platform/upstream/dotnet/runtime.git
4 years agoMerge pull request mono/mono#18053 from EgorBo/fix-fneg-llvm
Egor Bogatov [Thu, 5 Dec 2019 18:13:11 +0000 (21:13 +0300)]
Merge pull request mono/mono#18053 from EgorBo/fix-fneg-llvm

LLVM: use fneg instead of fsub for `-x`

Commit migrated from https://github.com/mono/mono/commit/84839ab297e0e5e2e51744933e7e5d71bfa53db6

4 years ago[netcore] Remove approx. 10 icalls. (mono/mono#18018)
Jay Krell [Thu, 5 Dec 2019 18:10:29 +0000 (10:10 -0800)]
[netcore] Remove approx. 10 icalls. (mono/mono#18018)

* [netcore] Remove ves_icall_System_Runtime_InteropServices_Marshal_PtrToStringUni and _len icalls.
They are implemented in C#.
The C# is equivalent to the C, and could be used for classic runtime, except for the treatment of large/negative lengths.

* Remove a few more icalls.
Instead of ifdef and leaving them uncompiled when compiling netcore,
declare them and put asserts in them.

Commit migrated from https://github.com/mono/mono/commit/ae9192f96febdb5ceea3b73d6aa226d601006371

4 years ago[llvm] Avoid decomposing OP_STOREV_MEMBASE early for types which contains no referenc...
Zoltan Varga [Thu, 5 Dec 2019 15:02:43 +0000 (16:02 +0100)]
[llvm] Avoid decomposing OP_STOREV_MEMBASE early for types which contains no references. (mono/mono#18041)

Commit migrated from https://github.com/mono/mono/commit/e67673fac85d13882256893d1b61c08ca0fc13c6

4 years agoFix ifndef include guard in icall-decl.h. (mono/mono#18055)
Jay Krell [Thu, 5 Dec 2019 09:19:24 +0000 (01:19 -0800)]
Fix ifndef include guard in icall-decl.h. (mono/mono#18055)

There was no clash so ok either way, but this is more like intended.

Commit migrated from https://github.com/mono/mono/commit/5a5e092a53111153b50742f1d021da45c09288b3

4 years agoConvert to 0/1 in OP_XEQUAL (mono/mono#18038)
Egor Bogatov [Thu, 5 Dec 2019 01:38:34 +0000 (04:38 +0300)]
Convert to 0/1 in OP_XEQUAL (mono/mono#18038)

Commit migrated from https://github.com/mono/mono/commit/a3fbb644938bdfe008796f281e958f7707069dfe

4 years agoremove some items from rsp
EgorBo [Wed, 4 Dec 2019 23:09:11 +0000 (02:09 +0300)]
remove some items from rsp

Commit migrated from https://github.com/mono/mono/commit/55a3c10e7235e2ace77a6dc502e9349568a68b7f

4 years agouse fneg instead of fsub
EgorBo [Wed, 4 Dec 2019 23:03:24 +0000 (02:03 +0300)]
use fneg instead of fsub

Commit migrated from https://github.com/mono/mono/commit/92047c453ffe89dc8283c0705ddc8a7f457371b3

4 years ago[amd64] align application stack pointer in signal handler (mono/mono#18021)
Bernhard Urban-Forster [Wed, 4 Dec 2019 21:23:19 +0000 (22:23 +0100)]
[amd64] align application stack pointer in signal handler (mono/mono#18021)

* [amd64] align application stack pointer in signal handler

16byte stack pointer alignment is only required in certain places by the ABI, so the assumption that it holds all the time is wrong (though, it passed CI 😅). We have to setup the frame properly, i.e. we have to ensure the stack pointer is properly aligned. That's what we have done previously too.

Regression of https://github.com/mono/mono/pull/17922
Fixes https://github.com/mono/mono/issues/18006

* add ALIGN_DOWN_TO macro

* remove unnecessary ifndefs

Commit migrated from https://github.com/mono/mono/commit/9d58bc1d12085f19ac2ba198c2bc4f9798810532

4 years agoUpdate submodules before running external/bdwgc/autogen.sh (mono/mono#18042)
Alexander Köplinger [Wed, 4 Dec 2019 21:16:21 +0000 (22:16 +0100)]
Update submodules before running external/bdwgc/autogen.sh (mono/mono#18042)

Otherwise we'll run into issues because in configure.ac we update submodules _after_ we try to call external/bdwgc/configure (which won't be there if autogen.sh didn't run).
This would manifest itself as `*** No rule to make target 'all'.` in external/bdwgc when you build as there's no Makefile.

Commit migrated from https://github.com/mono/mono/commit/59554712cc1e385dc73eb309cc810c08e28ef23f

4 years agoRemove mk/ folder (mono/mono#18039)
Alexander Köplinger [Wed, 4 Dec 2019 18:55:22 +0000 (19:55 +0100)]
Remove mk/ folder (mono/mono#18039)

The plan to put additional helpers into this directory didn't materialize and we can just inline the one thing we have in common.mk where needed.

Commit migrated from https://github.com/mono/mono/commit/4e8ead7caf330dcc271fac38cdbf5b0ef90e2361

4 years ago[android] don't probe directories for .config files (mono/mono#18024)
Jonathan Peppers [Wed, 4 Dec 2019 10:50:59 +0000 (04:50 -0600)]
[android] don't probe directories for .config files (mono/mono#18024)

* [android] don't probe directories for .config files

In Xamarin.Android, on startup we see logging like this for every .NET
assembly:

    12-03 14:27:42.104 26842 26842 D Mono    : Prepared to set up assembly 'FormsViewGroup' (/storage/emulated/0/Android/data/Xamarin.Forms_Performance_Integration/files/.__override__/FormsViewGroup.dll)
    12-03 14:27:42.104 26842 26842 D Mono    : Assembly FormsViewGroup[0xebcee9c0] added to domain RootDomain, ref_count=1
    12-03 14:27:42.104 26842 26842 D Mono    : Assembly Loader loaded assembly from location: '/storage/emulated/0/Android/data/Xamarin.Forms_Performance_Integration/files/.__override__/FormsViewGroup.dll'.
    12-03 14:27:42.104 26842 26842 D Mono    : Config attempting to parse: '/storage/emulated/0/Android/data/Xamarin.Forms_Performance_Integration/files/.__override__/FormsViewGroup.dll.config'.
    12-03 14:27:42.104 26842 26842 D Mono    : Config attempting to parse: '/Users/builder/jenkins/workspace/archive-mono/2019-10/android/release/sdks/out/android-x86-release/etc/mono/assemblies/FormsViewGroup/FormsViewGroup.config'.

In most cases, no `.dll.config` file exists and so it probes this
macOS-looking path:

    /Users/builder/jenkins/workspace/archive-mono/2019-10/android/release/sdks/out/android-x86-release/etc/mono/assemblies/FormsViewGroup/FormsViewGroup.config

It looks like this is happening because nothing in Xamarin.Android
calls `mono_set_dirs` or `mono_set_config_dir` and so it falls back to
using `MONO_CFG_DIR`. `MONO_CFG_DIR` is set to a path on the build
machine.

Even if Xamarin.Android *did* call `mono_set_config_dir`, the code
appends `mono/assemblies`:

    cfg = g_build_filename (cfg_dir, "mono", "assemblies", aname, cfg_name, (const char*)NULL);

So it doesn't seem useful to use this fallback at all, there would
never be a file found.

I wrapped the appropriate code in `#ifndef TARGET_ANDROID` blocks.

