platform/upstream/dotnet/runtime.git
2 years agoRemove BB limit from importer_vectorization (#66534)
Egor Bogatov [Mon, 14 Mar 2022 21:05:00 +0000 (00:05 +0300)]
Remove BB limit from importer_vectorization (#66534)

2 years agoRemove 4146 from libunwind (#66427)
Aaron Robinson [Mon, 14 Mar 2022 21:00:38 +0000 (17:00 -0400)]
Remove 4146 from libunwind (#66427)

* Remove 4146 from libunwind

Use existing libunwind align macro and use casting
  when the operation's target should reflect a signed value.

2 years agoFix macOS ARM64 stack overflow detection (#66599)
Jan Vorlicek [Mon, 14 Mar 2022 20:50:46 +0000 (21:50 +0100)]
Fix macOS ARM64 stack overflow detection (#66599)

We were incorrectly using exception_data_type_t instead of mach_exception_data_type_t
in the PAL due to a missing change when we have added support for ARM64 macOS.

The wrong type was 32 bit and it caused the address of a hardware exception to have
the upper 32 bits cut off. That prevented us from detecting stack overflow correctly.

2 years ago[wasm] Use `npm` from emsdk when running `sanitize.py` (#66533)
Ankit Jain [Mon, 14 Mar 2022 19:02:48 +0000 (15:02 -0400)]
[wasm] Use `npm` from emsdk when running `sanitize.py` (#66533)

The issue manifests are build errors:
`UnhandledPromiseRejectionWarning: Error: Cannot find module 'is-fullwidth-code-point'`

.. eventually failing the build:
`*** No rule to make target 'src/cjs/runtime.cjs.iffe.js', needed by 'dotnet.js'`

2 years ago[Mono] Intrinsify Vector WidenLower and WidenUpper on Arm64 (#66512)
Simon Rozsival [Mon, 14 Mar 2022 18:44:47 +0000 (19:44 +0100)]
[Mono] Intrinsify Vector WidenLower and WidenUpper on Arm64 (#66512)

* Implement WidenLower and WidenUpper

* Restrict code only to Arm64

2 years agoARM64 - Optimizing a % b operations (#65535)
Will Smith [Mon, 14 Mar 2022 18:29:12 +0000 (11:29 -0700)]
ARM64 - Optimizing a % b operations (#65535)

* Initial work for ARM64 mod optimization

* Updated comment

* Updated comment

* Updated comment

* Fixing build

* Remove uneeded var

* Use '%' morph logic for both x64/arm64

* Adding back in divisor check for x64

* Formatting

* Update comments

* Update comments

* Fixing

* Updated comment

* Updated comment

* Tweaking x64 transformation logic for the mod opt

* Tweaking x64 transformation logic for the mod opt

* Using IntCon

* Fixed build

* Minor tweak

* Fixing x64 diffs

* Removing flag set

* Feedback

* Fixing build

* Feedback

* Fixing tests

* Fixing tests

* Fixing tests

* Formatting

* Fixing tests

* Feedback

* Fixing build

2 years agoFix JsonSerializer src-gen issues with reference handler feature (#66422)
Layomi Akinrinade [Mon, 14 Mar 2022 17:40:01 +0000 (10:40 -0700)]
Fix JsonSerializer src-gen issues with reference handler feature (#66422)

* Fix JsonSerializer src-gen issues with reference handler feature

* Add Newtonsoft.Json reference to src-gen test project

* Fix failing tests

2 years agoMake ConfigurationManager.Sources public (#66485)
fredjeck [Mon, 14 Mar 2022 17:24:58 +0000 (18:24 +0100)]
Make ConfigurationManager.Sources public (#66485)

* Make ConfigurationManager.Sources public

ConfigurationManager.Sources is now a public property rather than
an explicit interface implementation of  ConfigurationBuilder.Sources.

#61675

2 years agoAdd time prefix to superpmi.py output (#66575)
Bruce Forstall [Mon, 14 Mar 2022 16:00:26 +0000 (09:00 -0700)]
Add time prefix to superpmi.py output (#66575)

Also, add per-operation completion times for collection operations.

These two things will make it easier to see what is taking time
during SuperPMI runs, especially those run in the CI system.

2 years agoDo not assume containment (#66385)
SingleAccretion [Mon, 14 Mar 2022 10:40:48 +0000 (13:40 +0300)]
Do not assume containment (#66385)

Codegen was assuming lowering would always contain immediates
for RMW ops. It did not for some fuzzer-generated code.

Check for containment explicitly instead.

2 years agoReenable a test in stress mode (#66501)
Jakob Botsch Nielsen [Mon, 14 Mar 2022 10:05:03 +0000 (11:05 +0100)]
Reenable a test in stress mode (#66501)

The issue #35687 was closed in #59784 and I can confirm the original
repro mentioned there no longer reproduces with this test.

Fix #65934

2 years agoBetter fix for #66242 (#66335)
SingleAccretion [Mon, 14 Mar 2022 10:00:56 +0000 (13:00 +0300)]
Better fix for #66242 (#66335)

Lowering was moving nodes without checking legality first.

2 years agoStop generating CLS_VAR (#66298)
SingleAccretion [Mon, 14 Mar 2022 09:57:58 +0000 (12:57 +0300)]
Stop generating CLS_VAR (#66298)

2 years agoGet rid of fallback to process ids enumeration in ProcessManager.IsProcessRunning...
Evgeny Peshkov [Mon, 14 Mar 2022 09:01:19 +0000 (12:01 +0300)]
Get rid of fallback to process ids enumeration in ProcessManager.IsProcessRunning (Windows) (#65041)

* ProcessManager.IsRunning on Windows: Idle process handle couldn't be opened

* Peformance optimization in ProcessManager.IsRunning on Windows: Enumerate processes on local machine only when handle check failed

2 years agoFix typo in managed port of strong name validation (#66554)
Jan Kotas [Mon, 14 Mar 2022 02:55:27 +0000 (19:55 -0700)]
Fix typo in managed port of strong name validation (#66554)

Fixes #66545

2 years ago[mono][wasm] Initialize debugger_tls_id on WASM as well. (#66560)
Zoltan Varga [Mon, 14 Mar 2022 00:52:16 +0000 (20:52 -0400)]
[mono][wasm] Initialize debugger_tls_id on WASM as well. (#66560)

Some code paths are using mono_native_tls_get_value () instead of GET_DEBUGGER_TLS ().

