Alexis Christoforides [Mon, 13 Feb 2017 08:55:16 +0000 (03:55 -0500)]
[MacSDK] Use full Mono version, incl. build number in build
Also add a VERSION file in the package.
Alexis Christoforides [Mon, 13 Feb 2017 10:32:56 +0000 (05:32 -0500)]
[MacSDK] Update MSBuild
Alexis Christoforides [Mon, 13 Feb 2017 04:40:37 +0000 (23:40 -0500)]
Move Mac package command to scripts so that it's available without configuring
Marek Safar [Mon, 13 Feb 2017 09:09:01 +0000 (10:09 +0100)]
[System.ServiceModel] Disable unreliable test
Marek Safar [Fri, 10 Feb 2017 15:13:25 +0000 (16:13 +0100)]
Bump roslyn to rc4
Ludovic Henry [Sun, 12 Feb 2017 16:10:42 +0000 (11:10 -0500)]
[finalizer] Suspend finalizer thread on after shutdown timeout (#4363)
Fixes bug https://bugzilla.xamarin.com/show_bug.cgi?id=52429
Zoltan Varga [Sat, 11 Feb 2017 04:41:24 +0000 (23:41 -0500)]
[runtime] Remove an obsolete option from --help=devel.
Alexander Köplinger [Fri, 10 Feb 2017 19:56:08 +0000 (20:56 +0100)]
[tools] Install pdb files for security tools (#4364)
We don't use the global executable.make install logic for the tools
in security so we missed updating this to copy .pdb's as well.
Bernhard Urban [Fri, 10 Feb 2017 07:38:07 +0000 (08:38 +0100)]
[exceptions] introduce suspend-on-native-crash that suspends on every native crash
not just `SIGSEGV`. rename `suspend-on-sigsegv` but keep it as an alias around.
Zoltan Varga [Fri, 10 Feb 2017 18:47:23 +0000 (13:47 -0500)]
[runtime] Avoid holding the init lock while calling the cctor in mono_runtime_class_init (), its not needed any more. (#4361)
monojenkins [Fri, 10 Feb 2017 16:47:59 +0000 (16:47 +0000)]
[msvc] Update csproj files
Alexis Christoforides [Fri, 10 Feb 2017 16:26:21 +0000 (11:26 -0500)]
Bump bockbuild
Marek Safar [Fri, 10 Feb 2017 15:03:16 +0000 (16:03 +0100)]
[System.Data] Add RESX_RESOURCE_STRING
Eberhard Beilharz [Thu, 9 Feb 2017 17:07:33 +0000 (18:07 +0100)]
[swf] Fix pasting of Unicode text
Previously pasting of Unicode text, e.g. from LibreOffice, resulted in
question marks being displayed.
This fixes Xamarin-7359.
Change-Id: I78ec4cb1bbbd0e14da91d06a6fc517bc5c91e135
Marek Safar [Tue, 17 Jan 2017 15:03:41 +0000 (16:03 +0100)]
[System] CodeDom from CoreFX
Alexis Christoforides [Fri, 10 Feb 2017 04:52:15 +0000 (23:52 -0500)]
Bump bockbuild
Alexis Christoforides [Fri, 10 Feb 2017 04:29:30 +0000 (23:29 -0500)]
"WIP (auto-committed by bockbuild)"
Alexis Christoforides [Fri, 10 Feb 2017 04:00:59 +0000 (23:00 -0500)]
Bump Bockbuild
Ludovic Henry [Fri, 10 Feb 2017 03:02:43 +0000 (22:02 -0500)]
[threads] Make OSEvent alertable to fix bug #51653 (#4347)
Alexis Christoforides [Fri, 10 Feb 2017 00:00:37 +0000 (19:00 -0500)]
Bump bockbuild
Ludovic Henry [Thu, 9 Feb 2017 19:57:59 +0000 (14:57 -0500)]
[process] Improve error message for inaccessible process (#4354)
Fixes bug https://bugzilla.xamarin.com/show_bug.cgi?id=51561
Alexander Köplinger [Thu, 9 Feb 2017 17:30:53 +0000 (18:30 +0100)]
[Mono.Security] Check NewLocalMachinePath too for deciding if store is machine store (#4357)
The code in X509Store only checked the legacy X509StoreManager.LocalMachinePath
when deciding whether CspProviderFlags.UseMachineKeyStore should be
set when dealing with private keys.
