the test for that.

8 years ago[tests] Use 2.1 profile assembly names for mobile
Marek Habersack [Thu, 11 Jun 2015 18:15:01 +0000 (20:15 +0200)]
[tests] Use 2.1 profile assembly names for mobile

8 years ago[tests] Make failures more descriptive and fix XA
Marek Habersack [Thu, 11 Jun 2015 18:13:21 +0000 (20:13 +0200)]
[tests] Make failures more descriptive and fix XA

The test failures didn't really indicate which of the Fail statements
caused the test failure. Also, the 10k offset turned out to be not
enough for the XA test suite (it's much bigger than that)

8 years ago[tests] Assembly.Location doesn't work on XA
Marek Habersack [Thu, 11 Jun 2015 18:11:14 +0000 (20:11 +0200)]
[tests] Assembly.Location doesn't work on XA

Xamarin.Android stores assemblies inside apk (or bundled, or AOT-ed) and
so Assembly.Location will not point to a real path on the
filesystem. It merely contains assembly name.

8 years ago[tests] No Mono.CompilerServices.SymbolWriter in XA
Marek Habersack [Thu, 11 Jun 2015 18:09:45 +0000 (20:09 +0200)]
[tests] No Mono.CompilerServices.SymbolWriter in XA

8 years ago[tests] No declarative security in mobile profile
Marek Habersack [Thu, 11 Jun 2015 18:08:24 +0000 (20:08 +0200)]
[tests] No declarative security in mobile profile

8 years ago[tests] Fix for when CWD is /
Marek Habersack [Thu, 11 Jun 2015 18:02:03 +0000 (20:02 +0200)]
[tests] Fix for when CWD is /

The test case doesn't account for the situation when the tests are ran
at the root of the Unix filesystem. This is, admittedly, a rare case on
Desktop but it's precisely what CWD is for Android processes. The test
would fail because it would get a path of the //Form instead of the /Form

8 years ago[tests] Works in stand-alone app, fails in nunit
Marek Habersack [Thu, 11 Jun 2015 18:00:43 +0000 (20:00 +0200)]
[tests] Works in stand-alone app, fails in nunit

8 years ago[tests] Fails on mobile with no details
Marek Habersack [Thu, 11 Jun 2015 17:59:45 +0000 (19:59 +0200)]
[tests] Fails on mobile with no details

8 years ago[tests] Doesn't work on mobile
Marek Habersack [Thu, 11 Jun 2015 17:58:56 +0000 (19:58 +0200)]
[tests] Doesn't work on mobile

8 years ago[tests] Works in stand-alone app, fails in nunit
Marek Habersack [Thu, 11 Jun 2015 17:58:07 +0000 (19:58 +0200)]
[tests] Works in stand-alone app, fails in nunit

8 years ago[tests] Xamarin.Android has the desktop behavior
Marek Habersack [Thu, 11 Jun 2015 17:56:59 +0000 (19:56 +0200)]
[tests] Xamarin.Android has the desktop behavior

8 years ago[tests] Xamarin.Android already has the namespace
Marek Habersack [Thu, 11 Jun 2015 17:56:12 +0000 (19:56 +0200)]
[tests] Xamarin.Android already has the namespace

8 years ago[build] Update missing file
Marek Safar [Wed, 15 Jul 2015 16:06:11 +0000 (18:06 +0200)]
[build] Update missing file

8 years ago[build] Update profiles path
Marek Safar [Wed, 15 Jul 2015 15:57:01 +0000 (17:57 +0200)]
[build] Update profiles path

8 years ago[build] Update profiles path
Marek Safar [Wed, 15 Jul 2015 15:34:54 +0000 (17:34 +0200)]
[build] Update profiles path

8 years ago[Facades] More missing types
Marek Safar [Wed, 15 Jul 2015 15:30:20 +0000 (17:30 +0200)]
[Facades] More missing types

8 years ago[Facades] Update to 4.6 api
Marek Safar [Wed, 15 Jul 2015 14:35:59 +0000 (16:35 +0200)]
[Facades] Update to 4.6 api

8 years ago[build] Update docs profile paths
Marek Safar [Wed, 15 Jul 2015 14:34:56 +0000 (16:34 +0200)]
[build] Update docs profile paths

8 years ago[build] Switch to 4.6 profile as default profile
Marek Safar [Wed, 15 Jul 2015 14:10:19 +0000 (16:10 +0200)]
[build] Switch to 4.6 profile as default profile

8 years agoBump binary-reference-assemblies
Marek Safar [Wed, 15 Jul 2015 07:34:57 +0000 (09:34 +0200)]
Bump binary-reference-assemblies

8 years ago[xbuild]: Fix ResolveAssemblyReference.ResolveReference() with empty search paths.
Martin Baulig [Tue, 14 Jul 2015 21:38:40 +0000 (23:38 +0200)]
[xbuild]: Fix ResolveAssemblyReference.ResolveReference() with empty search paths.