2 years agoEnable Fast Tail Call Optimization for ARM32 (#66282)
Dong-Heon Jung [Sun, 13 Mar 2022 11:50:20 +0000 (20:50 +0900)]
Enable Fast Tail Call Optimization for ARM32 (#66282)

2 years agoFix setting the default value of WasmExceptionHandling. (#66548)
Zoltan Varga [Sun, 13 Mar 2022 06:40:43 +0000 (01:40 -0500)]
Fix setting the default value of WasmExceptionHandling. (#66548)

2 years ago[mono] Less MonoClassField arithmetic and less direct offset access (#66519)
Aleksey Kliger (λgeek) [Sun, 13 Mar 2022 01:54:51 +0000 (20:54 -0500)]
[mono] Less MonoClassField arithmetic and less direct offset access (#66519)

This is in preparation for adding the ability for hot reload to add instance fields to existing classes.

Mostly the PR does 3 things:

Add /* TODO: metadata-update */ comments to places where we will need to do work
Add g_assert (!m_field_is_from_update (field)) to places where we either don't expect to see updated fields (for example the JIT or AOT compilers), or places where we will need to do work for hot reload.
Switch from directly iterating over MonoClass:fields to using mono_class_get_fields_internal. For hot reload we will update the iterator method to also return added fields.
Avoid accessing MonoClassField:offset directly. Go via m_field_get_offset. This is primarily to make it easier to grep for all the places that access field offsets. (Hot reload added fields aren't stored with the object, so "offset" won't make sense - they will be in separate storage that is associated with the object instances)

* Cleanup some field index arithmetic

Or add comments why its ok.

The issue: we don't want to use MonoClassField* pointer artithmetic when there is a possibility that a field was added by an EnC update.

We also don't want to iterate over all the fields or all the fields' tokens by just doing computations based on MonoClassField* addresses and first_field_idx. Instead use mono_class_get_fields_internal which will be updated to iterate over the added fields, too

* Use m_field_get_offset in more places. Add metadata-update asserts and TODO

   In places where we access the offset directly, add TODO notes and asserts that we're not looking at an added field.

2 years agoThe problem was that the assert didn't take into account the fact that some regions...
Peter Sollich [Sat, 12 Mar 2022 17:11:44 +0000 (18:11 +0100)]
The problem was that the assert didn't take into account the fact that some regions may be on the decommit list already due to aging of regions. (#66495)

2 years agoReenable HostApplicationBuilderTests disabled on Mono/Windows (#66521)
Alexander Köplinger [Sat, 12 Mar 2022 15:27:04 +0000 (16:27 +0100)]
Reenable HostApplicationBuilderTests disabled on Mono/Windows (#66521)

They were disabled against https://github.com/dotnet/runtime/issues/34582 which was fixed by https://github.com/dotnet/runtime/pull/64834

2 years agoArm64: Always use SIMD features (#66411)
Kunal Pathak [Sat, 12 Mar 2022 04:35:06 +0000 (20:35 -0800)]
Arm64: Always use SIMD features  (#66411)

* Use supportSIMDTypes() throughout

* Make featureSIMD only for non-arm64

2 years agoRe-enable RevokeEndEntity_PolicyErrors_NotTimeValid
Kevin Jones [Sat, 12 Mar 2022 02:31:43 +0000 (21:31 -0500)]
Re-enable RevokeEndEntity_PolicyErrors_NotTimeValid

2 years agoFix compiler warning 4244 in host (#66518)
Elinor Fung [Sat, 12 Mar 2022 01:22:26 +0000 (17:22 -0800)]
Fix compiler warning 4244 in host (#66518)

2 years agoAdd Activity.HasRemoteParent (#66489)
Tarek Mahmoud Sayed [Fri, 11 Mar 2022 22:05:40 +0000 (14:05 -0800)]
Add Activity.HasRemoteParent (#66489)

Co-authored-by: Stephen Toub <stoub@microsoft.com>
2 years agoRevert "Avoid Attribute.GetCustomAttributes() returning null for open generic type...
Jan Kotas [Fri, 11 Mar 2022 21:12:20 +0000 (13:12 -0800)]
Revert "Avoid Attribute.GetCustomAttributes() returning null for open generic type (#65237)" (#66508)

This reverts commit 20294957616061616e672b43f1d962460b7f4234.

2 years agoAdd comment for use of DynamicDebugViewEmptyException.Empty (#66510)
Charles Stoner [Fri, 11 Mar 2022 21:12:00 +0000 (13:12 -0800)]
Add comment for use of DynamicDebugViewEmptyException.Empty (#66510)

2 years ago[WIP] Enable regions again. (#64688)
Manish Godse [Fri, 11 Mar 2022 21:11:33 +0000 (13:11 -0800)]
[WIP] Enable regions again. (#64688)

* Enable regions again.

* disable regions for MacOS for now.

* disable for NativeAOT for now

2 years agoAdd Android 32 to RID graph (#66022)
Steve Pfister [Fri, 11 Mar 2022 21:09:58 +0000 (16:09 -0500)]
Add Android 32 to RID graph (#66022)

Supports Android 12L, which will be stable most likely by summer or earlier.

Co-authored-by: Steve Pfister <steve.pfister@microsoft.com>
2 years agoWrite the exit code tag from the Apple app wrapper (#66511)
Přemek Vysoký [Fri, 11 Mar 2022 20:42:43 +0000 (21:42 +0100)]
Write the exit code tag from the Apple app wrapper (#66511)

Co-authored-by: Alexander Köplinger <alex.koeplinger@outlook.com>
2 years agoFix early JITDUMP disasm in ARM/ARM64 (#66502)
Jakob Botsch Nielsen [Fri, 11 Mar 2022 20:36:16 +0000 (21:36 +0100)]
Fix early JITDUMP disasm in ARM/ARM64 (#66502)

2 years agoEnable emitting intrinsics for System.Numerics.Vector<T> on arm64 (#66476)
Fan Yang [Fri, 11 Mar 2022 20:09:01 +0000 (15:09 -0500)]
Enable emitting intrinsics for System.Numerics.Vector<T> on arm64 (#66476)

