Ryan Lucia [Tue, 15 Sep 2020 01:55:45 +0000 (21:55 -0400)]
[mono] Introduce MonoMemoryManager to hold memory previously owned by the domain (#41754)
* Introduce MonoMemoryManager
* Move things out of the domain
* Update callers
This does *not* handle cases where we call mono_domain_alloc and friends, just direct access to domain->mp etc. As of this, everything is allocated from the default ALC on netcore and the domain on legacy.
* Update locks in domain_alloc functions
* Take correct lock around mp
* Introduce memory manager allocation helpers
* Shorten mono_domain_default_memory_manager
* Add memory_manager alloc nolock variants
* Fix locking
* More use of nolock functions
* Add trailing newline to monovm.h
This is obviously unrelated, but it was causing issues with my sed commands
* mono_memory_manager -> mono_mem_manager
Bit shorter at least, but I can cut it down further if deemed necessary
* Add to msvc build
* Fix mono/mono build
* Fix test failure
Oops
* Remove erroneous returns
* Store domain in memory manager, take domain lock
* Build fixes
* Remove comment
* Free memory manager during domain unload
* C++ fix
* Don't double free
* Remove designated initializer
MSVC C++ fails it seems :(
Sergey Andreenko [Tue, 15 Sep 2020 01:21:26 +0000 (18:21 -0700)]
SPMI: fix performance of `repRecordCallSite`. (#42192)
* Fix a long replaying test.
* add a comment.
Tarek Mahmoud Sayed [Mon, 14 Sep 2020 22:14:20 +0000 (15:14 -0700)]
Fix the globalization test. (#42212)
Stephen Toub [Mon, 14 Sep 2020 22:08:03 +0000 (18:08 -0400)]
Redo annotation of Enumerable.Cast and friends to use oblivious TResult (#42215)
* Revert "Make Enumerable.Cast<T> and friends return IEnumerable<T?> (#40929)"
This reverts commit
e75670889321611505115e4561a98e80732b58a0.
* Redo annotation of Enumerable.Cast and friends to use oblivious T
Some folks aren't happy with the fallout of Cast being annotated to return a `TResult?`, and others aren't happy with having it be `TResult`. Consensus is to make the TResult oblivious, such that we don't declare one way or another what it is, since there's no good way to annotate it appropriately.
monojenkins [Mon, 14 Sep 2020 20:48:16 +0000 (16:48 -0400)]
[utils] Fix MONO_SOLIB_EXT in mono-dl.h on Apple ARM64 (now .dylib) (#42181)
Co-authored-by: k15tfu <k15tfu@users.noreply.github.com>
Sergey Andreenko [Mon, 14 Sep 2020 19:59:00 +0000 (12:59 -0700)]
Don't use BADCODE when inlining. (#42193)
Ankit Jain [Mon, 14 Sep 2020 19:12:21 +0000 (15:12 -0400)]
[wasm][debugger] Avoid infinite loop when we have a boxed `new object` (#42059)
* [wasm][debugger] Avoid infinite loop when we have a boxed `new object`
Eg. `object o = new object(); object o1 = o;`
- Avoid infinitely looping for `o1`
* [wasm][debugger] Handle valuetypes boxed in classes that are not
.. type `object`.
Prompted by @lambdageek's comment - https://github.com/dotnet/runtime/pull/42059#issuecomment-
690310274
Ankit Jain [Mon, 14 Sep 2020 19:09:49 +0000 (15:09 -0400)]
[wasm][debugger] Skip downloading pdb, for empty urls (#42060)
Mitchell Hwang [Mon, 14 Sep 2020 19:02:35 +0000 (15:02 -0400)]
Mark System.Runtime APIs Supported on Windows (#42030)
* [wasm] System.Runtime enable platform attributes
* System.Runtime Mark APIs Supported on Windows
* Avoid nullable token error
Co-authored-by: Mitchell Hwang <mitchell.hwang@microsoft.com>
Eric Erhardt [Mon, 14 Sep 2020 18:35:21 +0000 (13:35 -0500)]
AccessViolation when using ValueTypes as a Service (#42152)
* AccessViolation when using ValueTypes as a Service
When using ValueTypes as services in DependencyInjection, we are generating incorrect IL in a few cases:
- When the ValueType is a top level service
- When the ValueType is in an IEnumerable of services
We were double unboxing the ValueType, which was causing AccessViolations.