I was able to see a minor *maybe* ~20ms performance improvement on
startup:

    Before:
    12-03 14:27:48.494 26903 26903 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:601::443900
    12-03 14:27:54.356 26963 26963 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:587::495400
    12-03 14:28:00.263 27021 27021 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:586::147100
    12-03 14:28:06.135 27074 27074 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:593::639100
    12-03 14:28:11.988 27129 27129 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:571::194900
    After:
    12-03 14:46:50.590 27737 27737 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:503::918700
    12-03 14:46:56.517 27794 27794 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:570::533600
    12-03 14:47:02.368 27850 27850 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:537::435000
    12-03 14:47:08.289 27904 27904 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:587::830900
    12-03 14:47:14.130 27965 27965 I monodroid-timing: Runtime.init: end, total time; elapsed: 0s:555::593400

This was the Xamarin.Forms integration project in xamarin-android.
Running a `Debug` build on a HAXM x86 emulator on Windows.

https://github.com/xamarin/xamarin-android/tree/master/tests/Xamarin.Forms-Performance-Integration

* Add --enable-minimal=cfgdir_config

* This also defines `DISABLE_CFGDIR_CONFIG`
* Used this setting for android

Commit migrated from https://github.com/mono/mono/commit/b816189b947ab590e82509f6eb6130546927c988

4 years ago[configure.ac] Build on FreeBSD/PowerPC64 (mono/mono#18015)
Aleksey Kliger (λgeek) [Wed, 4 Dec 2019 00:56:30 +0000 (19:56 -0500)]
[configure.ac] Build on FreeBSD/PowerPC64 (mono/mono#18015)

Community contribution from https://github.com/mono/mono/issues/18009
from @igorzigorz

Commit migrated from https://github.com/mono/mono/commit/467d46f8abbd55c242b267771aa7575730a2a72c

4 years agoMerge pull request mono/mono#18013 from EgorBo/intrin-string-memcpy
Egor Bogatov [Tue, 3 Dec 2019 23:19:32 +0000 (02:19 +0300)]
Merge pull request mono/mono#18013 from EgorBo/intrin-string-memcpy

[netcore] Improve jitdiff tool

Commit migrated from https://github.com/mono/mono/commit/85826e5c2771d9b30dabde51f6fcce83988f4b7f

4 years agoImprove TryParseFunctionName in jitdiff
EgorBo [Tue, 3 Dec 2019 15:13:48 +0000 (18:13 +0300)]
Improve TryParseFunctionName in jitdiff

Commit migrated from https://github.com/mono/mono/commit/5f58bb0959d6a50f8577e52b715f46185ab8e429

4 years agoRemove dead docs
EgorBo [Tue, 3 Dec 2019 15:11:42 +0000 (18:11 +0300)]
Remove dead docs

Commit migrated from https://github.com/mono/mono/commit/ea0824cb2276ca756f667219ce44a135a0ff4318

4 years agoDon't throw NRE when len == 0 in Buffer.Memmove (mono/mono#18004)
Egor Bogatov [Tue, 3 Dec 2019 14:21:41 +0000 (17:21 +0300)]
Don't throw NRE when len == 0 in Buffer.Memmove (mono/mono#18004)

Commit migrated from https://github.com/mono/mono/commit/8150a38dae6fbf60a8a5ffc20a1d07b911fbc31b

4 years agoBump version to 6.11 (mono/mono#17990)
Jo Shields [Mon, 2 Dec 2019 20:58:38 +0000 (15:58 -0500)]
Bump version to 6.11 (mono/mono#17990)

Commit migrated from https://github.com/mono/mono/commit/a2f9fb77a3704049ffe5de32362029ece154ee4c

4 years agoRemove handles/MonoError from Mono.RuntimeClassHandle. (mono/mono#17805)
Jay Krell [Mon, 2 Dec 2019 18:18:03 +0000 (10:18 -0800)]
Remove handles/MonoError from Mono.RuntimeClassHandle. (mono/mono#17805)

Commit migrated from https://github.com/mono/mono/commit/72aa81e9010088191328002da887035b793b56cb

4 years ago[profiler] disable inlining when coverage profiler is used and fix allocation sizes...
Bernhard Urban-Forster [Mon, 2 Dec 2019 18:04:13 +0000 (19:04 +0100)]
[profiler] disable inlining when coverage profiler is used and fix allocation sizes (mono/mono#17986)

* [profiler] let compiler decide about the size of an entry

* [profiler] disable inlining when coverage profiler is used

Commit migrated from https://github.com/mono/mono/commit/a9a0a1e5b95a93cd3d3f73c8eeafe7e64da4acd8

4 years ago[sgen] Don't trigger collections during allocation of thread objects (mono/mono#17970)
Vlad Brezae [Mon, 2 Dec 2019 17:54:21 +0000 (19:54 +0200)]
[sgen] Don't trigger collections during allocation of thread objects (mono/mono#17970)

Allocation of thread objects usese the mono_object_new_mature API. These objects are allocated before the thread is finished attaching to the runtime. This means that a collection can happen on an unattached thread as well as all its callbacks, which is a counter-intuitive behavior.

On android this is problematic because at the end of the SGen collection we need to call the java collection, which needs to have the thread attached. This could probably be fixed instead from the Xamarin.Android side, but this approach seems simpler and saner.

Fixes https://github.com/mono/mono/issues/17878

Commit migrated from https://github.com/mono/mono/commit/640ffa7a62a861b048fb311c0bdcb0a500336ad2

4 years agoRemove handles/MonoError from Mono.Security.Cryptography.KeyPairPersistence icalls...
Jay Krell [Mon, 2 Dec 2019 17:51:16 +0000 (09:51 -0800)]
Remove handles/MonoError from Mono.Security.Cryptography.KeyPairPersistence icalls. (mono/mono#17967)

Commit migrated from https://github.com/mono/mono/commit/d7f5a5fdfedf2caa1b8cb5995b3ab1cd3e1c7a09

4 years ago[DIM] Fix behavior when there is an override of a method in a generic interface....
Thays Grazia [Mon, 2 Dec 2019 16:20:00 +0000 (13:20 -0300)]
[DIM] Fix behavior when there is an override of a method in a generic interface.  (mono/mono#17963)

* Fix behavior when there is an override of a method in a generic interface.
Fixes mono/mono#17869

Commit migrated from https://github.com/mono/mono/commit/3834be3d2242312a6284f89ed77b60ff8b2327ab

4 years agoRefactor Monitor to eliminate near identical duplicated code. (mono/mono#17964)
Jay Krell [Mon, 2 Dec 2019 16:06:11 +0000 (08:06 -0800)]
Refactor Monitor to eliminate near identical duplicated code. (mono/mono#17964)

The only difference in the copies, was that one had an infinite timeout
so implied TRUE result unless interrupted.

Where the other copy had a timeout and could return FALSE as a result.

They both used the same underlying, with timeout, so it wasn't really an optimization,
just duplicated code.

Also there is subtelty around having a return value or an out parameter.
That is carefully preserved.

Also, coop is fixed, by creating a frame only in a slow path.
Presumably creating a thread is quick and stateless enough such that this
is not too late.

Commit migrated from https://github.com/mono/mono/commit/7356aedc77020cca65af4948dda3859cdb72b6e6

4 years ago[wasm] Add beginnings of SIMD support in AOT mode. (mono/mono#17971)
Zoltan Varga [Mon, 2 Dec 2019 06:21:17 +0000 (07:21 +0100)]
[wasm] Add beginnings of SIMD support in AOT mode. (mono/mono#17971)

* [jit] Implement beginnings of SIMD support for WASM.

* [wasm] Add a DISABLE_WASM_NO_NETCORE Make.local conditional to disable non-netcore builds during development.

* [wasm] Add a --simd argument to the packager.

* Remove some duplicate defines.

* Fix an windows assertion.

Commit migrated from https://github.com/mono/mono/commit/f13e6ba91ab184d90aa4ad05c72a30367656247e

4 years agoenable only for netcore
EgorBo [Tue, 26 Nov 2019 19:26:45 +0000 (22:26 +0300)]
enable only for netcore

Commit migrated from https://github.com/mono/mono/commit/d33ce1c3ec6c477af5d28f3c1a3b5c29b2374cb0

4 years agofix failing tests
EgorBo [Mon, 25 Nov 2019 17:50:24 +0000 (20:50 +0300)]
fix failing tests

Commit migrated from https://github.com/mono/mono/commit/376d3acb292ca878cb2bde8993b5ba4ddaaecc14

4 years agofix build
EgorBo [Fri, 22 Nov 2019 22:32:03 +0000 (01:32 +0300)]
fix build

Commit migrated from https://github.com/mono/mono/commit/b77f757679ebf6aea39ce1a20d4f7d9d88ec35b2

4 years agoAddress feedback and fix C++ warnings
EgorBo [Fri, 22 Nov 2019 21:49:49 +0000 (00:49 +0300)]
Address feedback and fix C++ warnings

Commit migrated from https://github.com/mono/mono/commit/1b7c7e60758af9bee4d1d6993b1472b117d225a6

4 years agoOptimize box(X) == null to false
EgorBo [Fri, 22 Nov 2019 21:26:12 +0000 (00:26 +0300)]
Optimize box(X) == null to false

Commit migrated from https://github.com/mono/mono/commit/e2241969f56342ef1e38e730bf470a8df379e1aa

4 years agosockets: Remove piecemeal ifndef ENABLE_NETCORE since the file has it broadly.
Jay Krell [Sun, 1 Dec 2019 10:23:52 +0000 (02:23 -0800)]
sockets: Remove piecemeal ifndef ENABLE_NETCORE since the file has it broadly.