2 years agoAdd OSR to DacpTieredVersionData (#66507)
Andy Ayers [Fri, 11 Mar 2022 19:57:23 +0000 (11:57 -0800)]
Add OSR to DacpTieredVersionData (#66507)

2 years ago[mono] Fixup name of define (#66523)
Alexander Köplinger [Fri, 11 Mar 2022 19:54:04 +0000 (20:54 +0100)]
[mono] Fixup name of define (#66523)

2 years agoMove to Windows.10.Amd64.Server2022.ES.Open (#66404)
Elinor Fung [Fri, 11 Mar 2022 19:48:05 +0000 (11:48 -0800)]
Move to Windows.10.Amd64.Server2022.ES.Open (#66404)

2 years ago[mono] Cleanup todo items from defines-todo.cmake (#66408)
Alexander Köplinger [Fri, 11 Mar 2022 19:43:58 +0000 (20:43 +0100)]
[mono] Cleanup todo items from defines-todo.cmake (#66408)

There are roughly three categories:

1. Already implemented or completely unused defines
2. Defines we can remove because they're no longer useful:
- `MAJOR_IN_MKDEV` and `MAJOR_IN_SYSMACROS`
Not used in the runtime anymore.

- `HAVE_EPOLL` and `HAVE_KQUEUE`
Not used in the runtime anymore.

- `ANDROID_UNIFIED_HEADERS` and `HAVE_ANDROID_NDK_VERSION_H`
We're requiring a recent Android NDK now that only supports unified headers.

- `MONO_DL_NEED_USCORE`
It doesn't seem needed anymore.
dlsym() on macOS automatically prepends an underscore and from what I can find it was mostly for old *BSDs.

- `GLIBC_BEFORE_2_3_4_SCHED_SETAFFINITY`
We don't support such an old glibc anymore.

- `HAVE_SIOCGIFCONF`
We don't use this and related networking code in the runtime anymore.

- `DISABLE_HW_TRAPS`
We have `MONO_ARCH_DISABLE_HW_TRAPS` instead.

- `MONO_BIG_ARRAYS`
No longer needed and probably broken due to missing BCL changes.

- `ENABLE_MONODROID` and replace `ENABLE_MONOTOUCH` with `TARGET_APPLE_MOBILE`
ENABLE_MONODROID was unused and MonoTouch was an earlier name for Xamarin.iOS, replace it with something more descriptive

- `MONO_JEMALLOC`
We haven't used it and don't plan to.

- `MONO_XEN_OPT`
Xen is not a primary scenario for Mono anymore so we probably don't need this optimization.

3. Defines which are still useful and are implemented by this PR
- Fix `MONO_KEYWORD_THREAD` build and implement `HAVE_TLS_MODEL_ATTR`

- Boehm/Null GC support

- `MONO_SMALL_CONFIG`

- `USE_MALLOC_FOR_MEMPOOLS`

- `ENABLE_DTRACE`

- `ENABLE_EXPERIMENT_TIERED` and remove dynamic experiments

- `ENABLE_OVERRIDABLE_ALLOCATORS`

- `MONO_ARCH_ILP32`

- `RISCV_FPABI`

- `__mono_ppc64__`

- `HAVE_ARMV5/6`. There is an open question about what we'll do with `HAVE_ARMV7`, added a TODO comment.

2 years ago[Mono] Intrinsify Vector Dot product and Sum on Arm64 (#66391)
Simon Rozsival [Fri, 11 Mar 2022 19:38:16 +0000 (20:38 +0100)]
[Mono] Intrinsify Vector Dot product and Sum on Arm64 (#66391)

* Add Dot intrinsic

* WIP saddvq/uaddvq

* Add Dot and Sum impl for Arm64

* Add faddv

* Fix the Dot and Sum implementations

* Cleanup

* Cleanup

* Fix type checking

* Fix type checking

* Fix indentation

* Fix edge case with single element vectors

* Improve implementation

2 years agoUse ApiCompatTask in ApiCompat.proj (#66073)
Viktor Hofer [Fri, 11 Mar 2022 17:51:22 +0000 (18:51 +0100)]
Use ApiCompatTask in ApiCompat.proj (#66073)

* Use ApiCompatTask in ApiCompat.proj

As ApiCompat now has a task that allows to run in-proc, use that instead
of shelving out. Also use items instead of properties to define
the attribute exclusion list.

* Update ApiCompat version

2 years agoAdd support for the new WASM Exception Handling feature (#66435)
Zoltan Varga [Fri, 11 Mar 2022 16:50:33 +0000 (11:50 -0500)]
Add support for the new WASM Exception Handling feature (#66435)

* [mono] Fix parsing options added by utils/options.h when using --response.

* [mono][jit] Add support for WASM EH.

WASM EH uses the newly added llvm catchpad instructions instead of
the old landingpad instructions. Add a --wasm-exceptions option
to control whenever the old JS style or the new WASM style code
is generated.

* Add support for the new WASM Exception Handling feature.

It is controlled by the $(WasmExceptionHandling) msbuild property.

Its only used in AOT mode, the interpreter mode wouldn't benefit from it.

The only file in the runtime which contains a throw or a catch is
llvm-runtime.cpp, so create 2 small static libraries from this file:
* libmono-wasm-eh-js - compiled to use JS style EH
* libmono-wasm-eh-wasm - compiled to use WASM style EH
This avoids the need to compile 2 versions of the whole runtime.

Requires emscripten 3.x.