This was a regression caused by a previous change when ValueTypes were used in constructor parameters of services.
To fix this, I lifted the unboxing into the VisitConstructor and VisitIEnumerable methods instead of forcing VisitConstant to always do the unboxing. VisitConstant doesn't have enough context to know whether it should do the unboxing or not.
Fix #42037
* Add transient tests
Kevin Gosse [Mon, 14 Sep 2020 17:39:13 +0000 (19:39 +0200)]
Test only the BodyContent property (#42177)
Parse the response as JSON in order to test only the BodyContent property.
Tanner Gooding [Mon, 14 Sep 2020 16:28:31 +0000 (09:28 -0700)]
Removing the T4 templates used by Vector<T> (#42050)
* Removing the T4 templates used by Vector<T>
* Fixing System.Numerics.Vectors.Tests
* Adding back Register so morph doesn't break
* Renaming Vector.cs to Vector_1.cs
* Renaming Vector_Operations.cs to Vector.cs
Leslie Zhai [Mon, 14 Sep 2020 16:05:22 +0000 (00:05 +0800)]
Implement MIPS64 RID (#42194)
https://github.com/gsvm/loongson-dotnet/issues/6
Ryan Lucia [Mon, 14 Sep 2020 14:47:47 +0000 (10:47 -0400)]
[mono] Enable startup stats timer on wasm (#41994)
* Set clock in mono_clock_init if available and add wasm definitions
This matches the OSX behavior and cleans things up a bit. We also now call this cross-platform, so don't error when called on Windows
* Add new "real time" counter based on clock_gettime
The initial value is set early in mono_main as part of mono_counters_init. This is needed because /proc/stat does not exist with emscripten and we need a way to measure time on wasm
Jan Vorlicek [Mon, 14 Sep 2020 13:56:50 +0000 (15:56 +0200)]
Temporarily disable Linux musl ARM testing in Helix (#42195)
We need to disable it until the CLI for musl arm is produced.
Tomáš Rylek [Mon, 14 Sep 2020 12:17:10 +0000 (14:17 +0200)]
Remove the directory src/coreclr/tests/src (#42139)
Move CLRTest.*.targets under src/tests/Common;
Consolidate Directory.Build.props in src/tests, src/tests/Common and src/coreclr/tests/src - src/tests is now the "master" version, src/tests/Common just points to it and src/coreclr/tests/src no longer exists;
Move Directory.Build.targets from src/coreclr/tests/src to src/tests/Directory.Build.targets that was previously just importing the src/coreclr/tests/src version;
Move the few remaining scripts under src/coreclr/tests/src to src/tests/Common (dirs.proj, nobuild.targets, runonly.targets, xunitconsolerunner.depproj).
Thanks
Tomas
dotnet-maestro[bot] [Mon, 14 Sep 2020 10:20:31 +0000 (10:20 +0000)]
[master] Update dependencies from Microsoft/vstest dotnet/xharness mono/linker dotnet/arcade dotnet/icu (#41796)
[master] Update dependencies from Microsoft/vstest dotnet/xharness mono/linker dotnet/arcade dotnet/icu
- Merge branch 'master' into darc-master-
7eb28eb8-d9b3-4ab3-8732-
196d2424c589
- Merge branch 'master' into darc-master-
7eb28eb8-d9b3-4ab3-8732-
196d2424c589
Maxim Lipnin [Mon, 14 Sep 2020 08:57:29 +0000 (11:57 +0300)]
Mark some System.Net.HttpListener APIs as unsupported on browser (#42063)
* Mark some System.Net.HttpListener APIs as unsupported on browser
* FIx windows builds
* Fix windows builds
* Remove browser annotations from the properties with windows-specific setters.