Commit migrated from https://github.com/mono/mono/commit/fffeb232394e2e762bb1d0f9ac67b808696379c6

4 years ago[interp] track valuetype stack for CEE_RET too (mono/mono#17968)
Bernhard Urban-Forster [Fri, 29 Nov 2019 23:06:06 +0000 (00:06 +0100)]
[interp] track valuetype stack for CEE_RET too (mono/mono#17968)

[interp] track valuetype stack for CEE_RET too

Usually the stack depth is determined via basic block predecessors, but  that only works for forward branches.

Fixes https://github.com/mono/mono/issues/17924

Commit migrated from https://github.com/mono/mono/commit/4d0feffa42fd96a19985c239aaa85aa44142dc67

4 years ago[wasm] Add more netcore test suites. (mono/mono#17965)
Zoltan Varga [Fri, 29 Nov 2019 17:17:40 +0000 (18:17 +0100)]
[wasm] Add more netcore test suites. (mono/mono#17965)

* [wasm] Handle more exceptions in the xunit runner.

* [wasm] Add more corefx test suites.

* Add more icall signatures.

* Add support for running the System.Compression test suites.

Commit migrated from https://github.com/mono/mono/commit/6c0bfdc3f3d5855d27628112a505ba01bdfc4584

4 years agoInterpreter fixes. (mono/mono#17966)
Zoltan Varga [Fri, 29 Nov 2019 15:26:05 +0000 (16:26 +0100)]
Interpreter fixes. (mono/mono#17966)

Interpreter fixes.

Commit migrated from https://github.com/mono/mono/commit/d1735a8356cd8a56c9382cca0b59ed2a301f3fcf

4 years ago[netcore] Improve Jitdiff tool (mono/mono#17960)
Egor Bogatov [Fri, 29 Nov 2019 10:54:01 +0000 (13:54 +0300)]
[netcore] Improve Jitdiff tool (mono/mono#17960)

use `objdump` arguments to reduce verbosity (no raw bytes, etc) and make the dasm more diff-friendly.
On Linux all functions have a random prefix (address) - replace it with `0xD1FFAB1E` constant). On macOS it's fine by default.

Also, allow to generate diffs for a single assembly via `dump-asm-lib` rule.

Commit migrated from https://github.com/mono/mono/commit/9a18b6fd6fc4d2da86430ddfdb5ab58c2986d92b

4 years agoRevert "Use range metadata (mono/mono#17956)" (mono/mono#17958)
Zoltan Varga [Thu, 28 Nov 2019 20:52:01 +0000 (21:52 +0100)]
Revert "Use range metadata (mono/mono#17956)" (mono/mono#17958)

This reverts commit mono/mono@fde058518d4741f2db2aca3e01f0513fd151961f.

Commit migrated from https://github.com/mono/mono/commit/59a9af6d7ce3ff7785e9038e3902fc77291475b8

4 years agoUse range metadata (mono/mono#17956)
Egor Bogatov [Thu, 28 Nov 2019 13:26:35 +0000 (16:26 +0300)]
Use range metadata (mono/mono#17956)

Commit migrated from https://github.com/mono/mono/commit/fde058518d4741f2db2aca3e01f0513fd151961f

4 years ago[interp] Make calli non recursive as well. (mono/mono#17953)
Zoltan Varga [Thu, 28 Nov 2019 05:26:36 +0000 (06:26 +0100)]
[interp] Make calli non recursive as well. (mono/mono#17953)

Commit migrated from https://github.com/mono/mono/commit/79db9b2a8776c5a19ab79b713726577c3d8bea16

4 years agoAdd quick diff
EgorBo [Wed, 27 Nov 2019 15:29:25 +0000 (18:29 +0300)]
Add quick diff

Commit migrated from https://github.com/mono/mono/commit/6360ccbfe31ebacdfb9e65ab05f8758513bb19be

4 years agoAdd jit-diff tools
EgorBo [Wed, 27 Nov 2019 14:45:20 +0000 (17:45 +0300)]
Add jit-diff tools

Commit migrated from https://github.com/mono/mono/commit/e333f13710af0c3487bd55b3e8c4fdd4df4310e6

4 years agoChange signature, add cast
Ryan Lucia [Mon, 25 Nov 2019 19:01:16 +0000 (14:01 -0500)]
Change signature, add cast

Commit migrated from https://github.com/mono/mono/commit/9c1cae02cac33cfa215611211932a4d4c78dcd5f

4 years ago[mini] Move bundled native library loading into the runtime
Ryan Lucia [Mon, 25 Nov 2019 16:55:39 +0000 (11:55 -0500)]
[mini] Move bundled native library loading into the runtime

Commit migrated from https://github.com/mono/mono/commit/0c3036abc701da73986f399e17124c2415c95c64

4 years ago[cxx] MonoJitICallId same between C and C++.
Jay Krell [Fri, 25 Oct 2019 02:51:33 +0000 (19:51 -0700)]
[cxx] MonoJitICallId same between C and C++.
This is optional.
The real problem was -enable-cxx not actually compiling some files as C++.
This is still reasonable, to ward off the danger of that occuring again.

Commit migrated from https://github.com/mono/mono/commit/b2817e6404d573e837e1d8c7e047135974a3b311

4 years agoUpdate netcore.
Jay Krell [Sat, 23 Nov 2019 10:11:46 +0000 (02:11 -0800)]
Update netcore.

Commit migrated from https://github.com/mono/mono/commit/c3da0ad518537e8f912b94427c15143f7499408d

4 years agoRemove handles/MonoError from Mono.RuntimeMarshal and Mono.SafeStringMarshal.
Jay Krell [Fri, 15 Nov 2019 05:47:47 +0000 (21:47 -0800)]
Remove handles/MonoError from Mono.RuntimeMarshal and Mono.SafeStringMarshal.

Commit migrated from https://github.com/mono/mono/commit/4a15e73429d5ac994ff7cd183c14bbce92806e6b

4 years ago[threadpool] Decrement max_working when worker times out (mono/mono#17927)
Aleksey Kliger (λgeek) [Wed, 27 Nov 2019 21:06:17 +0000 (16:06 -0500)]
[threadpool] Decrement max_working when worker times out (mono/mono#17927)

[threadpool] Decrement max_working when worker times out

The corresponding code in CoreCLR:
https://github.com/dotnet/runtime/blob/mono/mono@bf88f146412d1e2d41779422337184a802f186c9/src/coreclr/src/vm/win32threadpool.cpp#L2156-L2164

The issue is that counter._.max_working is only ever changed by:
1. the monitor_thread when it detects starvation (increment by 1).
2. the hill climbing algorithm (increment or decrement).

Creating a periodic load of creating many workers at once followed by a minute
or more of quiescence increases the variable worker.counters._.max_working in
monitor_thread(). Once that counter matches worker.limit_worker_max,
monitor_thread() keeps looping without unparking or creating a thread, even
though the actual number of threads/workers is small. If the existing threads
are all waiting on work that needs a new thread, then a deadlock occurs.

With this change, when a parked worker times out, it will lower max_working to
the number of active (working + parked + starting) threads minus itself (but no
less than limit_worker_min). As a result, monitor_thread will only increment
max_working as long as none of the already running workers are timing out.

Attempt to address https://github.com/mono/mono/issues/17833

---

Also allow monitor_thread to start workers if there aren't enough active, even if the max worker limit has been reached.

Work around a pathological condition where the work_item_count is non-zero, and
the max_working limit has been reached, but the number of active threads is
still below the max.  In that case, unpark some workers and start some new
threads.

May address http://work.azdo.io/827206

Commit migrated from https://github.com/mono/mono/commit/b6a351c330836dd59f1458beecda62de6cbee3e5

4 years agoCleanup initialize_object_slots, remove unused GetHashCode (ghc) handling. (mono...
Jay Krell [Wed, 27 Nov 2019 19:25:14 +0000 (11:25 -0800)]
Cleanup initialize_object_slots, remove unused GetHashCode (ghc) handling. (mono/mono#17834)

Fix possible race conditions involving multiple writes to globals, by
removing one of the two globals involved in finalize.