* Add missing quote

* Update src/mono/mono/mini/CMakeLists.txt

Co-authored-by: Aleksey Kliger (λgeek) <akliger@gmail.com>
* Update src/mono/CMakeLists.txt

Co-authored-by: Aleksey Kliger (λgeek) <akliger@gmail.com>
Co-authored-by: Radek Doulik <radek.doulik@gmail.com>
Co-authored-by: Aleksey Kliger (λgeek) <akliger@gmail.com>
2 years agoUpdate dependencies from https://github.com/dotnet/linker build 20220310.1 (#66505)
dotnet-maestro[bot] [Fri, 11 Mar 2022 16:29:49 +0000 (17:29 +0100)]
Update dependencies from https://github.com/dotnet/linker build 20220310.1 (#66505)

Microsoft.NET.ILLink.Tasks
 From Version 7.0.100-1.22158.1 -> To Version 7.0.100-1.22160.1

Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
2 years ago[main] Update dependencies from dotnet/runtime-assets dotnet/roslyn-analyzers (#66504)
dotnet-maestro[bot] [Fri, 11 Mar 2022 16:29:18 +0000 (17:29 +0100)]
[main] Update dependencies from dotnet/runtime-assets dotnet/roslyn-analyzers (#66504)

* Update dependencies from https://github.com/dotnet/runtime-assets build 20220310.1

Microsoft.DotNet.CilStrip.Sources , System.ComponentModel.TypeConverter.TestData , System.Drawing.Common.TestData , System.IO.Compression.TestData , System.IO.Packaging.TestData , System.Net.TestData , System.Private.Runtime.UnicodeData , System.Runtime.Numerics.TestData , System.Runtime.TimeZoneData , System.Security.Cryptography.X509Certificates.TestData , System.Text.RegularExpressions.TestData , System.Windows.Extensions.TestData
 From Version 7.0.0-beta.22128.1 -> To Version 7.0.0-beta.22160.1

* Update dependencies from https://github.com/dotnet/roslyn-analyzers build 20220310.2

Microsoft.CodeAnalysis.NetAnalyzers
 From Version 7.0.0-preview1.22157.1 -> To Version 7.0.0-preview1.22160.2

Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
2 years agodon't run WriteChars_VeryLargeArray_DoesNotOverflow test on 32 bit platforms (#66503)
Adam Sitnik [Fri, 11 Mar 2022 15:28:34 +0000 (16:28 +0100)]
don't run WriteChars_VeryLargeArray_DoesNotOverflow test on 32 bit platforms (#66503)

2 years agoFix typo in comment (#66497)
JK Park [Fri, 11 Mar 2022 11:18:20 +0000 (20:18 +0900)]
Fix typo in comment (#66497)

thta => that

2 years agoAdd armv8l virtualized 32-bit ARM core detection (#66477)
Antoine Martin [Fri, 11 Mar 2022 09:29:12 +0000 (04:29 -0500)]
Add armv8l virtualized 32-bit ARM core detection (#66477)

2 years agoRefactor and improve MLL tests (#66332)
Vitek Karas [Fri, 11 Mar 2022 08:34:45 +0000 (00:34 -0800)]
Refactor and improve MLL tests (#66332)

* Remove duplicated code - I know that tests should be descriptive, but repeating 100 times that we want to capture output doesn't feel necessary.
* For some of them move more stuff into the shared test state - this improves perf as we avoid repeating the setup (copying files around) for each test case, we do it once for the entire class
I also changed some of the tests to "Theory" as it's easier to read that way.
  * `SDKLookup.cs` - moved most of the state in to the shared state to speed up the tests.

* Adding more cases into the "theory" tests
  * Most notably for the framework resolution I added variations on the TFM (which will be needed when we implement disable of MLL)
* Adding new tests mostly around "list runtimes" (various forms), "list sdks" (various forms) and errors (which also list runtimes or sdks)

* Ported all of the `MultiLevelLookupFramework` tests over to the `FrameworkResolution` and `DependencyResolutions` suites which have a more robust test infra and can run the same tests much faster. Along the way I added lot more variations on top of the existing tests:
  * `PerAssemblyVersionResolutionMultipleFrameworks.cs` - this is actually not an MLL test, but I moved it to the new infra to make it much faster
  * `MultipleHives.cs` - MLL framework resolution tests

For SDK resolution I kept the `MultiLevelSDKLookup.cs` just removed code duplication and added new variants.

For the core reviewers: I promise I didn't remove any single test case in spirit with these exceptions:

* We had tests which validated that framework resolution is not affected by frameworks in current directory and also by frameworks in the user's directory. I left some basic test for the current directory check, but I completely removed the user's directory variant as the product simply doesn't have any code around that anymore.

* Remove "user" directory from SDK resolution tests

The product simply doesn't have any knowledge about user directory SDK lookup, so there's little value in testing for it (even though the test expects it to not have any impact).

Co-authored-by: Elinor Fung <elfung@microsoft.com>
2 years agoUpdate repo CMake configuration to target /W4 by default (#66474)
Aaron Robinson [Fri, 11 Mar 2022 04:59:21 +0000 (23:59 -0500)]
Update repo CMake configuration to target /W4 by default (#66474)

* Update repo CMake configuration to target /W4 by default

Disable all failing warnings.

* Mono API tests only use W3

2 years ago[metadata] Avoid realloc and memcpy when >1 gparams on generics (#66458)
Aleksey Kliger (λgeek) [Fri, 11 Mar 2022 04:21:01 +0000 (23:21 -0500)]
[metadata] Avoid realloc and memcpy when >1 gparams on generics (#66458)

* [metadata] Avoid realloc and memcpy when >1 gparams on generics

* Read the first gparam row again at the beginning of the second pass

2 years agoUpdate dependencies from https://github.com/dotnet/linker build 20220308.1 (#66390)
dotnet-maestro[bot] [Fri, 11 Mar 2022 00:33:51 +0000 (00:33 +0000)]
Update dependencies from https://github.com/dotnet/linker build 20220308.1 (#66390)

[main] Update dependencies from dotnet/linker

2 years agoOne-shot decrypt in to user buffers
Kevin Jones [Fri, 11 Mar 2022 00:17:07 +0000 (19:17 -0500)]
One-shot decrypt in to user buffers

This changes the one-shot decryption to allow decrypting directly in to user-supplied buffers, even in the presence of padding, so that a copy and CryptoPool rental can be avoided.

If the buffer is large enough to accommodate the padding, then decrypt directly in to the supplied buffer, zero the padding, and indicate the de-padded amount in bytesWritten.

If the buffer is possibly large enough, then decrypt all but the last block in to the buffer, and do the last block on the stack. If the de-padded amount in the stack is small enough, then copy the remaining. If it is too large, then zero everything in the user buffer.

2 years agoAdd table of testing configurations (#66347)
Andy Gocke [Thu, 10 Mar 2022 23:48:39 +0000 (15:48 -0800)]
Add table of testing configurations (#66347)

* Add table of testing configurations

* Remove "cross" from specification

2 years agoOverhaul how regex generated code is structured (#66432)
Stephen Toub [Thu, 10 Mar 2022 23:42:19 +0000 (18:42 -0500)]
Overhaul how regex generated code is structured (#66432)

* Overhaul how regex generated code is structured

* Address PR feedback

* Change parser to return RegexMethod

Clean up how the data is structured.