* Mark only setter of ExtendedProtectionPolicy
* Remove another redundant annotation
Aaron Robinson [Mon, 14 Sep 2020 07:59:42 +0000 (00:59 -0700)]
Disable UnmanagedCallersOnlyTest on WASM. (#42191)
Layomi Akinrinade [Mon, 14 Sep 2020 04:31:54 +0000 (21:31 -0700)]
Re-add linker attributes to generic type params of JsonSerializer methods (#42186)
David Cantú [Sun, 13 Sep 2020 16:25:55 +0000 (09:25 -0700)]
Add specification for extending supported types of TKey in ditionaries. (#32676)
* Add specification for extending supported types of TKey in ditionaries.
* Fix enumeration in notes
* fix notes
* Fix supported types.
Aaron Robinson [Sun, 13 Sep 2020 02:39:29 +0000 (19:39 -0700)]
Convert invalid C# uses of UnmanagedCallersOnly to IL. (#42146)
* Convert invalid C# uses of UnmanagedCallersOnly to IL for negative testing.
Adeel Mujahid [Sun, 13 Sep 2020 00:11:17 +0000 (03:11 +0300)]
Delete unused multibyte methods from PAL (#42157)
Thays Grazia [Sat, 12 Sep 2020 20:54:32 +0000 (17:54 -0300)]
[wasm][debugger][test] Change DateTime format test (#42136)
* Changing DateTime format tests, we cannot test how we were doing before because wasm and dotnet core uses different versions of ICU.
* Changing what @radical suggested.
monojenkins [Sat, 12 Sep 2020 16:57:23 +0000 (12:57 -0400)]
[gc] add stop-the-world api (#42138)
It's sometimes useful to have a way for the runtime to stop the world. Only works on SGen.
Co-authored-by: lambdageek <lambdageek@users.noreply.github.com>
monojenkins [Sat, 12 Sep 2020 14:28:54 +0000 (10:28 -0400)]
[metadata] Add #JTD heap MonoImage:minimal_delta bit (#42137)
The presence of the heap indicates that it's a minimal delta image (where the heaps are meant to be appended to the previous heaps), as opposed to a full delta image (where the heap in the delta image contains the previous and the
new data)
---
Also add `MONO_TRACE_METADATA_UPDATE` (set `MONO_LOG_MASK` to `metadata-update`) for tracing
Co-authored-by: lambdageek <lambdageek@users.noreply.github.com>
Tomáš Rylek [Fri, 11 Sep 2020 20:45:44 +0000 (22:45 +0200)]
Move the remaining test dependencies under src/tests (#42103)
Move ilasm, test_dependencies and test_dependencies_fs under src/tests.
Jan Kotas [Fri, 11 Sep 2020 20:29:20 +0000 (13:29 -0700)]
Fix ldtoken for generic RuntimeMethodHandle (#42112)
ldtoken for generic RuntimeMethodHandle in non-generic method triggered generic dictionary lookup due to normalization of MethodDescs for reflection. It lead to InvalidProgramException later.
Fixes #41674
David Wrighton [Fri, 11 Sep 2020 19:58:59 +0000 (12:58 -0700)]
Fix Unix build issue (#42104)
When native build fails, handle exit code properly by checking it immediately after cmake is called
Michal Strehovský [Fri, 11 Sep 2020 19:56:28 +0000 (21:56 +0200)]
Fix building JIT without runtime on Linux (#42090)
* Don't build createdump
* We need mscorrc
Roman Marusyk [Fri, 11 Sep 2020 19:50:57 +0000 (22:50 +0300)]
PhysicalFileProvider and PhysicalFilesWatcher should implement IDisposable correctly (#41918)
Tanner Gooding [Fri, 11 Sep 2020 19:27:39 +0000 (12:27 -0700)]
Removing the Vector*_Intrinsics.cs file, regions from the numerics code, and normalizing the member order (#41898)
* Removing the Vector*_Intrinsics.cs file, regions from the numerics code, and normalizing the member order
* Updating the doc comment for Quaternion
Juan Hoyos [Fri, 11 Sep 2020 15:39:58 +0000 (08:39 -0700)]
Enable build + pack of cross-OS diagnostic artifacts on Linux-musl-arm (#42096)
Adeel Mujahid [Fri, 11 Sep 2020 15:02:49 +0000 (18:02 +0300)]
Delete unused kernel32 QCalls (#42058)
Aleksey Kliger (λgeek) [Fri, 11 Sep 2020 13:36:15 +0000 (09:36 -0400)]
[mono] Preserve GetterAdapterFrame if CreateGetterDelegate is used (#42091)
Fixes https://github.com/dotnet/runtime/issues/41811
Tomáš Rylek [Fri, 11 Sep 2020 13:20:13 +0000 (15:20 +0200)]
Delete the obsolete PerfHarness project (#42092)
According to Drew's clarification this is an old version of the
perf harness that is no longer used. Deleting it as part of my
ongoing work on CoreCLR test system cleanup.