However, with BTLS we now also have X509StoreManager.NewLocalMachinePath.
This caused an issue/regression in the recent 4.8 builds that
started importing into the BTLS cert store when using cert-sync:
https://github.com/mono/mono/commit/
968c60a0f32c7a4c92568c3feda7503f99b98c94
Because cert-sync is ran at package installation time as root
on Linux we would erroneously not pass UseMachineKeyStore (because
the paths between legacy and BTLS store are different), which in turn
causes KeyPairPersistence to look into the user directory, e.g.
~/.config/.mono/keypairs. Since on a clean machine that directory
hierarchy doesn't exist, it goes on and creates it -> we now have
the ~/.config directory owned by root which means a normal user
application can't write to it anymore (like NuGet) and crashes.
The fix is to check for the BTLS store path as well so KeyPairPersistence
uses the machine path and ~/.config is not created on package installation.
See e.g. https://github.com/travis-ci/travis-ci/issues/7271
Marek Safar [Thu, 9 Feb 2017 16:12:43 +0000 (17:12 +0100)]
[build] Add simple target for easier corefx strings updating
Rodrigo Kumpera [Thu, 9 Feb 2017 06:11:25 +0000 (22:11 -0800)]
Merge pull request #4248 from Unity-Technologies/boehm-gc-alloc-fixed
Implement mono_gc_alloc_fixed/mono_gc_free_fixed on Boehm to match SGen
Ludovic Henry [Thu, 9 Feb 2017 03:29:03 +0000 (22:29 -0500)]
[process] Fix Process.GetProcessesByName (#4351)
To get a process by name, we need to access the other process ProcessName, but it can fails on non-win32 platforms. This failure wouldn't be caught in the referencesource implementation trigerring this bug.
The exception we would observe would be:
Unhandled Exception:
System.InvalidOperationException: Process has exited, so the requested information is not available.
at System.Diagnostics.Process.get_ProcessName () [0x0002d] in /Users/alexander/dev/mono/mcs/class/System/System.Diagnostics/Process.cs:336
at (wrapper remoting-invoke-with-check) System.Diagnostics.Process:get_ProcessName ()
at Program.Main (System.String[] args) [0x0003b] in <
9c838a66cdc44c81b38c974edcc9c029>:0
[ERROR] FATAL UNHANDLED EXCEPTION: System.InvalidOperationException: Process has exited, so the requested information is not available.
at System.Diagnostics.Process.get_ProcessName () [0x0002d] in /Users/alexander/dev/mono/mcs/class/System/System.Diagnostics/Process.cs:336
at (wrapper remoting-invoke-with-check) System.Diagnostics.Process:get_ProcessName ()
at Program.Main (System.String[] args) [0x0003b] in <
9c838a66cdc44c81b38c974edcc9c029>:0
Fixes bug https://bugzilla.xamarin.com/show_bug.cgi?id=52345
Zoltan Varga [Thu, 9 Feb 2017 03:25:18 +0000 (22:25 -0500)]
[runtime] Fix the reflected type of Delegate.Method if the method is an inherited virtual method. Fixes #39444. (#4353)
Ludovic Henry [Thu, 9 Feb 2017 02:12:44 +0000 (21:12 -0500)]
[threads] Allow resetting the name of a threadpool thread (#4350)
Fixes bug https://bugzilla.xamarin.com/show_bug.cgi?id=47221
Zoltan Varga [Wed, 8 Feb 2017 23:16:09 +0000 (18:16 -0500)]
[runtime] Add a cache for generic array helper wrappers to reduce the number created. This works because there are a lot of duplicates due to the fact that the interfaces passed to setup_generic_array_ifaces () have the same context. (#4343)
Zoltan Varga [Wed, 8 Feb 2017 23:11:53 +0000 (18:11 -0500)]
[jit] Remove the MONO_INIT_VARINFO macro, its only used in one place. (#4349)
Martin Baulig [Wed, 8 Feb 2017 16:12:02 +0000 (17:12 +0100)]
[btls]: Fix MonoBtlsX509Store.cs.AddTrustedRoots(). (#4340)
This internal method is only called from Mono.Btls.Interface.BtlsStore
and used by the web-tests.