Commit migrated from https://github.com/mono/mono/commit/3e369ba8c9a456d26679405e96d01ab285f697d9

4 years ago[sdks] Add netcore option for iOS SDK
Alexander Köplinger [Mon, 25 Nov 2019 17:12:47 +0000 (18:12 +0100)]
[sdks] Add netcore option for iOS SDK

Commit migrated from https://github.com/mono/mono/commit/a0d79a4886e76b077f9377185945b959adb6fc8c

4 years ago[merp] Use timeline (and generate timeline breadcrumbs) for SendExceptionToTelemetry...
Alexis Christoforides [Wed, 27 Nov 2019 16:56:12 +0000 (11:56 -0500)]
[merp] Use timeline (and generate timeline breadcrumbs) for SendExceptionToTelemetry() (mono/mono#17913)

Fixes https://github.com/mono/mono/issues/17790

Commit migrated from https://github.com/mono/mono/commit/167913524fd5029a4762d0b497a84739d7227fb3

4 years ago[arm] if mtriple is provided, do not set -march (mono/mono#17937)
Bernhard Urban-Forster [Wed, 27 Nov 2019 16:36:03 +0000 (17:36 +0100)]
[arm] if mtriple is provided, do not set -march (mono/mono#17937)

[arm] if mtriple is provided, do not set -march

E.g. when passing `-march=arm -mtriple=armv7s-ios`, `-march` will win and thus ignore `armv7s`.

Fixes https://github.com/mono/mono/issues/17931

* Caused by mono/mono@5e318afd2e1d5f9ef0fd692abbceadcf615389a5
* Related with mono/mono@4cd506823079bc15c9ee72da4f94a653f237a5e5
* Almost fixed by mono/mono@e431093f73360b5726e5256fd0de3108ec21b38a

Hopefully we are done with it.

Commit migrated from https://github.com/mono/mono/commit/4c595069e72e6dc39e13c1e322087a502ea000c9

4 years ago[amd64] do not stack allocate on the application stack for the transtion from alstack...
Bernhard Urban-Forster [Wed, 27 Nov 2019 11:20:47 +0000 (12:20 +0100)]
[amd64] do not stack allocate on the application stack for the transtion from alstack handling (mono/mono#17922)

### Backstory
This test is failing in the interpreter (ONLY on Debian9/amd64. Not Ubuntu/amd64. Not macOS/amd64):
https://github.com/mono/mono/blob/mono/mono@0fed03ed63ed4ea742c4511d8edc3bc1c6f4044f/mcs/class/Mono.Debugger.Soft/Test/dtest.cs#L2484-L2499
Where the debuggee is just some unsafe code that tries to read from unmapped memory, thus causing a segfault.

So the test expects the runtime to send a CRASH event via the managed debugger interface. This happens very late in the crash handling machinery:
https://github.com/mono/mono/blob/mono/mono@0fed03ed63ed4ea742c4511d8edc3bc1c6f4044f/mono/mini/mini-posix.c#L1105

However, the runtime would never reach that, because it silently crashes when calling `backtrace`:
https://github.com/mono/mono/blob/mono/mono@0fed03ed63ed4ea742c4511d8edc3bc1c6f4044f/mono/mini/mini-posix.c#L939

At the point where the SIGSEGV happens, the native stack trace looks like this:
```
Thread 1 "mono-sgen" received signal SIGSEGV, Segmentation fault.
interp_exec_method_full (frame=0x555555c51f18, context=0x555555c38350, clause_args=clause_args@entry=0x0, error=0x555555c53f18, error@entry=0x7fffffffd8b0) at interp/interp.c:4463
4463                            sp[-1].data.i = *(gint32*)sp[-1].data.p;
(gdb) bt
%0  interp_exec_method_full (frame=0x555555c51f18, context=0x555555c38350, clause_args=clause_args@entry=0x0, error=0x555555c53f18, error@entry=0x7fffffffd8b0) at interp/interp.c:4463
%1  0x000055555569237a in interp_runtime_invoke (method=<optimized out>, obj=<optimized out>, params=<optimized out>, exc=0x0, error=0x7fffffffd8b0) at interp/interp.c:1900
%2  0x0000555555590745 in mono_jit_runtime_invoke (method=0x555555bfdc38, obj=<optimized out>, params=0x7fffffffd858, exc=0x0, error=0x7fffffffd8b0) at mini-runtime.c:3001
%3  0x0000555555778c4c in do_runtime_invoke (method=0x555555bfdc38, obj=<optimized out>, params=<optimized out>, exc=<optimized out>, error=0x7fffffffd8b0) at object.c:3052
%4  0x000055555577b7d0 in do_exec_main_checked (method=0x555555bfdc38, args=<optimized out>, error=0x7fffffffd8b0) at object.c:5184
%5  0x000055555559a2aa in mono_jit_exec_internal (argv=0x7fffffffdcb8, argc=1, assembly=0x0, domain=0x555555bf9d40) at driver.c:1320
%6  mono_jit_exec (domain=domain@entry=0x555555bf9d40, assembly=assembly@entry=0x555555c66ae0, argc=argc@entry=1, argv=argv@entry=0x7fffffffdcb8) at driver.c:1265
%7  0x000055555559b8e2 in main_thread_handler (user_data=<synthetic pointer>) at driver.c:1402
%8  mono_main (argc=<optimized out>, argv=<optimized out>) at driver.c:2622
%9  0x000055555558b757 in mono_main_with_options (argv=<optimized out>, argc=<optimized out>) at main.c:52
%10 main (argc=<optimized out>, argv=<optimized out>) at main.c:434
```
Stepping into the crash handling machinery, the picture looks a bit different:
```
Thread 1 "mono-sgen" hit Breakpoint 1, altstack_handle_and_restore (ctx=0x7fffffffd2d0, obj=0x0, flags=0) at exceptions-amd64.c:871
871     {
(gdb) bt
%0  altstack_handle_and_restore (ctx=0x7fffffffd2d0, obj=0x0, flags=0) at exceptions-amd64.c:871
%1  0x00005555556957de in interp_exec_method_full (frame=0x555555c53f08, context=0x555555c38350, clause_args=0x0, error=0x555555c53f18) at interp/interp.c:4461
%2  0x0000000000000000 in ?? ()
```
So it looks like we messed up in the transition from altstack back to the program stack.

 ### The fix

The problem is that `backtrace` consults the unwind information emitted at the point where SIGSEGV happens in `interp_exec_method_full ()`. However at that given program code, there is no function call that allocates space to pass arguments on the stack. Still, we massage the stack to do that in the signal handler, which does not add up with the emitted unwind information by the C compiler.

Note that without the fix, crashing in `backtrace` is the worst case. In other cases we might (1) get still a correct stack trace, or (2) we get a wrong stack trace, and `backtrace` will not crash. So this PR might improve the crash experience overall.

This also reverts commit mono/mono@cff400f3fae9775ac734f8a691339e59a272872c.

Commit migrated from https://github.com/mono/mono/commit/690af29ad904c2005f3d9387d7f875571fcf0fcd

4 years ago[interp] Non-recursive interpreter (mono/mono#16461)
Zoltan Varga [Tue, 26 Nov 2019 03:31:32 +0000 (04:31 +0100)]
[interp] Non-recursive interpreter (mono/mono#16461)

* Aling alloca_size to 8.

* Prevent some false pinning in the monitor-resurrection.cs test when using the non-recursive interpreter.

* [runtime] Add a mono_compiler_barrier () function.

* Add a MONO_GET_SP opcode to pass to enter/exit_gc_safe_region, ldloca might not return a native stack address with the interpreter.

* Add an interpreter/gc callback to mark the interpreter stack.

* Copy only an integer in ENDFILTER.

* Add a native_stack_addr field to InterpFrame so the EH code has something to compare against native stack addresses.

* Add a FrameStack structure to allow allocation of InterpFrame structures and stack data outside the native stack.

* Allocate InterpFrames/stack data from a frame stack instead of the native stack.

* Make some calls non-recursive.

* Remove an assert.

* Use non-recursive calls in CALLVIRT_FAST as well.

* Fix stack size allocation for vararg calls.

* Add some micro optimizations to the calling code.

* Add more micro optimizations.

* [dtest] disable Crash test on interpreter

* Add some implicit conversions to STSFLD to fix a corefx test failure.

* Disable the new test.

Commit migrated from https://github.com/mono/mono/commit/c4136ebd94567bb29557c94e7480e24fe72ffdac

4 years agoWASM + Netcore. (mono/mono#17915)
Zoltan Varga [Tue, 26 Nov 2019 01:59:31 +0000 (02:59 +0100)]
WASM + Netcore. (mono/mono#17915)

* [runtime] Throw an exception from Thread::StartInternal () if threads are disabled.