Thanks
Tomas
Anirudh Agnihotry [Fri, 11 Sep 2020 13:06:50 +0000 (06:06 -0700)]
move template (#42098)
Jan Vorlicek [Fri, 11 Sep 2020 12:39:00 +0000 (14:39 +0200)]
Fix vtable chunk sharing with covariant returns (#42077)
* Fix vtable chunk sharing with covariant returns
There is an issue causing invalid sharing of a vtable chunk when a
method is overriden using method impl in a class and its parent class
and the slot in the parent class is in a previous vtable chunk. The
algorithm in MethodTableBuilder::ChangesImplementationOfVirtualSlot that
detects whether a slot in a vtable chunk was changed doesn't know that
the override will be propagated to all slots occupied by the method
being overriden. In the test attached to this fix, it leads to a stack
overflow due to an infinite recursion in the type system code. It could
also result in an overwrite of the method in the parent class, so when
another class derives from it, it would get a method from its sibling.
This change fixes it by checking if the parent method of a current slot
was overriden in some previous slot in the vtable.
Maxim Lipnin [Fri, 11 Sep 2020 09:42:32 +0000 (12:42 +0300)]
Mark System.Security.Cryptography.Algorithms as unsupported on Browser WASM (#42010)
* Mark System.Security.Cryptography.Algorithms as unsupported on Browser WASM
* Address the comment regarding IncrementalHash.CreateHMAC
* Group System.* usings
* Restore the blank lines between the meta-comment and the code for all the HMAC classes
* Fix style issues
* Mark MD5 class as unsupported on Browser; clean up the api exclusion list and proj file.
* Using
* Style fix
Zoltan Varga [Fri, 11 Sep 2020 07:59:57 +0000 (03:59 -0400)]
Remove some accidental duplicates from mini/CMakeLists.txt. (#42095)
Maxim Lipnin [Fri, 11 Sep 2020 06:10:44 +0000 (09:10 +0300)]
Mark some System.Net.Http APIs as unsupported on browser (#42074)
Viktor Hofer [Fri, 11 Sep 2020 00:09:31 +0000 (02:09 +0200)]
Remove machine state modifications on Unix (#42083)
Changing /proc/sys/kernel/core_pattern is handled by core-eng and should not be done by helix consuming repositories. Also removing a workaround that has been fixed since (__CrashDumpFolder=/cores) on macOS. Also removing ulimit which is handled by core-eng as well.
Tanner Gooding [Fri, 11 Sep 2020 00:05:51 +0000 (17:05 -0700)]
Fixing TargetHasAVXSupportCheck to read ECX not EDX (#42089)
Tomáš Rylek [Thu, 10 Sep 2020 22:32:27 +0000 (00:32 +0200)]
Remove Windows / Unix duplications in send-to-helix script variants (#42038)
After cleaning up send-to-helix-step / send-to-helix-inner-step,
I have noticed two more scripts unnecessarily duplicating the
Windows / Unix variant the same way, both of which can be adapted
to my (slightly generalized) send-to-helix-inner-step.