Call MonoBtlsProvider.SetupCertificateStore(store) to make its
behavior consistent with the internal implementation.
Ludovic Henry [Wed, 8 Feb 2017 15:33:22 +0000 (10:33 -0500)]
Revert "[threads] Make OSEvent alertable to fix bug #51653" (#4346)
Ludovic Henry [Wed, 8 Feb 2017 15:01:08 +0000 (10:01 -0500)]
[threads] Make OSEvent alertable to fix bug #51653 (#4341)
Martin Baulig [Wed, 8 Feb 2017 13:33:47 +0000 (14:33 +0100)]
[btls]: Fix MonoBtlsProvider.SetupCertificateStore(MonoTlsSettings,bool). (#4337)
* When 'settings != null' and 'settings.CertificateSearchPaths == null',
then we add the default paths, including the 'settings.TrustAnchors'.
This applies to all platforms, including Android.
* In MonoBtlsContext, we can now simply call
MonoTlsProvider.SetupCertificateStore(MonoTlsSettings,bool) instead of
having a custom version of it.
Niklas Therning [Wed, 8 Feb 2017 10:18:55 +0000 (11:18 +0100)]
Merge pull request #4335 from ntherning/skip-unwind-info-functions-on-windows-with-disabled-jit
Skip unwind info functions on Windows when JIT is disabled
Niklas Therning [Wed, 8 Feb 2017 10:16:17 +0000 (11:16 +0100)]
Merge pull request #4333 from ntherning/Mono-Posix-Stdlib-do-not-clobber-Marshal.GetLastWin32Error
Make sure Stdlib.GetLastError() never clobbers Marshal.GetLastWin32Error()
Marek Safar [Tue, 7 Feb 2017 22:21:38 +0000 (23:21 +0100)]
Bump corefx
Zoltan Varga [Wed, 8 Feb 2017 03:03:43 +0000 (22:03 -0500)]
[lldb] Avoid a crash with the mono_tramp_info_register () call in load_aot_module ().
Zoltan Varga [Tue, 7 Feb 2017 23:35:20 +0000 (18:35 -0500)]
[sdb] Add an attach test, not yet enabled.
Marek Safar [Tue, 7 Feb 2017 18:54:18 +0000 (19:54 +0100)]
[xbuild] Add deprecation warning
Alexander Köplinger [Tue, 7 Feb 2017 21:50:36 +0000 (22:50 +0100)]
[ci] Temporarily use http url to download server until cert is fixed
Martin Baulig [Tue, 7 Feb 2017 21:24:02 +0000 (22:24 +0100)]
[btls]: MonoBtlsError.GetError(out string, out int) returns 0, not null. (#4338)
Ludovic Henry [Tue, 7 Feb 2017 20:09:47 +0000 (15:09 -0500)]
[threadpool] Fix hang on threadpool cleanup (#4330)
Waiting for the threadpool threads to exit on shutdown triggers a hand. The simplest solution is to not wait for them, and let the OS clean them up.
This is not an ideal solution, but this is to fix a months-long chase for this bug, and to release a less-buggy monthly release.
Zoltan Varga [Tue, 7 Feb 2017 19:35:50 +0000 (14:35 -0500)]
[runtime] Add a workaround for #50529 by locking around pthread_create()/pthread_join () calls so they can't happen at the same time. (#4332)
monojenkins [Tue, 7 Feb 2017 16:36:06 +0000 (16:36 +0000)]
[msvc] Update csproj files (#4336)
Niklas Therning [Tue, 7 Feb 2017 14:07:02 +0000 (15:07 +0100)]
Skip unwind info functions on Windows when JIT is disabled
mono_arch_code_chunk_new()/mono_arch_code_chunk_destroy() are not compiled in
on Windows when DISABLED_JIT is defined.
code_manager_chunk_new()/code_manager_chunk_destroy() in mini-runtime.c, which
use these functions, will however be compiled in on Windows regardless of
whether DISABLED_JIT is defined or not. This causes linking failures due to
missing mono_arch_code_chunk_new()/mono_arch_code_chunk_destroy() symbols when
building on Windows and DISABLE_JIT is defined.