When using shared projects, the search path and/or framework paths may sometimes
contain empty elements.

This fixes commit 5b12d9ae37ee34aad5f1ceada34cb36f11fb41ce.

8 years ago[llvm] Add support for pass/returning vtypes by ref.
Zoltan Varga [Tue, 14 Jul 2015 21:19:13 +0000 (17:19 -0400)]
[llvm] Add support for pass/returning vtypes by ref.

8 years ago[system] Add missing APIs
Marek Safar [Tue, 14 Jul 2015 18:07:40 +0000 (20:07 +0200)]
[system] Add missing APIs

8 years ago[threadpool-ms-io] Fix epoll socket close/remove race condition
Ludovic Henry [Fri, 10 Jul 2015 21:51:51 +0000 (18:51 -0300)]
[threadpool-ms-io] Fix epoll socket close/remove race condition

With the previous commit, it fixes a range of bug that happens only with epoll. This is because, contrary to poll and kqueue, epoll keeps an internal state that has to be kept in sync with the threadpool_io->states hash table.

This threadpool_io->states hash table is keeping a list of file descriptor to watch, with the corresponding SocketAsyncResult to callback in case of event on the fd. But epoll also keeps its internal state that needs to me modified with EPOLL_CTL_ADD, EPOLL_CTL_MOD and EPOLL_CTL_DEL.
Contrary to kqueue, EPOLLONESHOT does not force epoll to remove the socket from its internal state once an event have been triggered; but it will just not fire this event again, until this event is reset. Also, epoll will remove a fd from its internal state when this fd is closed.

This will lead to a bunch of out-of-sync states like the following :
 - the user start reading asynchronously the STDOUT of a child process, the child process quit (closing the STDOUT fd), this fd is reused by the OS for something else (like a socket), the IO threadpool call epoll_ctl with EPOLL_CTL_MOD (because the fd is still in the states hashtable), but the OS returns an error because the underlying file is different, even if the fd is the same.
 - the user opens a socket, does some asynchronous operation on it, and close it; removing it from the table but not from epoll (as it does not call epoll_ctl with EPOLL_CTL_DEL). For another reason, the fd has not been closed in the OS (and thus in epoll), because it might, for example, still be open by another process. After that, we receive an event for this fd (as it is still in epoll), and between the time of the epoll_wait and the epoll_ctl with EPOLL_CTL_DEL, the fd is actually closed in the OS, leading to a EBADF error.

The g_warning have also been replaced by g_error, so we do not silently discard this kind of error in testing.

8 years ago[bcl] Remove ProcessAsyncReader handle from IO ThreadPool on closing
Ludovic Henry [Mon, 13 Jul 2015 17:54:30 +0000 (14:54 -0300)]
[bcl] Remove ProcessAsyncReader handle from IO ThreadPool on closing

8 years ago[threadpool-ms-io] Rework some utility functions
Ludovic Henry [Fri, 10 Jul 2015 19:16:08 +0000 (16:16 -0300)]
[threadpool-ms-io] Rework some utility functions

8 years ago[threadpool-ms-io] Replace the 2 locks by only 1
Ludovic Henry [Fri, 10 Jul 2015 19:15:39 +0000 (16:15 -0300)]
[threadpool-ms-io] Replace the 2 locks by only 1

8 years ago[runtime] Fix the mono_get_lmf_addr () optimization for x86-ios. Fixes #31921.
Zoltan Varga [Tue, 14 Jul 2015 17:42:51 +0000 (13:42 -0400)]
[runtime] Fix the mono_get_lmf_addr () optimization for x86-ios. Fixes #31921.

8 years ago[utils] Factor lazy initialization of threadpool-ms and threadpool-ms-io for use...
Ludovic Henry [Mon, 22 Jun 2015 17:14:27 +0000 (14:14 -0300)]
[utils] Factor lazy initialization of threadpool-ms and threadpool-ms-io for use by other modules

8 years agoDisable the test added by 37ad145b2f00beb3e1aa9b86e306fbcbc50df3fe, it doesn't work...
Zoltan Varga [Tue, 14 Jul 2015 13:00:40 +0000 (09:00 -0400)]
Disable the test added by 37ad145b2f00beb3e1aa9b86e306fbcbc50df3fe, it doesn't work on all platforms.