Thanks
Tomas
Kevin Jones [Thu, 10 Sep 2020 22:31:28 +0000 (18:31 -0400)]
Correct fallback logic for ECDiffieHellmanCng with other ECDHPublicKey types
Use the concrete ECDiffieHellmanCng type (which is the internal type in Algorithms and the public type in Cng) instead of calling Create (which always returns the internal type).
Zoltan Varga [Thu, 10 Sep 2020 19:34:22 +0000 (15:34 -0400)]
Ongoing work on the cmake build. (#42072)
* Ongoing work on the cmake build.
* Add wasm cross compiler support.
* Add ios target support.
* Make ranlib deterministic on osx.
* Add GC_SUSPEND option.
* Add a 'preemptive' value to GC_SUSPEND.
* Add an ENABLE_WERROR option.
* Split mono/mini/CMakeLists.txt file into per-subdirectory files.
Tomas Weinfurt [Thu, 10 Sep 2020 18:18:59 +0000 (11:18 -0700)]
improve ConnectAsync_CancellationRequestedAfterConnect_ThrowsOperation test (#42054)
Tanner Gooding [Thu, 10 Sep 2020 17:51:09 +0000 (10:51 -0700)]
Enabling CA1416: Validate platform compatibility (#41760)
* Enabling CA1416
* Resolving PR feedback
* Updating the .NET analyzers to 5.0.0-rc2.20458.2
* Resolving compilation errors on full framework
* Fixing two more tfm issues with the platform compat checks
* Adjusting various platform compat checks to work with netstandard and netfx
Michal Strehovský [Thu, 10 Sep 2020 17:31:32 +0000 (19:31 +0200)]
Delete AreEqualOrdinalIgnoreCase (#42067)
This is unused after #40910 and breaks building standalone System.Globalization.Native (`error G94D986AD: unused function 'AreEqualOrdinalIgnoreCase' [-Werror,-Wunused-function]`).
Geoff Kizer [Thu, 10 Sep 2020 16:13:39 +0000 (09:13 -0700)]
Implement SocketsHttpHandler.ConnectCallback (#41955)
* Implement SocketsHttpHandler.ConnectCallback
Tomas Weinfurt [Thu, 10 Sep 2020 15:49:42 +0000 (08:49 -0700)]
move Http2_MultipleConnectionsEnabled_IdleConnectionTimeoutExpired_ConnectionRemovedAndNewCreated to outerloop (#42051)
monojenkins [Thu, 10 Sep 2020 15:47:54 +0000 (11:47 -0400)]
[crashing] update s390x handling (#41961)
commit
2a4e66af4dc628a84a42ff443bc2c09a4bcc7dec broke build on s390x after
changing the signature of `mono_handle_native_crash()`. We need to update
exceptions-s390x.c as well.
Fixes: https://github.com/mono/mono/issues/20356
Co-authored-by: sharkcz <sharkcz@users.noreply.github.com>
Jan Vorlicek [Thu, 10 Sep 2020 13:36:47 +0000 (15:36 +0200)]
Enable Alpine ARM support (#41982)
* Enable Alpine ARM support
This change enables Alpine ARM support in the lab build scripts,
packages and RID graph. It enables CI and official builds.
No test leg execution until packages are produced.
Přemek Vysoký [Thu, 10 Sep 2020 12:45:09 +0000 (14:45 +0200)]
Add test leg to the PR build to run libraries tests on iOS x64 Simulators (#37476)
Forked from https://github.com/dotnet/runtime/pull/37057
Adds the iOS test leg that will execute selected BCL tests on an iOS x64 Simulator (13.5) using the `OSX.1015.Amd64.Open` queue which is our on-prem Mac Mini infrastructure (60 machines).
To prepare the Helix job, we are using the Arcade Helix SDK, more specifically the [XHarness Helix SDK](https://github.com/dotnet/arcade/blob/master/src/Microsoft.DotNet.Helix/Sdk/tools/xharness-runner/Readme.md) which uses the [XHarness tool](https://github.com/dotnet/xharness) to control the test run.