Niklas Therning [Tue, 7 Feb 2017 14:00:34 +0000 (15:00 +0100)]
Merge pull request #4334 from ntherning/exclude-mono_w32socket_disconnect-on-non-desktop-uwp
Exclude mono_w32socket_disconnect() on unsupported Windows families
Niklas Therning [Tue, 7 Feb 2017 09:19:55 +0000 (10:19 +0100)]
Exclude mono_w32socket_disconnect() on unsupported Windows families
Marek Safar [Tue, 7 Feb 2017 09:22:01 +0000 (10:22 +0100)]
[corlib] Update TimeZoneTest ios simulator check
Niklas Therning [Tue, 7 Feb 2017 08:45:06 +0000 (09:45 +0100)]
Make sure Stdlib.GetLastError() never clobbers Marshal.GetLastWin32Error()
Niklas Therning [Mon, 6 Feb 2017 12:53:53 +0000 (13:53 +0100)]
Temp fix for broken TimeZoneInfo on Windows families without registry access
Band aid fix until we can switch to CoreRT's implementation.
Andi McClure [Mon, 6 Feb 2017 22:33:48 +0000 (17:33 -0500)]
Remove comment from Thread.cs describing field removed in 4.8
Martin Baulig [Mon, 6 Feb 2017 22:27:59 +0000 (23:27 +0100)]
Remove debugging spew in MonoBtlsX509LookupMonoCollection.cs.
Zoltan Varga [Mon, 6 Feb 2017 22:25:53 +0000 (17:25 -0500)]
[arm] Use TARGET_IOS instead of __APPLE__ so we don't use the ios abi when cross compiling from osx to android.
Martin Baulig [Mon, 6 Feb 2017 22:18:21 +0000 (23:18 +0100)]
[btls]: Add option to specify the certificate search paths. (#4318)
* [btls]: Add option to specify the certificate search paths.
* MonoTlsSettings.CertificateSearchPaths: new property.
Valid values are
- @default - add default search paths
- @user - add default user search path
- @machine - add default machine search path
- @trust - add 'MonoTlsSettings.TrustedRoots'
- @pem:<directory> - custom directory containing certificates in PEM format
- @der:<directory> - same, in DER format
* Make it internal for the moment.
* Cosmetic; use "?.".
Vlad Brezae [Mon, 6 Feb 2017 19:20:21 +0000 (21:20 +0200)]
[sgen] Don't assert when suspending detaching thread (#4292)
This means that the thread will be part of the stw (not skipped), so if the stack pointer is not between the expected ranges we dont scan the thread. In other parts of the runtime, suspending here is handled by the suspend_can_continue flag.
Alexander Köplinger [Mon, 6 Feb 2017 11:32:48 +0000 (12:32 +0100)]
[tests] Fix duplicate variable setting in assemblyresolve
Probably bad merge conflict resolution in
5d054b1852f24b620e5800c06141201602e69d0e
Alexander Kyte [Mon, 6 Feb 2017 06:12:54 +0000 (01:12 -0500)]
[runtime] Get mono/tests running on bitcode/arm (#4233)
Alexander Köplinger [Sat, 4 Feb 2017 15:50:01 +0000 (16:50 +0100)]
[acceptance-tests] Baseline failing coreclr tests for now (#4325)
We can look into them when we have time, but for now let's
just ensure no new failures sneak in.
Alexander Köplinger [Sat, 4 Feb 2017 15:48:56 +0000 (16:48 +0100)]
[xbuild] Add .NET 4.6.2 entries and frameworklist.xml (#4326)
Jon Purdy [Fri, 3 Feb 2017 20:46:29 +0000 (12:46 -0800)]
[runtime] Require C# namespace to be quoted.