8 years ago[ilasm] Imlicitly set enums value__ special flags
Marek Safar [Tue, 14 Jul 2015 09:31:20 +0000 (11:31 +0200)]
[ilasm] Imlicitly set enums value__ special flags

8 years agoLook for resgen2 where mcs is.
Rolf Bjarne Kvinge [Fri, 26 Jun 2015 14:44:18 +0000 (16:44 +0200)]
Look for resgen2 where mcs is.

Fixes an issue on El Capitán where resgen2 isn't necessarily in PATH.

8 years agoMerge pull request #1730 from BrzVlad/feature-arm-fast-tls
Vlad Brezae [Tue, 14 Jul 2015 02:42:56 +0000 (19:42 -0700)]
Merge pull request #1730 from BrzVlad/feature-arm-fast-tls

Feature arm fast tls

8 years ago[threadpool-ms] Add tracing
Ludovic Henry [Mon, 13 Jul 2015 23:22:04 +0000 (20:22 -0300)]
[threadpool-ms] Add tracing

8 years ago[ilasm] Make only structs automatically sealed (partially reverts 541a65ddd3f4b49c4d5...
Marek Safar [Mon, 13 Jul 2015 21:14:43 +0000 (23:14 +0200)]
[ilasm] Make only structs automatically sealed (partially reverts 541a65ddd3f4b49c4d5bb6324bae63f0ed569115)

8 years ago[ilasm] Don't make enums and structs sealed when sealed modifier is not used
Marek Safar [Mon, 13 Jul 2015 20:37:18 +0000 (22:37 +0200)]
[ilasm] Don't make enums and structs sealed when sealed modifier is not used

8 years ago[jit] Fix float32 support for the interlocked opcodes.
Zoltan Varga [Mon, 13 Jul 2015 19:14:24 +0000 (15:14 -0400)]
[jit] Fix float32 support for the interlocked opcodes.

8 years ago[amd64] Fix marshalling of empty structures.
Zoltan Varga [Mon, 13 Jul 2015 17:56:14 +0000 (13:56 -0400)]
[amd64] Fix marshalling of empty structures.

8 years ago[System] Don't assume WebConnectionStream of unknown http methods require write strea...
Marek Safar [Mon, 13 Jul 2015 12:08:08 +0000 (14:08 +0200)]
[System] Don't assume WebConnectionStream of unknown http methods require write stream. Fixes #31830

8 years agoMerge pull request #1927 from borgdylan/x86fix
Zoltan Varga [Mon, 13 Jul 2015 07:52:18 +0000 (03:52 -0400)]
Merge pull request #1927 from borgdylan/x86fix

fix the x86 build

8 years agofix the x86 build
Dylan Borg [Mon, 13 Jul 2015 07:18:03 +0000 (09:18 +0200)]
fix the x86 build

8 years ago[runtime] Add support for portable pdb files without a MODULE table. Fix mono_ppdb_ge...
Zoltan Varga [Sun, 12 Jul 2015 20:34:45 +0000 (16:34 -0400)]
[runtime] Add support for portable pdb files without a MODULE table. Fix mono_ppdb_get_seq_points ().

8 years ago[runtime] Make mono_trace_set_printerr_handler () call g_set_printerr_handler ()...
Zoltan Varga [Sun, 12 Jul 2015 16:37:33 +0000 (12:37 -0400)]
[runtime] Make mono_trace_set_printerr_handler () call g_set_printerr_handler (). Fixes #31451.

8 years ago[runtime] Update portable pdb support for the latest spec.
Zoltan Varga [Sun, 12 Jul 2015 04:14:36 +0000 (00:14 -0400)]
[runtime] Update portable pdb support for the latest spec.

8 years ago[aot] Emit the got into the bss section on osx too.
Zoltan Varga [Fri, 10 Jul 2015 19:01:41 +0000 (15:01 -0400)]
[aot] Emit the got into the bss section on osx too.

8 years ago[tests] Add tls stress test
Vlad Brezae [Fri, 1 May 2015 18:55:54 +0000 (11:55 -0700)]
[tests] Add tls stress test

8 years ago[arm] Add MONO_DEBUG option for using only fallback tls.
Vlad Brezae [Sat, 25 Apr 2015 00:23:20 +0000 (17:23 -0700)]
[arm] Add MONO_DEBUG option for using only fallback tls.