Levi Broderick [Thu, 10 Sep 2020 11:06:35 +0000 (04:06 -0700)]
Temporarily disable arm64-intrin. UTF-16 validation (#42052)
Ryan Lucia [Thu, 10 Sep 2020 10:50:53 +0000 (06:50 -0400)]
[mono] Propagate -Werror properly, fix warning (#41996)
Viktor Hofer [Thu, 10 Sep 2020 08:50:09 +0000 (10:50 +0200)]
Update SDK min and target version to >= Preview8 (#41947)
* Update SDK min and target version to >= Preview8
Kunal Pathak [Thu, 10 Sep 2020 05:32:50 +0000 (22:32 -0700)]
Adds instruction count and summarized variable debug info (#42040)
* Print instruction count in Jitasm
* Print summarized variable debug info
* Make instrCount a local variable
Kenneth Pouncey [Thu, 10 Sep 2020 04:25:46 +0000 (06:25 +0200)]
[browser][http] Add support for AllowAutoRedirect. (#41394)
* [browser][http] Add support for AllowAutoRedirect.
* Update src/libraries/System.Net.Http/src/System/Net/Http/BrowserHttpHandler/BrowserHttpHandler.cs
Co-authored-by: campersau <buchholz.bastian@googlemail.com>
* Address review comments and add flag to determine if the _allowAutoRedirect was explicitly set or not.
* Remove Disposed and Handler checks to address review comments
* Update src/libraries/System.Net.Http/src/System/Net/Http/BrowserHttpHandler/BrowserHttpHandler.cs
Co-authored-by: Marie Píchová <11718369+ManickaP@users.noreply.github.com>
* MaxAutomaticRedirections moved back to PlatformNotSupportedException.
* Address review comments about wrong order.
Co-authored-by: campersau <buchholz.bastian@googlemail.com>
Co-authored-by: Marie Píchová <11718369+ManickaP@users.noreply.github.com>
Anirudh Agnihotry [Thu, 10 Sep 2020 03:00:45 +0000 (20:00 -0700)]
add build rotation doc (#42005)
* add build rotation doc
* Update docs/buildtriage.md
* address feedback
* removing the extra table
* Update docs/buildtriage.md
Co-authored-by: Jan Kotas <jkotas@microsoft.com>
* remove pass rate
Co-authored-by: Dan Moseley <danmose@microsoft.com>
Co-authored-by: Jan Kotas <jkotas@microsoft.com>
Tomas Weinfurt [Thu, 10 Sep 2020 01:09:47 +0000 (18:09 -0700)]
rework UseCallback_BadCertificate_ExpectedPolicyErrors test to use Loopback server to avoid external dependency (#41850)
* rework UseCallback_BadCertificate_ExpectedPolicyErrors test to use Loppbackserver to avoid external dependency
* feedback from review
* attempt to fix old winhttp
Anirudh Agnihotry [Thu, 10 Sep 2020 00:52:13 +0000 (17:52 -0700)]
set isShipping=false for internal.transport (#42032)
Khalid Abuhakmeh [Thu, 10 Sep 2020 00:27:45 +0000 (20:27 -0400)]
Avoid Unnecessary Cast (#42033)
By changing to we no longer have a
cast from double to float.
Fixes #41140
Egor Bogatov [Thu, 10 Sep 2020 00:16:16 +0000 (03:16 +0300)]
[iOS] Make Environment.GetFolderPathCore Xamarin.iOS compatible (#41959)
Jeremy Koritzinsky [Wed, 9 Sep 2020 22:44:03 +0000 (15:44 -0700)]
Remove empty transform file. (#42042)
Signed-off-by: Jeremy Koritzinsky <jekoritz@microsoft.com>
David Wrighton [Wed, 9 Sep 2020 21:49:42 +0000 (14:49 -0700)]
Build infra changes to enable building a more universal cross target jit builds (#41126)
This change is part of a larger effort to enable crossgen2 to perform general purpose cross compilation from any architecture to any other architecture. This portion of the change is to begin generating cross targeting jits with a consistent naming scheme with a minimal amount of code change.
The practical output of this change is replacement of the current practice of building protojit/protononjit/etc. with `clrjit_<targetos>_<target_arch>_<host_arch>` binaries.