* Add more test suites and a corefx exclusions file.

Commit migrated from https://github.com/mono/mono/commit/5b4cd5ce50e293548fd097622e41010807fcf2b8

4 years agoRemove mkinstalldirs invocation from two places (mono/mono#17910)
Alexander Köplinger [Mon, 25 Nov 2019 20:10:26 +0000 (21:10 +0100)]
Remove mkinstalldirs invocation from two places (mono/mono#17910)

These can be easily replaced so we don't need to mirror over the mkinstalldirs file to the new repo

Commit migrated from https://github.com/mono/mono/commit/e437330808f6f7838eb025e0f46898d69b044823

4 years ago[configure.ac] Don't run compiler server check in netcore mode (mono/mono#17911)
Alexander Köplinger [Mon, 25 Nov 2019 20:10:08 +0000 (21:10 +0100)]
[configure.ac] Don't run compiler server check in netcore mode (mono/mono#17911)

We don't need it there.

Commit migrated from https://github.com/mono/mono/commit/8aa1cdfe178f462aa3bdcd787533b5def073b83d

4 years agoFix boehm on Windows i386 builds (mono/mono#17901)
Alexander Köplinger [Mon, 25 Nov 2019 17:58:42 +0000 (18:58 +0100)]
Fix boehm on Windows i386 builds (mono/mono#17901)

After https://github.com/mono/mono/pull/16832 was merged the run-msbuild.bat
script has an additional required parameter for specifying the GC.

However the script only passed the MONO_TARGET_GC property to msbuild
if the script had no additional parameters which doesn't make sense.

Always pass the parameter, this should fix the boehm build on Windows i386.

Commit migrated from https://github.com/mono/mono/commit/bdf4790fa061704a6e40d55c07b214961e5083cc

4 years ago[WIP] Make debugger agent startup faster (mono/mono#17898)
Marek Habersack [Mon, 25 Nov 2019 16:52:58 +0000 (16:52 +0000)]
[WIP] Make debugger agent startup faster (mono/mono#17898)

Mono SDB's debugger agent takes a `host:port` parameter to specify the host and
port on which to listen for debugger connection. `host` is passed, via
`mono_get_address_info`, to the `getaddrinfo(3)` standard library call which
does the job of translating the name to the IP address. However, there's no
special treatment for situations when `host` is already an IP address. In this
case `getaddrinfo` takes a lot of time to perform at least two DNS lookups (if
there's just a single DNS server configured) which will time out as the IP
address in `host` will not be resolved by any of the servers.

`getaddrinfo` takes a "hints" parameter where caller can specify some
information about the `host:port` pair to make the lookup faster/easier.
However, Mono doesn't specify the address family in the `hints` parameter, thus
forcing the slow path of DNS lookups on `getaddrinfo`.

This commit adds a new hint named `MONO_HINT_NUMERIC_HOST` (which maps to
`AI_NUMERICHOST` for `getaddrinfo`) which will allow, if set, `getaddrinfo` to
take a faster route in address discovery. Debugger agent will, in most cases,
get an IP address to instead of host name so we call `mono_get_address_info` up
to 3 times (IPv4, IPv6 and unspecified) in order to make the lookup faster.

Commit migrated from https://github.com/mono/mono/commit/10795da1c065c5349bf78ca4d39cfc7992fe5f7c

4 years ago[interp] Resolve tokens also from wrappers during optimization (mono/mono#17880)
Vlad Brezae [Mon, 25 Nov 2019 13:45:56 +0000 (15:45 +0200)]
[interp] Resolve tokens also from wrappers during optimization (mono/mono#17880)

[interp] Resolve tokens also from wrappers during optimization

Remove some duplicated code.

Commit migrated from https://github.com/mono/mono/commit/1ef7a298363799989bdfd08ef21bdcb367922c3b

4 years ago[eglib] In AIX/generic POSIX, handle null addresses better (mono/mono#17892)
Calvin Buckley [Mon, 25 Nov 2019 07:52:54 +0000 (03:52 -0400)]
[eglib] In AIX/generic POSIX, handle null addresses better (mono/mono#17892)

Fixes crash in tests on AIX.

Matches mono/mono@b0d966bd7ab887f075f6ce3d1c9858af6b52f30f

Commit migrated from https://github.com/mono/mono/commit/004fd06b967e0d2fd0e28bc516e8e33fe8a75a21

4 years agoAllow runtime to be built with C++ on AIX (mono/mono#17672)
Calvin Buckley [Sun, 24 Nov 2019 22:08:19 +0000 (18:08 -0400)]
Allow runtime to be built with C++ on AIX (mono/mono#17672)

* Allow runtime to be built with C++ on AIX

Many AIX/PPC/BE specific codepaths didn't do casting properly,
since C++ is much stricter than C about pointer typing.

Also specify the C99 format macros early as possible, since
inttypes may get included before eglib gets a chance to set the
macros.

Also explicitly use -pthread, since not using it is the cause of
many libstdc++ crashes on AIX.

* Oops, should have been that define instead

* Suggested changes from Jay on style

* Remove unneeded cast

Commit migrated from https://github.com/mono/mono/commit/a3f4def8b6951fed5c248bd1875e90a7ab748d15

4 years agoFix warnings in runtime build (mono/mono#17887)
Alexander Köplinger [Sat, 23 Nov 2019 23:16:20 +0000 (00:16 +0100)]
Fix warnings in runtime build (mono/mono#17887)

This makes at least the netcore-mono build in dotnet/runtime warning-free.

Commit migrated from https://github.com/mono/mono/commit/280e9d2423549d86686716f0818bcdbac9702ea1

4 years ago[netcore] Disable libmonoruntime-support.la build on netcore (mono/mono#17885)
Alexander Köplinger [Sat, 23 Nov 2019 09:06:40 +0000 (10:06 +0100)]
[netcore] Disable libmonoruntime-support.la build on netcore (mono/mono#17885)

It's used for the zlib compression helpers used by System.IO.Compression in the old Mono BCL.
On corefx this is handled differently so we don't need it there.

To avoid automake complaining about missing files we include the .c files directly instead of in the Makefile.

Commit migrated from https://github.com/mono/mono/commit/165e28ae97cc3bcac2f36484c74e8cd7c39ee9f1

4 years ago[netcore] Disable a test which fails on the interpreter
Alexander Köplinger [Fri, 22 Nov 2019 22:00:28 +0000 (23:00 +0100)]
[netcore] Disable a test which fails on the interpreter

The test is wrong, will be fixed upstream.

Commit migrated from https://github.com/mono/mono/commit/2ccd6122bd89533436e68198409350dc0fd99d3c

4 years agoMake configure.ac and Makefile.am work without certain directories (mono/mono#17871)
Alexander Köplinger [Fri, 22 Nov 2019 21:48:11 +0000 (22:48 +0100)]
Make configure.ac and Makefile.am work without certain directories (mono/mono#17871)

Those folders won't be present in the dotnet/runtime repository.

We need to factor out AC_OUTPUT into a separate file so we can continue
to share configure.ac but not the list of files in AC_OUTPUT since
autotools doesn't support variables there.

Commit migrated from https://github.com/mono/mono/commit/0d8037217302a15c83687dcd31e4136baf4f360f

4 years ago[runtime] Treat calling a non-virtual method through an open delegate the same as...
Zoltan Varga [Fri, 22 Nov 2019 19:34:14 +0000 (20:34 +0100)]
[runtime] Treat calling a non-virtual method through an open delegate the same as the normal case instead of the OPEN_VIRTUAL case. (mono/mono#17832)

* [runtime] Treat calling a non-virtual method through an open delegate the same as the normal case instead of the OPEN_VIRTUAL case.

* [amd64] Fix the return value mapping in gsharedvt calls, the return address might be in different registers in the caller and callee.

* [runtime] Always initialize del->method_ptr even if interp_method is set, JITted code makes calls through it in mixed mode.

* Add a test for open delegates and gsharedvt.

* Increase nrgctx trampoline count, its needed by dynamic-method-churn.exe in full-aot-interp mode.