* Add a test for same method name in multiple types

2 years agoUpdating Unicode files to use 14.0.0 (#66362)
Jose Perez Rodriguez [Thu, 10 Mar 2022 22:28:35 +0000 (14:28 -0800)]
Updating Unicode files to use 14.0.0 (#66362)

2 years ago[Mono] Intrinsify Narrow for for Vector{64, 128} on Arm64 (#66409)
Fan Yang [Thu, 10 Mar 2022 22:00:05 +0000 (17:00 -0500)]
[Mono] Intrinsify Narrow for for Vector{64, 128} on Arm64 (#66409)

* Add arm64 intrinsics for Narrow

* Fix integer types

* Fix more issues with integer types

* Add some comments

* Nit

* Change var names

* Fix format

* Fix format again

2 years ago[Mono] Add intrinsics for several operators (#66317)
Meri Khamoyan [Thu, 10 Mar 2022 21:56:07 +0000 (01:56 +0400)]
[Mono] Add intrinsics for several operators (#66317)

This PR contributes to  #64072

Added intrinsics for following operators
- Operator *
- Operator /
- Operator |
- Operator &
- Operator ^
- Unary operator -
- Unary operator ~

Co-authored-by: Fan Yang <yangfan@microsoft.com>
2 years ago[mono] Stop setting time_date_stamp field in MonoImage (#66452)
Alexander Köplinger [Thu, 10 Mar 2022 21:47:57 +0000 (22:47 +0100)]
[mono] Stop setting time_date_stamp field in MonoImage (#66452)

We never read that field in the runtime and it was causing unnecessary disk IO during startup (5ms on my iOS device).
It was also never set on Windows already.

2 years agoEnable/fix compiler warning 4996 in host (#66431)
Elinor Fung [Thu, 10 Mar 2022 21:23:56 +0000 (13:23 -0800)]
Enable/fix compiler warning 4996 in host (#66431)

2 years agoUpdate dependencies from https://github.com/dotnet/msquic build 20220310.2 (#66468)
dotnet-maestro[bot] [Thu, 10 Mar 2022 20:58:09 +0000 (21:58 +0100)]
Update dependencies from https://github.com/dotnet/msquic build 20220310.2 (#66468)

System.Net.MsQuic.Transport
 From Version 7.0.0-alpha.1.22159.1 -> To Version 7.0.0-alpha.1.22160.2

Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
2 years agoRemove null check that is used later to handle size request scenario. (#66428)
Aaron Robinson [Thu, 10 Mar 2022 20:02:16 +0000 (15:02 -0500)]
Remove null check that is used later to handle size request scenario. (#66428)

2 years agoUse StartsWith(..., OrdinalIgnoreCase) in RegexCompiler / source generator (#66339)
Stephen Toub [Thu, 10 Mar 2022 19:11:56 +0000 (14:11 -0500)]
Use StartsWith(..., OrdinalIgnoreCase) in RegexCompiler / source generator (#66339)

* Use StartsWith(..., OrdinalIgnoreCase) in RegexCompiler / source generator

When we encounter a sequence of sets representing case-insensitive ASCII, we can simplify the code generated to just call StartsWith, which both makes it more readable but also takes advantage of the new JIT optimization to lower that into efficient vectorized comparisons based on the supplied literal.

This also cleans up some formatting in the source generator emitted code to make things much more concise and less noisy.

* Address PR feedback

2 years agoCheck gtClassProfileCandidateInfo before instrumentation (#66388)
Egor Bogatov [Thu, 10 Mar 2022 18:40:34 +0000 (21:40 +0300)]
Check gtClassProfileCandidateInfo before instrumentation (#66388)

2 years agoDisable timing out test on mono (#66381)
Eirik Tsarpalis [Thu, 10 Mar 2022 18:40:14 +0000 (18:40 +0000)]
Disable timing out test on mono (#66381)

* disable timing out test on mono

* Update src/libraries/System.Text.Json/tests/System.Text.Json.Tests/Serialization/CustomConverterTests/CustomConverterTests.cs

2 years agoClean up NonBacktracking DgmlWriter (#66363)
Stephen Toub [Thu, 10 Mar 2022 18:33:14 +0000 (13:33 -0500)]
Clean up NonBacktracking DgmlWriter (#66363)

* Clean up NonBacktracking DgmlWriter

* Address PR feedback, clean up some lang=regex usage, fix a test failure

2 years ago[main] Update dependencies from 7 repositories (#66330)
dotnet-maestro[bot] [Thu, 10 Mar 2022 18:10:24 +0000 (19:10 +0100)]
[main] Update dependencies from 7 repositories (#66330)

* Update dependencies from https://github.com/dotnet/arcade build 20220307.6

Microsoft.DotNet.XUnitConsoleRunner , Microsoft.DotNet.CodeAnalysis , Microsoft.DotNet.Build.Tasks.Workloads , Microsoft.DotNet.Build.Tasks.Templating , Microsoft.DotNet.Build.Tasks.TargetFramework , Microsoft.DotNet.Build.Tasks.Packaging , Microsoft.DotNet.Build.Tasks.Installers , Microsoft.DotNet.Build.Tasks.Feed , Microsoft.DotNet.Build.Tasks.Archives , Microsoft.DotNet.Arcade.Sdk , Microsoft.DotNet.ApiCompat , Microsoft.DotNet.XUnitExtensions , Microsoft.DotNet.GenAPI , Microsoft.DotNet.VersionTools.Tasks , Microsoft.DotNet.GenFacades , Microsoft.DotNet.SharedFramework.Sdk , Microsoft.DotNet.RemoteExecutor , Microsoft.DotNet.PackageTesting , Microsoft.DotNet.Helix.Sdk
 From Version 2.5.1-beta.22154.3 -> To Version 2.5.1-beta.22157.6

* Update dependencies from https://github.com/dotnet/icu build 20220307.1

Microsoft.NETCore.Runtime.ICU.Transport
 From Version 7.0.0-preview.3.22128.1 -> To Version 7.0.0-preview.3.22157.1