8 years ago[arm] Add fast tls support on ios.
Vlad Brezae [Sat, 4 Apr 2015 00:00:11 +0000 (17:00 -0700)]
[arm] Add fast tls support on ios.

8 years ago[sgen] Use the sgen allocator fastpath also when cross compilling.
Vlad Brezae [Wed, 22 Apr 2015 21:39:35 +0000 (14:39 -0700)]
[sgen] Use the sgen allocator fastpath also when cross compilling.

The platform needs to support tls for it to be generated.

8 years ago[sgen] Fix allocator creation when running with jit disabled.
Vlad Brezae [Fri, 1 May 2015 19:27:15 +0000 (12:27 -0700)]
[sgen] Fix allocator creation when running with jit disabled.

We don't allocate the method header when not jitting.

8 years ago[arm] Add fast tls support on android.
Vlad Brezae [Tue, 7 Apr 2015 21:39:38 +0000 (21:39 +0000)]
[arm] Add fast tls support on android.

8 years ago[sgen] Allocator defines depend on the target not the host.
Vlad Brezae [Fri, 3 Apr 2015 00:29:07 +0000 (17:29 -0700)]
[sgen] Allocator defines depend on the target not the host.

8 years ago[arm] Add fast tls support on linux
Vlad Brezae [Wed, 4 Feb 2015 23:37:33 +0000 (15:37 -0800)]
[arm] Add fast tls support on linux

Add support for OP_TLS operations. The code that implements this is comprised of getter and setter asm thunks, which is either inlined, or it uses external code. At runtime we check the implementation on the system, using the inlined thunks if we can.

8 years ago[mcs] Emit symbol info local variables for compiler generated storey variables. Fixes...
Marek Safar [Fri, 10 Jul 2015 18:06:51 +0000 (20:06 +0200)]
[mcs] Emit symbol info local variables for compiler generated storey variables. Fixes #31784

8 years ago[mcs] Remove NET_4_0 ifdefs
Marek Safar [Fri, 10 Jul 2015 15:25:42 +0000 (17:25 +0200)]
[mcs] Remove NET_4_0 ifdefs

8 years ago[jit] Use a feature macro instead of hardcoding users.
Rodrigo Kumpera [Fri, 10 Jul 2015 17:36:15 +0000 (13:36 -0400)]
[jit] Use a feature macro instead of hardcoding users.

8 years agoRemove duplicated code.
Rodrigo Kumpera [Fri, 10 Jul 2015 16:26:36 +0000 (12:26 -0400)]
Remove duplicated code.

8 years ago[arm] Fixed the max length of seq_point. Fixes #31829.
Rodrigo Kumpera [Fri, 10 Jul 2015 16:20:04 +0000 (12:20 -0400)]
[arm] Fixed the max length of seq_point. Fixes #31829.

8 years ago[jit] Disable fast tls on osx if MONO_HAVE_FAST_TLS is not defined.
Rodrigo Kumpera [Fri, 10 Jul 2015 16:00:17 +0000 (12:00 -0400)]
[jit] Disable fast tls on osx if MONO_HAVE_FAST_TLS is not defined.

8 years ago[utils] Disable fast TLS on darwin until we have proper fallbacks in place.
Rodrigo Kumpera [Fri, 10 Jul 2015 14:07:26 +0000 (10:07 -0400)]
[utils] Disable fast TLS on darwin until we have proper fallbacks  in place.

8 years ago[mini] Add some extra asserts around fast tls to ensure we don't mess up.
Rodrigo Kumpera [Fri, 10 Jul 2015 14:06:40 +0000 (10:06 -0400)]
[mini] Add some extra asserts around fast tls to ensure we don't mess up.

8 years ago[utils] Don't abort at startup if remote TLS reads is not available.
Rodrigo Kumpera [Thu, 9 Jul 2015 14:02:57 +0000 (10:02 -0400)]
[utils] Don't abort at startup if remote TLS reads is not available.

Remote TLS reads is not a required feature for the runtime to work on
mach systems. We still want it available, but can leave the assertion
to happen when someone actually use it.

Usage won't happen today as part of regular code execution, so moving
the assert will make us more robust.

8 years ago[utils] Add fallback TLS offset scanning to X86 and AMD64.
Rodrigo Kumpera [Thu, 9 Jul 2015 13:33:52 +0000 (09:33 -0400)]
[utils] Add fallback TLS offset scanning to X86 and AMD64.