Commit migrated from https://github.com/mono/mono/commit/93e46c687b0f1b3bea7ecf362a667f3bd8b59fc5

4 years agoReplace embedded libgc with Unity fork of recent Boehm (bdwgc) (mono/mono#16832)
Jonathan Chambers [Fri, 22 Nov 2019 17:03:23 +0000 (17:03 +0000)]
Replace embedded libgc with Unity fork of recent Boehm (bdwgc) (mono/mono#16832)

* Initial commit to migrate libgc to bdwgc.

* Fix submodule url

* Don't force Boehm as runtime wrapper. Was just for testing locally.

* Build bdwgc as single object file. Helps with performance.

* Bump bdwgc submodule.

* Attempt to fix windows MSVC based build.

* Remove any references to GC_INSIDE_DLL as we manually managed threads on Windows.

* Allow building boehm on Windows amd64.

* Allow building boehm with msvc.

* Bump bdwgc to fix line endings

* Use LF for sh, am, m4, and ac files in bdwgc

* Bump bdwgc & libatomic_ops with fix for cygwin

* Bump bdwgc to fix mono-boehm being generated as libtool wrapper script.

* Link the static boehm library.

* Revert "Link the static boehm library."

This reverts commit mono/mono@1f90d81d08308a61c890743cd79d8f6bb791bc99.

* Use --export-all-symbols on cygwin/mingw to fix issues with boehm causing mono symbols not to be exported.

* Fix

Commit migrated from https://github.com/mono/mono/commit/cc137237d86b0553e844dc34c9a3cbc5b20782c0

4 years agoFix InternalGetHashCode for boehm/null GC (mono/mono#17859)
Filip Navara [Fri, 22 Nov 2019 12:51:43 +0000 (13:51 +0100)]
Fix InternalGetHashCode for boehm/null GC (mono/mono#17859)

It was unintentionally shifting the address to left instead of right. It should be shift right to get rid of zeroes from the alignment.

Commit migrated from https://github.com/mono/mono/commit/0fed03ed63ed4ea742c4511d8edc3bc1c6f4044f

4 years agoFix automake warning
Alexander Köplinger [Fri, 22 Nov 2019 12:18:10 +0000 (13:18 +0100)]
Fix automake warning

Commit migrated from https://github.com/mono/mono/commit/5fedc42ccf6a182586280b6a324c221a7703b55f

4 years ago[wasm] Add aot+netcore support. (mono/mono#17798)
Zoltan Varga [Fri, 22 Nov 2019 11:42:55 +0000 (12:42 +0100)]
[wasm] Add aot+netcore support. (mono/mono#17798)

* [wasm] Compile corelib as 32 bit.

* [wasm] Add a netcore cross compiler.

* [wasm] Add netcore+aot support.

* [aot] Load the dedup module lazily instead of at startup.

The previous version didn't work on netcore because it couldn't resolve references to corlib etc.
since they were not yet registered in the root domain at the time of loading.

* Add more icall trampoline signatures.

* Add more netcore pinvokes.

* Fix the build.

* Add more icall trampoline signatures.

* Add an --embed-file option to the packager.

* Add more netcore pinvokes.

* [interp] Return FALSE for IsDynamicCodeCompiled.

* Add a netcore project and targets for running BenchmarkDotNet.

* Fix pinvoke tables.

* Add missing files.

* Fix make clean.

* Add a comment.

* Revert "Fix pinvoke tables."

This reverts commit mono/mono@998065e77614b7e621d2873510aabc3930669ca9.

* Make xunit-runner into a dotnet project. Add a netcore version.

Commit migrated from https://github.com/mono/mono/commit/ecd67a91c4bbeef046f8521da93f106da8fc0191

4 years ago[netcore] Run tests with Interpreter (mono/mono#17862)
Egor Bogatov [Fri, 22 Nov 2019 08:14:23 +0000 (11:14 +0300)]
[netcore] Run tests with Interpreter (mono/mono#17862)

* Run tests with interp and llvm

* fix build

* Run tests with llvm

* Add a few tests to rsp for LLVM

Commit migrated from https://github.com/mono/mono/commit/6a5c44b15e65782cac21673cd33138a92a6b6207

4 years agoRemove handles/MonoError from Mono.RuntimeGPtrArrayHandle. (mono/mono#17806)
Jay Krell [Thu, 21 Nov 2019 19:15:29 +0000 (11:15 -0800)]
Remove handles/MonoError from Mono.RuntimeGPtrArrayHandle. (mono/mono#17806)

Commit migrated from https://github.com/mono/mono/commit/dadc3d3117535ec44b887b8390181beda7b75b31

4 years agoStop referencing files from Mono BCL in Netcore System.Private.CoreLib (mono/mono...
Jo Shields [Thu, 21 Nov 2019 18:48:00 +0000 (19:48 +0100)]
Stop referencing files from Mono BCL in Netcore System.Private.CoreLib (mono/mono#17866)

Commit migrated from https://github.com/mono/mono/commit/9e00330af303bdc2b5f4ea2302e8bbcdfdc45524

4 years ago[interp] fix array_element_size intrinsic (mono/mono#17857)
Bernhard Urban-Forster [Thu, 21 Nov 2019 10:56:03 +0000 (11:56 +0100)]
[interp] fix array_element_size intrinsic (mono/mono#17857)

[interp] fix array_element_size intrinsic

`mono_class_array_element_size` gives us the size of the provided MonoClass if it would be an array element. But here we want the element size of a given array's MonoClass. That's what `mono_array_element_size` is returning.

This leads to all kind of weird crashes otherwise, specifically here:
https://github.com/mono/mono/blob/mono/mono@2c20649539ac16e069a65f2a750c793eb341e50f/netcore/System.Private.CoreLib/src/System/Array.Mono.cs#L64

Which is then used later to compute the size to memset in order to clear the content of an array. If it's larger than it should be, then this will cause memory corruption. For example this would crash eventually:

```csharp
using System;
using System.Collections.Generic;

namespace HelloWorld {
    class Program {
        static void Main(string[] args) {
            for (int j = 0; j < 0x1000; j++) {
                var d = new Dictionary<string, string> ();
                for (int i = 0; i < 197; i++)
                    d.Add (i + "", i + "foo");
                d.Clear ();
            }
        }
    }
}
```

Thanks to @EgorBo for reporting.

Commit migrated from https://github.com/mono/mono/commit/058d157648f593f41a2afbf878c0f401ecb804b7

4 years ago[metadata] Remove native threadpool from netcore build (mono/mono#17856)
Alexander Köplinger [Thu, 21 Nov 2019 09:01:39 +0000 (10:01 +0100)]
[metadata] Remove native threadpool from netcore build (mono/mono#17856)

We use the managed threadpool there.

Commit migrated from https://github.com/mono/mono/commit/b6e4ddd73c436ce23ae7036a7170f59b545e46b7

4 years ago[llvm] Recognize GEP (mono/mono#17844)
Egor Bogatov [Wed, 20 Nov 2019 20:25:24 +0000 (23:25 +0300)]
[llvm] Recognize GEP (mono/mono#17844)

* initial impl

* improve jit

* fix copy-paste

* disable for non-netcore

* clean up

Commit migrated from https://github.com/mono/mono/commit/2c20649539ac16e069a65f2a750c793eb341e50f

4 years ago[interp] Move the freeing of interp method info into a callback. (mono/mono#17852)
Zoltan Varga [Wed, 20 Nov 2019 18:56:09 +0000 (19:56 +0100)]
[interp] Move the freeing of interp method info into a callback. (mono/mono#17852)

[interp] Move the freeing of interp method info into a callback.

Commit migrated from https://github.com/mono/mono/commit/4961d411e7f6a425c4b7fad86b72f02947f9c817

4 years agoenable gss on android (mono/mono#17838)
Egor Bogatov [Wed, 20 Nov 2019 15:53:14 +0000 (18:53 +0300)]
enable gss on android (mono/mono#17838)

Commit migrated from https://github.com/mono/mono/commit/27e6e812f14163d1cb6555803148947043065386

4 years ago[debugger-agent] Fix CMD_VM_ALL_THREADS returning wrong value due to unitialized...
Alexander Köplinger [Wed, 20 Nov 2019 07:28:58 +0000 (08:28 +0100)]
[debugger-agent] Fix CMD_VM_ALL_THREADS returning wrong value due to unitialized variable (mono/mono#17847)

`remove_gc_finalizing` wasn't initialized so its value is undefined.
The `count_thread_check_gc_finalizer()` function only sets it in some cases so we could end up with the variable having an undefined (and probably not 0) value.
This resulted in the returned thread count from the debuggee being wrong because we'd decrement `count` even though we shouldn't.