In GENERATE_GET_CLASS_WITH_CACHE and GENERATE_TRY_GET_CLASS_WITH_CACHE,
we were just quoting the namespace immediately.
Jon Purdy [Fri, 3 Feb 2017 00:37:46 +0000 (16:37 -0800)]
[runtime] Avoid 'strncpy'.
Its null-padding behavior can be surprising.
Jon Purdy [Thu, 2 Feb 2017 23:42:45 +0000 (15:42 -0800)]
[runtime] Use double-underscored attributes.
E.g., 'aligned' => '__aligned__'. We use the following attributes:
* aligned
* always_inline
* cold
* deprecated
* error
* format
* naked
* noinline
* packed
* printf
* tls_model
* unavailable
* used
* visibility
* warn_unused_result
Jon Purdy [Thu, 2 Feb 2017 23:49:06 +0000 (15:49 -0800)]
[runtime] Remove redundant #includes.
Jon Purdy [Thu, 2 Feb 2017 23:27:46 +0000 (15:27 -0800)]
[runtime] Avoid invalid characters.
There were some unquoted backticks and form-feed characters.
vkargov [Fri, 3 Feb 2017 20:52:56 +0000 (12:52 -0800)]
[runtime] Use ins->prev in mono_analyze_liveness. (#4296)
* [runtime] Add the missing MONO_BB_FOR_EACH_INS_REVERSE macro.
* [runtime] Use ins->prev instead of building reverse ins array in mono_analyze_liveness.
Zoltan Varga [Fri, 3 Feb 2017 12:43:14 +0000 (07:43 -0500)]
[runtime] Fix the computation of token sizes in ppdb files, the PDB stream has an additional table which contains the token sizes for the metadata tables referenced by the ppdb tables. (#4321)
Ankit Jain [Fri, 3 Feb 2017 12:19:05 +0000 (07:19 -0500)]
[xbuild] Disable pre-2.0 csc hack to use /debug:portable (#4322)
This hack should not be required now, as roslyn emits the correct debug
files with /debug:full and /debug:pdbonly, on non-windows platform.
This manifested as https://bugzilla.xamarin.com/show_bug.cgi?id=50710 ,
where the Xamarin.Mac targets override `$(CscToolExe)` to use `mcs`, but
xbuild's `ManagedCompiler` *always* converts `DebugType={full,pdbonly}`
to `/debug:portable`, which would cause mcs to complain.
Zoltan Varga [Thu, 2 Feb 2017 19:30:06 +0000 (14:30 -0500)]
[runtime] Factor out some code from mono_metadata_compute_size () to make it easier to add ppdb support later. NFC. (#4320)
Vlad Brezae [Thu, 2 Feb 2017 19:11:33 +0000 (21:11 +0200)]
Merge pull request #4306 from BrzVlad/fix-roslyn-ppc
[ppc] Save all the registers in the context
Vlad Brezae [Wed, 1 Feb 2017 16:29:26 +0000 (18:29 +0200)]
Revert "Default powerpc to build with mcs (#4313)"
This reverts commit
6a2a23dcebbf5bd19560b3bdeec54e13e26901f8.
Vlad Brezae [Tue, 31 Jan 2017 22:09:41 +0000 (00:09 +0200)]
[ppc] Save all the registers in the context
When suspending a thread, we were only saving the callee save registers in the context. Therefore SGen was failing to pin some objects. In order to store the registers in order, a code refactoring was necessary.
Martin Baulig [Thu, 2 Feb 2017 17:48:20 +0000 (18:48 +0100)]
[btls]: Make the 'MonoBtlsX509LookupMonoCollection' logic consistent with the native implementation. (#4319)
Martin Baulig [Thu, 2 Feb 2017 17:43:02 +0000 (18:43 +0100)]
[BTLS]: Improve error handling. (#4317)
* Add new native mono_btls_error_peek_error_line() and
mono_btls_error_get_error_line().
* Add `MonoBtlsError.PeekError(out string, out int)` and
`MonoBtlsError.GetError(out string, out int)` overloads.
* MonoBtlsContext.GetException(): use it here.