* Update dependencies from https://github.com/dotnet/xharness build 20220307.2

Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit
 From Version 1.0.0-prerelease.22153.2 -> To Version 1.0.0-prerelease.22157.2

* Update dependencies from https://github.com/dotnet/roslyn-analyzers build 20220307.1

Microsoft.CodeAnalysis.NetAnalyzers
 From Version 7.0.0-preview1.22153.1 -> To Version 7.0.0-preview1.22157.1

* Update dependencies from https://github.com/dotnet/hotreload-utils build 20220307.2

Microsoft.DotNet.HotReload.Utils.Generator.BuildTool
 From Version 1.0.2-alpha.0.22128.2 -> To Version 1.0.2-alpha.0.22157.2

* Update dependencies from https://github.com/dotnet/llvm-project build 20220307.1

runtime.win-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.win-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.osx.10.12-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.osx.11.0-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-x64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-musl-arm64.Microsoft.NETCore.Runtime.ObjWriter , runtime.linux-arm64.Microsoft.NETCore.Runtime.ObjWriter
 From Version 1.0.0-alpha.1.22128.1 -> To Version 1.0.0-alpha.1.22157.1

* Use full names for GenericHandler passed to xharness

* Update dependencies from https://github.com/dotnet/msquic build 20220309.1

System.Net.MsQuic.Transport
 From Version 7.0.0-alpha.1.21529.3 -> To Version 7.0.0-alpha.1.22159.1

* Update dependencies from https://github.com/dotnet/xharness build 20220310.1

Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit
 From Version 1.0.0-prerelease.22153.2 -> To Version 1.0.0-prerelease.22160.1

Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Co-authored-by: Ankit Jain <radical@gmail.com>
2 years agoStop collecting disposables if changed token encountered (#66265)
Pavel Ivanov [Thu, 10 Mar 2022 16:49:34 +0000 (21:49 +0500)]
Stop collecting disposables if changed token encountered (#66265)

* reproduce issue: collecting disposables even if composite token has alredy been changed

* stop collecting disposables if composite token has already been changed

* Update src/libraries/Microsoft.Extensions.Primitives/tests/CompositeChangeTokenTest.cs

Co-authored-by: Eric Erhardt <eric.erhardt@microsoft.com>
* Update src/libraries/Microsoft.Extensions.Primitives/tests/CompositeChangeTokenTest.cs

Co-authored-by: Eric Erhardt <eric.erhardt@microsoft.com>
Co-authored-by: Eric Erhardt <eric.erhardt@microsoft.com>
2 years agoRemove asserts checking OpenSSL error queues (#66443)
Radek Zikmund [Thu, 10 Mar 2022 14:45:04 +0000 (15:45 +0100)]
Remove asserts checking OpenSSL error queues (#66443)

Fixes #44689

As of #65148, libraries use different approach to handling OpenSSL errors. The original assert which would be hit if a previous operation left errors in the queue is no longer necessary as all OpenSSL calls are prepended by `ERR_clear_error()` to make sure the latest (and most relevant) error is reported.

2 years agoFix the accounting for gen 1 allocations by taking into account what has been already...
Peter Sollich [Thu, 10 Mar 2022 10:26:41 +0000 (11:26 +0100)]
Fix the accounting for gen 1 allocations by taking into account what has been already consumed in this GC. (#60248)

Details:

In gen 1 GCs, we recompute the budget at the end of the GC. However, the amount consumed in the current GC is nowhere taken into account. That causes us to do at most alternate gen 0 and gen 1 GCs, we cannot do back-to-back gen 1 GCs.

If we take the gen 1 budget as the amount of memory we are aiming to be used for gen 1, then it seems illogical to just drop the amount used in this GC on the floor.

The fix is to actually subtract the amount of gen 1 budget used in this GC from dd_new_allocation for gen 1. This is a bit awkward for server GC, because we even out the budget and the amount to subtract from dd_new_allocation across heaps after having computed it for each heap individually.

2 years agoFix implicit nullchecks for floats on arm64 (#66413)
Egor Bogatov [Thu, 10 Mar 2022 09:45:58 +0000 (12:45 +0300)]
Fix implicit nullchecks for floats on arm64 (#66413)

2 years agoAdd timestamp to some superpmi.py output (#66437)
Bruce Forstall [Thu, 10 Mar 2022 07:26:59 +0000 (23:26 -0800)]
Add timestamp to some superpmi.py output (#66437)

This should help determine where time is being spent in long-running
tasks such as PMI or crossgen2 collections in the CI system.

2 years agoAdd retries to `dotnet restore` in SPMI collection of benchmarks (#66433)
Bruce Forstall [Thu, 10 Mar 2022 06:18:55 +0000 (22:18 -0800)]
Add retries to `dotnet restore` in SPMI collection of benchmarks (#66433)

Also, run a `dotnet --info` command to verify the version of the CLI we are running.

2 years agoRevert "Do not mark implicit byrefs with GTF_GLOB_REF" (#66415)
Jakob Botsch Nielsen [Thu, 10 Mar 2022 05:10:10 +0000 (06:10 +0100)]
Revert "Do not mark implicit byrefs with GTF_GLOB_REF" (#66415)

* Revert "Do not mark implicit byrefs with GTF_GLOB_REF (#66266)"

This reverts commit 0a684281f661f5520ad7f358a5056a2eed5b50ba. This is
currently not safe to do in the JIT because we clear out address
exposure information in fgRetypeImplicitByRefArgs() and thus do not have
the necessary information in morph to know if indirections off of
implicit byrefs are global or not.

* Add a test

2 years agoAdd test for comhost with managed host (#66360)
Elinor Fung [Thu, 10 Mar 2022 04:52:18 +0000 (20:52 -0800)]
Add test for comhost with managed host (#66360)

2 years agoFix timeout checking in FindFinalStatePosition (#66365)
Stephen Toub [Thu, 10 Mar 2022 03:30:04 +0000 (22:30 -0500)]
Fix timeout checking in FindFinalStatePosition (#66365)

Exit the inner loop every now and then to do a timeout check.