Regression from https://github.com/mono/mono/pull/15618

Commit migrated from https://github.com/mono/mono/commit/0396037d31dacac06358faf30eb05d8e00b8b231

4 years ago[master] Update dependencies from dotnet/arcade dotnet/core-setup dotnet/corefx ...
dotnet-maestro[bot] [Tue, 19 Nov 2019 15:47:57 +0000 (16:47 +0100)]
[master] Update dependencies from dotnet/arcade dotnet/core-setup dotnet/corefx (mono/mono#17690)

* Update dependencies from https://github.com/dotnet/arcade build 20191102.1

- Microsoft.DotNet.Arcade.Sdk - 5.0.0-beta.19552.1
- Microsoft.DotNet.Helix.Sdk - 5.0.0-beta.19552.1

* Update dependencies from https://github.com/dotnet/core-setup build 20191104.1

- Microsoft.NETCore.App - 5.0.0-alpha.1.19554.1

* Update dependencies from https://github.com/dotnet/corefx build 20191103.8

- Microsoft.Private.CoreFx.NETCoreApp - 5.0.0-alpha.1.19553.8

* Update dependencies from https://github.com/dotnet/arcade build 20191108.11

- Microsoft.DotNet.Arcade.Sdk - 5.0.0-beta.19558.11
- Microsoft.DotNet.Helix.Sdk - 5.0.0-beta.19558.11

* Update dependencies from https://github.com/dotnet/core-setup build 20191109.5

- Microsoft.NETCore.App - 5.0.0-alpha.1.19559.5

* Update dependencies from https://github.com/dotnet/corefx build 20191109.6

- Microsoft.Private.CoreFx.NETCoreApp - 5.0.0-alpha.1.19559.6

* Update dependencies from https://github.com/dotnet/arcade build 20191117.2

- Microsoft.DotNet.Arcade.Sdk - 5.0.0-beta.19567.2
- Microsoft.DotNet.Helix.Sdk - 5.0.0-beta.19567.2

* Update dependencies from https://github.com/dotnet/core-setup build 20191114.1

- Microsoft.NETCore.App - 5.0.0-alpha.1.19564.1

* Update dependencies from https://github.com/dotnet/corefx build 20191113.6

- Microsoft.Private.CoreFx.NETCoreApp - 5.0.0-alpha.1.19563.6

* Use NETCoreApp version 5.0.0-alpha.1.19563.3

Later versions have an issue and the new dotnet/runtime assets aren't ready yet.

* Disable failing tests

* Fix merge error

Commit migrated from https://github.com/mono/mono/commit/a6db160596bb987af11fb2844d51c9a3a7c73133

4 years ago[interp] Add some missing netcore intrinsics. (mono/mono#17782)
Zoltan Varga [Mon, 18 Nov 2019 12:55:15 +0000 (13:55 +0100)]
[interp] Add some missing netcore intrinsics. (mono/mono#17782)

Commit migrated from https://github.com/mono/mono/commit/b8fe43da3ed89337c51dfa8f4e07955ed204761d

4 years agoBump corefx to pick up https://github.com/mono/corefx/pull/370 (mono/mono#17758)
Steve Pfister [Mon, 18 Nov 2019 11:13:16 +0000 (06:13 -0500)]
Bump corefx to pick up https://github.com/mono/corefx/pull/370 (mono/mono#17758)

Backport of https://github.com/dotnet/corefx/pull/34560

Added HAVE_LCHFLAGS & HAVE_STAT_FLAGS to configure.

Commit migrated from https://github.com/mono/mono/commit/0269f7b479f245c13b30c1b44cc91cc224c48b93

4 years ago[llvm] Add support for LLVM JIT on ARM64 (mono/mono#17827)
Filip Navara [Sun, 17 Nov 2019 23:24:11 +0000 (00:24 +0100)]
[llvm] Add support for LLVM JIT on ARM64 (mono/mono#17827)

* [llvm] Add support for LLVM JIT on ARM64

* Fix tabs

Commit migrated from https://github.com/mono/mono/commit/e2ccdac3f48d2c7726a69bf813df6992badfae01

4 years ago[mini] Fix NULL pointer crash in mono_decompose_vtype_opts(). (mono/mono#17801)
Nikolay Sivov [Fri, 15 Nov 2019 16:56:55 +0000 (19:56 +0300)]
[mini] Fix NULL pointer crash in mono_decompose_vtype_opts(). (mono/mono#17801)

* [mini] Fix NULL pointer crash in mono_decompose_vtype_opts().

All other call sites for mono_type_size() are already passing non-null pointer there.

* [mini] Rename variable dummy -> align.

Commit migrated from https://github.com/mono/mono/commit/18dd48cad81592f66ce7d21a5293c76d69864501

4 years ago[runtime] Add MONO_EMPTY_SOURCE_FILE to a couple of files. (mono/mono#17795)
Zoltan Varga [Fri, 15 Nov 2019 08:04:35 +0000 (09:04 +0100)]
[runtime] Add MONO_EMPTY_SOURCE_FILE to a couple of files. (mono/mono#17795)

* [runtime] Add MONO_EMPTY_SOURCE_FILE to a couple of files.

* Fix a typo.

* Revert unrelated changes.

Commit migrated from https://github.com/mono/mono/commit/25b7e6bd427f47ee08dbd2dd6333780da1ea97be

4 years agoRemove handles from System.Diagnostics.Debugger. (mono/mono#17713)
Jay Krell [Thu, 14 Nov 2019 07:18:02 +0000 (23:18 -0800)]
Remove handles from System.Diagnostics.Debugger. (mono/mono#17713)

* Remove handles from System.Diagnostics.Debugger.
memset (&x, 0, sizeof (type)) to memset (&x, 0, sizeof (x)).

* Bump API snapshot submodule

Commit migrated from https://github.com/mono/mono/commit/19a44cd32bbf4d993a0e833a76250e563828233f

4 years ago[interp] Add some inline checks from the JIT. (mono/mono#17784)
Zoltan Varga [Wed, 13 Nov 2019 21:34:01 +0000 (22:34 +0100)]
[interp] Add some inline checks from the JIT. (mono/mono#17784)

Commit migrated from https://github.com/mono/mono/commit/1d81534d27a01f23e0a75cee8097b565671080a1

4 years ago[runtime] Fix some handle usage in native-library.c (mono/mono#17781)
Zoltan Varga [Wed, 13 Nov 2019 17:12:09 +0000 (18:12 +0100)]
[runtime] Fix some handle usage in native-library.c (mono/mono#17781)

Commit migrated from https://github.com/mono/mono/commit/a0fcc29d5d67cf4f2e70072294b2ba4751324226

4 years agoRemove handles from ves_icall_System_Array_CanChangePrimitive. (mono/mono#17708)
Jay Krell [Wed, 13 Nov 2019 16:55:52 +0000 (08:55 -0800)]
Remove handles from ves_icall_System_Array_CanChangePrimitive. (mono/mono#17708)

Remove handles from ves_icall_System_Array_CanChangePrimitive.

Commit migrated from https://github.com/mono/mono/commit/73431403c7e7525122526eae016f80e9d9398d3e

4 years ago[debugger] Assert when async debug a generic method (mono/mono#17727)
Thays Grazia [Wed, 13 Nov 2019 15:29:33 +0000 (16:29 +0100)]
[debugger] Assert when async debug a generic method (mono/mono#17727)

* When we try to call a method to get the async_id to do an async debug and we are trying to do this in a generic method like this:

async Task<T> ExecuteAsync_Broken<T>()
 {
            await Task.Delay(2);
            return default;
  }

We need to inflate the generic type before call the method or we will get the error:  Could not execute the method because the containing type 'System.Runtime.CompilerServices.AsyncTaskMethodBuilder1[T_REF]’, is not fully instantiated.

Fixes mono/mono#17549
Fixes mono/mono#17569

Commit migrated from https://github.com/mono/mono/commit/dd18ec40dd60e9cde7732436f843942cb51605af

4 years ago[llvm] Use explicit null checks with LLVM. (mono/mono#16954)
imhameed [Wed, 13 Nov 2019 00:35:44 +0000 (01:35 +0100)]
[llvm] Use explicit null checks with LLVM. (mono/mono#16954)

* [llvm] Use explicit null checks with LLVM.

When the LLVM backend is used for a method, explicit null checks will
unconditionally be used at the mini IR level. We had previously used
implicit null checks at the mini IR level combined with llvm-volatile
loads and stores to avoid mis-optimization of the generated code.