Martin Baulig [Thu, 2 Feb 2017 11:06:23 +0000 (12:06 +0100)]
Merge pull request #4316 from mono/work-51562
[btls]: Allow importing a null `byte[]`. Fixes #51562.
Niklas Therning [Thu, 2 Feb 2017 07:11:36 +0000 (02:11 -0500)]
Merge pull request #4312 from ntherning/winaot-tests-on-jenkins
Adds CI scripts to run the test suites under the winaot profile
Zoltan Varga [Wed, 1 Feb 2017 20:36:40 +0000 (15:36 -0500)]
Revert "[runtime] Fix the computation of token sizes in ppdb files, the PDB stream has an additional table which contains the token sizes for the metadata tables referenced by the ppdb tables. (#4304)"
This reverts commit
c35cdfa2dec8c15fdf49a6032c437751a97a5e08.
Revert this as it causes test failures.
Jonathan Chambers [Wed, 1 Feb 2017 19:58:23 +0000 (14:58 -0500)]
Send assembly and type load events in the domain that contains them rather than the root domain in the attach/defer case. (#4314)
Martin Baulig [Wed, 1 Feb 2017 18:36:33 +0000 (19:36 +0100)]
[btls]: Allow importing a null `byte[]`. Fixes #51562.
Zoltan Varga [Wed, 1 Feb 2017 18:05:22 +0000 (13:05 -0500)]
[runtime] Fix the computation of token sizes in ppdb files, the PDB stream has an additional table which contains the token sizes for the metadata tables referenced by the ppdb tables. (#4304)
Zoltan Varga [Wed, 1 Feb 2017 17:56:34 +0000 (12:56 -0500)]
[x86] Fix a race in the single stepping seq point assembly, we were doing a null check using cmp_membase, following by a call_membase, but the value could change between the two instructions, leading to a nullref. Fixes #50117. (#4308)
Zoltan Varga [Wed, 1 Feb 2017 17:45:38 +0000 (12:45 -0500)]
[sdb] Add a Domain property to AssemblyMirror. (#4309)
Zoltan Varga [Wed, 1 Feb 2017 17:42:40 +0000 (12:42 -0500)]
[runtime] Update comments in mono-debug-debugger.h.
monojenkins [Wed, 1 Feb 2017 16:40:17 +0000 (16:40 +0000)]
[msvc] Update csproj files (#4315)
Alexander Köplinger [Wed, 1 Feb 2017 15:47:01 +0000 (16:47 +0100)]
Default powerpc to build with mcs (#4313)
It crashes during build with roslyn on Jenkins.
Egor Bogatov [Wed, 1 Feb 2017 14:49:21 +0000 (15:49 +0100)]
[System.Data] referencesource -> corefx (#4299)
* [System.Data] SQLTypes from corefx
* [System.Data] DataStorage + subclasses from corefx
* [System.Data] remove dead referencesource.sources references
* [System.Data] Generate SR.cs from corefx using resx2sr tool
* [System.Data] move common references to corefx.common.sources
* [System.Data] remove /ReferenceSources/SR.cs
* [System.Data] remove duplicates in corefx.common.sources (accidentally were added)
* [System.Data] simple types & enums from corefx
* [System.Data] referencesource -> corefx
* [System.Data] Referencesource -> corefx
* [System.Data] re-order (sort) files in corefx.common.sources, DbCommandBuilder from corefx
* [System.Data] Referencesource -> corefx
* [System.Data] DataRow, DataSet, Filters, etc from corefx. XmlDataDocument is internal in corefx but used to be public in mono.
* [System.Data] rollback referencesource changes
* [System.Data] Wrap /referencesource/DataSet.cs with #ifdef COREFX in order to include missing DataSetSchemaImporterExtension and ignore DataSet
Alexander Köplinger [Wed, 1 Feb 2017 14:20:11 +0000 (15:20 +0100)]
[System] Disable SocketTest.ConnectedProperty on testing_aot_full for now
It failed since we introduced this profile and it was disabled
on Mac a few years ago. A bug is filed to investigate why it only
fails on FullAOT on Linux but works on non-FullAOT builds.