8 years agoMerge pull request #1922 from antonmes/master
Rodrigo Kumpera [Fri, 10 Jul 2015 13:48:54 +0000 (09:48 -0400)]
Merge pull request #1922 from antonmes/master

Add TLS offset for pre-release OS X 10.11 on x86

8 years agoMerge pull request #1924 from bosmacs/master
João Matos [Fri, 10 Jul 2015 12:12:46 +0000 (13:12 +0100)]
Merge pull request #1924 from bosmacs/master

Add TLS offset for pre-release OS X 10.11 x86, as #1919 did for x64

8 years agoAdd TLS offset for pre-release OS X 10.11 x86, as in #1919 for x64
Craig S. Bosma [Fri, 10 Jul 2015 11:56:23 +0000 (06:56 -0500)]
Add TLS offset for pre-release OS X 10.11 x86, as in #1919 for x64

8 years agoAdd TLS offset for pre-release OS X 10.11 on x86
Toxa [Fri, 10 Jul 2015 05:32:54 +0000 (08:32 +0300)]
Add TLS offset for pre-release OS X 10.11 on x86

8 years ago[llvm] Remove the restriction on handlers without invokes, its not needed any more.
Zoltan Varga [Thu, 9 Jul 2015 19:35:30 +0000 (15:35 -0400)]
[llvm] Remove the restriction on handlers without invokes, its not needed any more.

8 years ago[jit] Fix a regression introduced by ce567f4026d0b0c22469dbb54f6bf0f0de18b6f2. cfg...
Zoltan Varga [Thu, 9 Jul 2015 19:32:53 +0000 (15:32 -0400)]
[jit] Fix a regression introduced by ce567f4026d0b0c22469dbb54f6bf0f0de18b6f2. cfg->flags was initialized with a JitFlags value.

8 years ago[llvm] Fix the handling of nested clauses.
Zoltan Varga [Thu, 9 Jul 2015 19:22:54 +0000 (15:22 -0400)]
[llvm] Fix the handling of nested clauses.

8 years agoMerge pull request #1888 from joelmartinez/mdoc-flagsbug
Jonathan Pryor [Thu, 9 Jul 2015 16:26:26 +0000 (12:26 -0400)]
Merge pull request #1888 from joelmartinez/mdoc-flagsbug

[mdoc] Adds custom formatting for the `ObjCRuntime.Platform` enum.

8 years agoBump reference sources
Marek Safar [Thu, 9 Jul 2015 16:01:17 +0000 (18:01 +0200)]
Bump reference sources

8 years ago[System.Net.Http] Reading ContentLength value actually set header property value...
Marek Safar [Thu, 9 Jul 2015 15:46:53 +0000 (17:46 +0200)]
[System.Net.Http] Reading ContentLength value actually set header property value. Fixes #30561

8 years agoMerge pull request #1919 from bosmacs/master
João Matos [Thu, 9 Jul 2015 00:19:07 +0000 (01:19 +0100)]
Merge pull request #1919 from bosmacs/master

Add TLS offset for pre-release OS X 10.11 on amd64

8 years agoAdd TLS offset for pre-release OS X 10.11 on amd64
Craig S. Bosma [Thu, 9 Jul 2015 00:01:55 +0000 (19:01 -0500)]
Add TLS offset for pre-release OS X 10.11 on amd64

8 years ago[sdb] Allow boxed vtypes as this when invoking vtype methods. Fixes #30023.
Zoltan Varga [Wed, 8 Jul 2015 21:09:08 +0000 (17:09 -0400)]
[sdb] Allow boxed vtypes as this when invoking vtype methods. Fixes #30023.

8 years ago[mdoc] Adds custom formatting for the `ObjCRuntime.Platform` enum.
Joel Martinez [Fri, 19 Jun 2015 21:20:30 +0000 (17:20 -0400)]
[mdoc] Adds custom formatting for the `ObjCRuntime.Platform` enum.

This enum, which can be found in Xamarin's iOS and Mac products use a custom encoding
scheme which resulted in invalid values being shown when processed with mdoc. This patch
adds a special case to watch for this particular enumeration value, and decodes the data
correctly.

Additionally, the code that formats the values for flags enumerations was refactored. Future extensibility of this feature can now be done by simply implementing a new `FlagsFormatter`, and applying the proper logic in the `MakeAttributesValueString` function.