2 years agoDowngrade the Roslyn version used by the Regex source generator to be 4.0.1 instead...
Jose Perez Rodriguez [Thu, 10 Mar 2022 03:28:15 +0000 (19:28 -0800)]
Downgrade the Roslyn version used by the Regex source generator to be 4.0.1 instead of 4.2.0 (#66423)

* Downgrade the Roslyn version used by the Regex source generator to be 4.0.1 instead of 4.2.0

* Update src/libraries/System.Text.RegularExpressions/gen/System.Text.RegularExpressions.Generator.csproj

Co-authored-by: Jeremy Koritzinsky <jkoritzinsky@gmail.com>
Co-authored-by: Jeremy Koritzinsky <jkoritzinsky@gmail.com>
2 years agoImplement System.Decimal.Scale (#66403)
MichalPetryka [Thu, 10 Mar 2022 03:26:36 +0000 (04:26 +0100)]
Implement System.Decimal.Scale (#66403)

Adds System.Decimal.Scale, a property that
returns the scaling factor of the decimal.

Closes #65074.

2 years agoRename `GeneratedDllImportAttribute` -> `LibraryImportAttribute` (#66307)
Elinor Fung [Thu, 10 Mar 2022 02:07:52 +0000 (18:07 -0800)]
Rename `GeneratedDllImportAttribute` -> `LibraryImportAttribute` (#66307)

2 years agoReenable compiler warning 4244 under coreclr/ (#66398)
Aaron Robinson [Thu, 10 Mar 2022 01:12:51 +0000 (20:12 -0500)]
Reenable compiler warning 4244 under coreclr/ (#66398)

* From vm/

* From util/

* From debug/

* GCInfo related

Use MAXDWORD instead of DWORD_MAX.
Remove DWORD_MAX from pal, since it is not used anywhere.

2 years agoFix random printf from JIT (#66420)
Jakob Botsch Nielsen [Thu, 10 Mar 2022 01:08:44 +0000 (02:08 +0100)]
Fix random printf from JIT (#66420)

2 years agoRun superpmi-replay pipeline on JIT PRs (#66063)
Bruce Forstall [Thu, 10 Mar 2022 01:01:21 +0000 (17:01 -0800)]
Run superpmi-replay pipeline on JIT PRs (#66063)

Currently, the `runtime-coreclr superpmi-replay` pipeline is run
when a JIT change is merged. This change moves the checking to happen
on the PR pre-merge, as a requirement for merging.

2 years ago[mono][wasm] Add changes required by emscripten 3.1.4. (#66410)
Zoltan Varga [Thu, 10 Mar 2022 00:27:01 +0000 (19:27 -0500)]
[mono][wasm] Add changes required by emscripten 3.1.4. (#66410)

2 years agochange the region related config names to be more reasonable (#66134)
Maoni Stephens [Thu, 10 Mar 2022 00:17:01 +0000 (16:17 -0800)]
change the region related config names to be more reasonable (#66134)

Co-authored-by: Maoni0 <maoni@microsoft.com>
2 years ago[MAUI][PERF] Add IPA file usage for iOS Maui SOD (#66358)
Parker Bibus [Wed, 9 Mar 2022 22:55:02 +0000 (16:55 -0600)]
[MAUI][PERF] Add IPA file usage for iOS Maui SOD (#66358)

* [PERF][MAUI] Add ios-arm64 logic to create .ipa for perf runs

Switches from simulator builds to device builds so we can track the size for the .ipa.

We generate a dummy codesigning cert and provisioning profile since device builds require that.

* First run with new ipa compatible ios perf setup and ipa app passthrough.

* Temporarily use preview 14 until all packages are synced for 6.0.300 for testing.

* Changed MacCatalyst folder copied on perf setup and made the mono ios app have it's subfolders copied.

* Update the calling args so xcopy knows the files are directories.

Co-authored-by: Alexander Köplinger <alex.koeplinger@outlook.com>
2 years agoDon't build IJW and COM tests for mono (#66400)
Elinor Fung [Wed, 9 Mar 2022 22:16:17 +0000 (14:16 -0800)]
Don't build IJW and COM tests for mono (#66400)

2 years agoExtend Equals/StartsWith auto-vectorization for OrdinalIgnoreCase (#66095)
Egor Bogatov [Wed, 9 Mar 2022 21:14:37 +0000 (00:14 +0300)]
Extend Equals/StartsWith auto-vectorization for OrdinalIgnoreCase (#66095)

2 years agoskip the ReadToEndAsync_WithCancellation test if there is not enough disk space ...
Adam Sitnik [Wed, 9 Mar 2022 19:26:09 +0000 (20:26 +0100)]
skip the ReadToEndAsync_WithCancellation test if there is not enough disk space (#66397)

* skip the ReadToEndAsync_WithCancellation test if there is not enough disk space to create 1GB test file

* move the test to OuterLoop

2 years agoUpdate CancellationToken.cs (#66399)
Aaron Robinson [Wed, 9 Mar 2022 16:16:21 +0000 (11:16 -0500)]
Update CancellationToken.cs (#66399)

Remove ObjectDisposedException from doc since the method doesn't throw the exception.

2 years ago[mono][jit] Add a helper function to check that a method has the [Intrinsic] custom...
Zoltan Varga [Wed, 9 Mar 2022 15:57:41 +0000 (10:57 -0500)]
[mono][jit] Add a helper function to check that a method has the [Intrinsic] custom attribute. (#66359)

2 years agoMove Array.CreateInstance methods to shared CoreLib (#66025)
Jan Kotas [Wed, 9 Mar 2022 15:56:10 +0000 (07:56 -0800)]
Move Array.CreateInstance methods to shared CoreLib (#66025)

2 years agoReduce the amount of memory allocated by System.IO.Tests (#66387)
Adam Sitnik [Wed, 9 Mar 2022 14:46:58 +0000 (15:46 +0100)]
Reduce the amount of memory allocated by System.IO.Tests (#66387)

* use 2.1 GB instead 6.5 GB to verify lack of Int32 overflow:

reduce the size
use same memory for input and output

* use SkipTestException to indicate that the test has been skipped

* don't run WriteChars_VeryLargeArray_DoesNotOverflow with other tests in parallel, as it can cause OOM

* move the test to Outerloop as suggested by Stephen

2 years ago[Mono] Add SIMD intrinsics for Vector64/128 "All" and "Any" variants of GT/GE/LT...
Simon Rozsival [Wed, 9 Mar 2022 14:24:14 +0000 (15:24 +0100)]
[Mono] Add SIMD intrinsics for Vector64/128 "All" and "Any" variants of GT/GE/LT/LE (#65889)