Disable the test for now to get closer to a green build.
https://bugzilla.xamarin.com/show_bug.cgi?id=52157
Alexander Köplinger [Wed, 1 Feb 2017 13:27:01 +0000 (14:27 +0100)]
[ci] Bump timeout for compile-runtime-tests
The new Azure Windows builders sometimes run into this timeout.
Niklas Therning [Wed, 1 Feb 2017 07:33:36 +0000 (08:33 +0100)]
Adds CI scripts to run the test suites under the winaot profile
Niklas Therning [Wed, 1 Feb 2017 12:58:54 +0000 (07:58 -0500)]
Merge pull request #4301 from ntherning/fix-Mono.Posix-API-breakage
Fixes the Mono.Posix API breakage introduced in #3985
Zoltan Varga [Wed, 1 Feb 2017 12:42:14 +0000 (07:42 -0500)]
[runtime] Fix parse_optimizations () so the optimization names don't … (#4300)
* [runtime] Fix parse_optimizations () so the optimization names don't need to be in a particular order.
* [jit] Avoid unused bits in optflags.
Zoltan Varga [Wed, 1 Feb 2017 10:37:17 +0000 (05:37 -0500)]
[ppdb] Fix support for embedded ppdb data.
Niklas Therning [Tue, 31 Jan 2017 10:03:15 +0000 (05:03 -0500)]
Fixes the Mono.Posix API breakage introduced in #3985
Also reverts the changes done to support/map.c. This is a generated file which
shouldn't be changed directly. An alternate fix for the problem in map.c on
Windows was to make sure L_SET et al are defined in mph.h which was already
done for Solaris.
Marek Safar [Wed, 1 Feb 2017 08:10:29 +0000 (09:10 +0100)]
Merge pull request #4310 from marek-safar/roslyn-rc3
Bump roslyn to rc3
Marek Safar [Wed, 1 Feb 2017 07:31:03 +0000 (08:31 +0100)]
Merge pull request #4302 from marek-safar/corefx-bump
Bump corefx
Marek Safar [Tue, 31 Jan 2017 22:52:58 +0000 (23:52 +0100)]
Bump roslyn to rc3
Marek Safar [Tue, 31 Jan 2017 10:10:24 +0000 (11:10 +0100)]
Bump corefx
Alexander Köplinger [Tue, 31 Jan 2017 21:05:31 +0000 (22:05 +0100)]
[tests] Fix assemblyresolve_event4.exe and reference-loader.exe FullAOT test fails
Fix assemblyresolve/Makefile.am so it correctly AOTs the
dependency of assemblyresolve_event4.exe (CLASS was empty before).
Disable reference-loader.exe since we can't AOT the
TestingReferenceAssembly.dll assembly which is a dependency
of the test so it would fail. Discussed with Aleksey, this is OK
because even if we could AOT the assembly we wouldn't use it since
the assembly loader looks at the managed assembly first, finds the
attribute and bails out before looking at the AOT image.
Alexander Köplinger [Tue, 31 Jan 2017 18:24:15 +0000 (19:24 +0100)]
[System] Fix AsyncOperationManager test that fails on the testing_full_aot profile
It seems the first assert only applies to XI/XM, so use those
in the condition instead of the profiles where it doesn't apply.
Alexander Köplinger [Tue, 31 Jan 2017 18:18:58 +0000 (19:18 +0100)]
Don't try to AOT test assemblies in runtime_preset=all (#4303)
This got broken in
ee048db72f867214ba2b9926d70a03badd79b33f
when the winaot profile was added to FULL_AOT_TESTS since we
enable that profile for building with runtime_preset=all which
means that FULL_AOT_TESTS would be defined and the runtime tests
would try to AOT the test assemblies.
Instead, we now only try to AOT if we're actually testing
a profile that needs it. Also introduced a new HYBRID_AOT_TESTS
variable and replaced the existing INSTALL_TESTING_AOT_HYBRID
since that profile would run into the same issue if we included
it in runtime_preset=all (using the INSTALL_ variables to decide
about tests is misleading!)
Marek Safar [Mon, 30 Jan 2017 16:40:32 +0000 (17:40 +0100)]
[mono] Update ignored assemblies list to include System.ValueTuple