During AOT compilation, LLVM's `ImplicitNullChecks` pass is enabled.
This conservatively elides these explicit null checks. This pass only
modifies branches tagged with `make.implicit` metadata.

Null check branches are now tagged with `make.implicit` metadata, with
three exceptions:

1. For platforms with backends that do not support zero page fault
handling, null check branches are not tagged with `make.implicit`.

2. This commit also adds a `MONO_DEBUG` option:
`llvm-disable-implicit-null-checks`. When it is enabled, null check
branches are not tagged with `make.implicit`.

3. This commit alters the behavior of the `explicit-null-checks`
`MONO_DEBUG` option. Enabling `explicit-null-checks` also implies
`llvm-disable-implicit-null-checks`, because this option is documented
(e.g. in https://www.mono-project.com/docs/debug+profile/debug/) to
completely disable signal/seh-based null checks.

* Add llvm-disable-implicit-null-checks to the MONO_DEBUG help string.

* Add llvm-disable-implicit-null-checks to mono.1.

* Don't tag null checks in EH regions with `make.implicit`.

This relocates the logic formerly used to guard individual LLVM-level
loads and stores with explicit null checks inside EH regions.

* Use atomic/volatile loads and stores for the terimation flag in thread-suspend-suspended.

LLVM was hoisting the load of and branch against `finished` out of
thread t1's loop, resulting in an infinite spin.

* Use ImplicitNullChecks with LLVM JIT.

* Suppress emission of the .llvm_faultmaps section.

Bump external/llvm to release_60
(mono/mono@7a8dc89adbe7e123220e070a527e096ee91e66d5).

* [ci] use LLVM built from repository

Commit migrated from https://github.com/mono/mono/commit/fe0f824db5fa1f3021aad6dc4a6eb428133d8a4e

4 years ago[wasm] Change netcore support to use a prebuilt corefx runtime. (mono/mono#17773)
Zoltan Varga [Wed, 13 Nov 2019 00:30:09 +0000 (01:30 +0100)]
[wasm] Change netcore support to use a prebuilt corefx runtime. (mono/mono#17773)

* [wasm] Change netcore support to use a prebuilt corefx runtime.

* [wasm] Build and test netcore support on CI.

* [wasm] Disable netcore tests on CI for now.

* [wasm] Add an error check.

* [wasm] Add a linker exclusion.

* [wasm] Add netcore runtimes to the nupkg.

Commit migrated from https://github.com/mono/mono/commit/eaf895f2d0ec8c6056ed6b4777579d2e08a8444d

4 years agoMirror changes from mono/corefx,coreclr (mono/mono#17777)
Filip Navara [Tue, 12 Nov 2019 17:46:00 +0000 (18:46 +0100)]
Mirror changes from mono/corefx,coreclr (mono/mono#17777)

* Rewrite the rest of the array non-generic fast paths in C# (mono/mono#27703)

* Rewrite the rest of the array non-generic fast paths in C#

* Add missing resource string

* Use CorElementType and handle native ints to match the unmanaged implementation

* Avoid try pattern for Sort and Reverse

* Match the native implementation more closely

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
* Rewrite Enum.CompareTo in C# (dotnet/coreclrmono/mono#27792)

* Rewrite Enum.CompareTo in C#

* Handle floating point too for match the existing native implementation

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
* Remove unused argument from GetRegionDisplayName (dotnet/coreclrmono/mono#27806)

* Remove unused argument from GetRegionDisplayName

* Address race conditions in CultureData

While most of the race conditions in this file are benign, in these cases, a thread could end up returning an intermediate value due to multiple potential writes to the same field.

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
* Fix StyleCop rule SA1129 ("Do not use default value type constructor") (mono/mono#27716)

Will enable the rule in corefx once these changes have propagated.

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
* Remove non-shared Enum.CompareTo/InternalCompareTo implementation, add new Array icalls used by shared code

Commit migrated from https://github.com/mono/mono/commit/bd278dd00dd24b3e8c735a4220afa6cb3ba317ee

4 years agoMove zlib to its own directory, and re-enable subdir-objects
Jo Shields [Tue, 12 Nov 2019 08:31:08 +0000 (09:31 +0100)]
Move zlib to its own directory, and re-enable subdir-objects

This has the following benefits:

* Clearer structure for mirroring
* Slight build time improvement on bundled zlib use (as it only builds once)
* Removes pages of subdir-objects spew from autogen

Commit migrated from https://github.com/mono/mono/commit/8bceb40878f3ff21cf3e521d0fe987a8fd6f1326

5 years agoDelete some LLVM test cases from `make dist` (mono/mono#17772)
Jo Shields [Mon, 11 Nov 2019 16:42:51 +0000 (17:42 +0100)]
Delete some LLVM test cases from `make dist` (mono/mono#17772)

Many of the tests from the llvm-project monorepo cause problems to `make dist`

Commit migrated from https://github.com/mono/mono/commit/c256fc86b6c011bcce691dd6f9415660c58306c5

5 years ago[netcore] Complete Monitor.LockContentionCount implementation (mono/mono#17719)
Alexis Christoforides [Mon, 11 Nov 2019 12:03:03 +0000 (13:03 +0100)]
[netcore] Complete Monitor.LockContentionCount implementation (mono/mono#17719)

* [netcore] Complete Monitor.LockContentionCount implementation

Fixes https://github.com/mono/mono/pull/16538

Commit migrated from https://github.com/mono/mono/commit/722c6f7dd59d235c2403c110869b74d8e63ac07d

5 years agoFixes build
Marek Safar [Sat, 9 Nov 2019 17:37:10 +0000 (18:37 +0100)]
Fixes build

Commit migrated from https://github.com/mono/mono/commit/d90030d2738dfc21dafa8de30d51b28bc83de0c3

5 years ago[netcore] Fix RuntimePropertyInfo.GetValue() in FullAOT scenarios (mono/mono#17757)
Alexander Köplinger [Fri, 8 Nov 2019 23:33:00 +0000 (00:33 +0100)]
[netcore] Fix RuntimePropertyInfo.GetValue() in FullAOT scenarios (mono/mono#17757)

It tried to use dynamic codegen which is not available in FullAOT.
We can use the `RuntimeFeature.IsDynamicCodeSupported` to check for this, added a JIT intrinsic which returns false in FullAOT.

Used a linker friendly way to check for TypeBuilder by adding an `internal virtual IsTypeBuilder` method in the `Type` class which returns false, and overwrote it in `TypeBuilder`.

Commit migrated from https://github.com/mono/mono/commit/ed2d5f69e0cf04242ac29dae851bf8ec38f2b651

5 years ago[wasm] Bump emscripten. (mono/mono#17748)
Zoltan Varga [Fri, 8 Nov 2019 19:43:32 +0000 (14:43 -0500)]
[wasm] Bump emscripten. (mono/mono#17748)

* [wasm] Bump emscipten to 1.39.2.

* Disable some code in libtest.c which causes emscripten to fail.

* [wasm] Source emscripten env before make install as well.

Commit migrated from https://github.com/mono/mono/commit/73b50890cc1fd2cb8ea4b7c209c03bfbd491359b

5 years agoImplement GC.GetGCMemoryInfo (mono/mono#17607)
Nathan Ricci [Fri, 8 Nov 2019 18:58:37 +0000 (13:58 -0500)]
Implement GC.GetGCMemoryInfo (mono/mono#17607)

Implement the GC.GetGCMemoryInfo API.

Commit migrated from https://github.com/mono/mono/commit/5b0a822afb97375ba90e0a6bcab53b4b4755b2c1

5 years ago[loader] Unmanaged library refactoring and NativeLibrary implementation (mono/mono...
Ryan Lucia [Fri, 8 Nov 2019 18:42:56 +0000 (13:42 -0500)]
[loader] Unmanaged library refactoring and NativeLibrary implementation (mono/mono#17369)

Commit migrated from https://github.com/mono/mono/commit/d5b6cf33b3d2a89189be4e6bddcf0207e9887ecd

5 years ago[netcore] Run individual CoreCLR test suites (mono/mono#17753)
Ryan Lucia [Fri, 8 Nov 2019 18:14:06 +0000 (13:14 -0500)]
[netcore] Run individual CoreCLR test suites (mono/mono#17753)

This is implemented in a lazy way for convenience when manually testing until the tests are wired up to xunit

Commit migrated from https://github.com/mono/mono/commit/68d142912d8a949c1874629391fb7d7b87fccd92