* Add All and Any variants of GT/GE/LT/LE

* Fix indentation

* Fix indentation

* Add xones to llvm

* Fix comparisons for floating point numbers

* Add missing unsigned type

* Reduce code duplication

* Fix build

* Reorganize code a bit

* Remove redundant assertion

* Fix bitwise operators intrinsics

* Simplify checking float args

* Fix typo

2 years agoRemove compiler warning suppression 2 (#66375)
Aaron Robinson [Wed, 9 Mar 2022 14:10:36 +0000 (09:10 -0500)]
Remove compiler warning suppression 2 (#66375)

* Remove 4146 - coreclr

* Remove 4302

2 years agoFix early stack offset and size computations for ARM32 with FEATURE_FASTTAILCALL...
Jakob Botsch Nielsen [Wed, 9 Mar 2022 12:59:09 +0000 (13:59 +0100)]
Fix early stack offset and size computations for ARM32 with FEATURE_FASTTAILCALL enabled (#66367)

When FEATURE_FASTTAILCALL is enabled we compute stack offsets for parameters early. These are used to check for interference when placing arguments for fast tailcalls. On ARM32 the assigned offsets were wrong in several cases involving alignment and when we have split parameters.

2 years agoAdd Stopwatch.GetElapsedTime (#66372)
Stephen Toub [Wed, 9 Mar 2022 12:42:15 +0000 (07:42 -0500)]
Add Stopwatch.GetElapsedTime (#66372)

2 years agoFix cancellation race condition in PipeStream cancellation callback (#65909)
Stephen Toub [Wed, 9 Mar 2022 11:49:16 +0000 (06:49 -0500)]
Fix cancellation race condition in PipeStream cancellation callback (#65909)

* Fix cancellation race condition in PipeStream cancellation callback

The public PipeStream.SafePipeHandle property throws an exception if the handle has already been closed.  This code should have been using the internal InternalHandle property, which just gets the SafePipeHandle object if it exists.

* Update src/libraries/System.IO.Pipes/src/System/IO/Pipes/PipeStream.ValueTaskSource.cs

Co-authored-by: Adam Sitnik <adam.sitnik@gmail.com>
Co-authored-by: Adam Sitnik <adam.sitnik@gmail.com>
2 years agoFirst try at implementing newly suggested ctor for providing tighter control over...
brianmed [Wed, 9 Mar 2022 10:56:44 +0000 (05:56 -0500)]
First try at implementing newly suggested ctor for providing tighter control over charset (#63231)

* First try at implementing newly suggested ctor for providing tighter control over charset

* Update reference assembly so mediaType is accurate

* Update per github code review

2 years agoDo not mark implicit byrefs with GTF_GLOB_REF (#66266)
Jakob Botsch Nielsen [Wed, 9 Mar 2022 10:22:25 +0000 (11:22 +0100)]
Do not mark implicit byrefs with GTF_GLOB_REF (#66266)

2 years agoFix some exception messages for BindHandle-like cases (#66326)
Koundinya Veluri [Wed, 9 Mar 2022 02:35:55 +0000 (18:35 -0800)]
Fix some exception messages for BindHandle-like cases (#66326)

* Fix some exception messages for BindHandle-like cases

- Updated `ThrowHelper.ThrowApplicationException` to get and include an error message
- Enabled the disabled tests, disabled a `BindHandle` test on Unixes where it now throws `PNSE` before `ArgumentNullException`

Fixes https://github.com/dotnet/runtime/issues/66273
Fixes https://github.com/dotnet/runtime/issues/66274

2 years agoDo not propagate RHS flags in block morphing (#66291)
SingleAccretion [Wed, 9 Mar 2022 02:18:54 +0000 (05:18 +0300)]
Do not propagate RHS flags in block morphing (#66291)

It is not necessary to copy flags from the source node to the field nodes
when doing a field-by-field copy in block morphing. In fact, it can be a
pessimization in cases where the source node had NO_CSE set, which would
block constant propagation.

2 years agoRemove compiler warning suppression (#66234)
Aaron Robinson [Wed, 9 Mar 2022 00:57:12 +0000 (19:57 -0500)]
Remove compiler warning suppression (#66234)

* Remove 4996

* Remove 4701

* Remove 4611

* Remove 4610

* Remove 4510

* Remove 4267

* Remove 4267 - libunwind

2 years agoPartial revert of Maui Install Fix #65904 (#66221)
Parker Bibus [Tue, 8 Mar 2022 23:57:38 +0000 (17:57 -0600)]
Partial revert of Maui Install Fix #65904 (#66221)

* Prep for test build run.

* Replace the rollback file in the install call as the emscription line has been added to the downloaded json.

* Revert "Prep for test build run."

This reverts commit 2d7cca803e1cb6b79f83583b4dfa9846be3cf95e.

* Fix comment detailing Install Maui Workload setup.

2 years agoFix NonBacktracking quadratic behavior with deep loops (#66038)
Olli Saarikivi [Tue, 8 Mar 2022 21:30:46 +0000 (13:30 -0800)]
Fix NonBacktracking quadratic behavior with deep loops (#66038)

Derivative construction now accumulates a list of concatenands and
builds concatenations making up a part of a derivative just once.
Repeated concatenation to the right side in an interaction of the
derivation rule for concatenations and loops was causing quadratic
rebuilding of concatenations.

To implement this fix TransitionRegex is no longer used and instead both
normal and effectful derivative variants use a monolithic base function
that implements the rules for regex derivatives. This also gets rid of
the additional inefficiency of allocating TransitionRegex trees.

Both kinds of derivatives now support simulating backtracking behavior
and always use OrderedOr nodes to maintain transition ordering. The
second reverse phase of matching, however, needs to not use backtracking
simulation. A new DisableBacktrackingSimulation is used to signal this.

Other notable fixes:
The eager concept in derivatives is renamed to simulateBacktracking.
Put TransitionRegex and SymbolicNFA behind DEBUG.
Fix ApplyEffects calls to not create closures.
Fix OrderedOr to always deduplicate: when the new element was on the
 left deduplication was skipped.
Improve CaptureStart/End ToString to not look like parentheses and
 include the group number.