In addition I've defined a couple of new subsets so that this can be less impactful to the inner dev loop.
clr.alljits, which will build all of the jits.
clr.jit, which will build only non-cross targeting jits.
clr.runtime which has been enhanced to not trigger the alljits build.
Current restrictions of the change:
1. This change does not support 32bit to 64bit compilation. I've built a larger change that includes the necessary work for that, but the changes are more disruptive, and I would prefer to scope this work to make it reviewable.
2. This work builds different compilers for the unix targetting and windows targetting compilers. This is due to the different ABIs that the architectures support. Upon investigation, I've determined that the actual product differences in these cases is small for ARM and ARM64. In the future, once we have testing, I believe it would be reasonable to unify the generated compiler binaries.
3. These changes do not build an armel targeting compiler for all scenarios, but it does build one with the current naming scheme as was done before. Again, like in condition 2, I believe we can make a single arm targetting compiler that will work.
4. These changes include changes to crossgen2 to use the new compiler binaries. However, there is no expanded testing to cover test the newly available cross compilation capabilities. That will appear in follow-on work.
Adeel Mujahid [Wed, 9 Sep 2020 21:41:13 +0000 (00:41 +0300)]
Add managed Marshal.AllocBSTRByteLen() (#41962)
monojenkins [Wed, 9 Sep 2020 21:39:11 +0000 (17:39 -0400)]
[interp] Remove dead basic blocks and merge adjacent ones (#41913)
We have a new pass that iterates over basic blocks list. This pass merges adjacent bblocks (who have a single out/in with is matches with one another and they are also consecutive in code order, `first_bb->next_bb == second_bb`) and removes dead basic blocks (who have `in_count == 0`). The linking between basic blocks is set during code generation and it is also updated, by removing links during constant folding and removal of conditional branches. If we merge a pair of basic blocks then we do another copy propagation pass, since cprop runs only at basic block level, and it can now propagate values to the new merged bblock.
This PR also enables inlining of methods with multiple returns, by having an exit_bb where all `ret` from inlined method can branch to.
Co-authored-by: BrzVlad <BrzVlad@users.noreply.github.com>
Jan Kotas [Wed, 9 Sep 2020 19:07:15 +0000 (12:07 -0700)]
Fix ilasm parsing of defines (#42002)
Regression introduced by #41611
Fixes #42001
Kevin Jones [Wed, 9 Sep 2020 18:32:25 +0000 (14:32 -0400)]
Handle IdLinkage X509 chain building status.
In macOS, when building an X509 chain, if a leaf and issuing certificate
form a chain, but the subject and authority key identifiers do not match,
an IdLinkage status is reported.
To match Windows, we ignore this chain status.
Linux appears to use these extensions when building a chain, and reports
PartialChain even if the attributes were not strictly needed to aid in
building a chain.
Additionally, this updates the debugging output for macOS to use an allocating
function to convert a CFString to a C string. CFStringGetCStringPtr will return
"NULL if the internal storage of theString does not allow this to be returned efficiently."
which it appears to have started doing for me.
Aaron Robinson [Wed, 9 Sep 2020 18:22:40 +0000 (11:22 -0700)]
Update comments to match official Win32 API documentation. (#42026)
Krzysztof Wicher [Wed, 9 Sep 2020 16:15:30 +0000 (18:15 +0200)]
Ensure <Nullable>enable<Nullable> on System.IO.Packaging and S.Resources.Extensions (#41731)
Fix src ref diffs
Kevin Jones [Wed, 9 Sep 2020 16:03:49 +0000 (12:03 -0400)]
Implement ExportSubjectPublicKeyInfo on ECDHPublicKey
Implement {Try}ExportSubjectKeyInfo on ECDHPK.
Kevin Jones [Wed, 9 Sep 2020 16:00:49 +0000 (12:00 -0400)]
Handle name constraint errors from chain building.
Map Linux name constraint status codes to the appropriate
X509ChainStatusFlags.
macOS does not have a unique status for different types of name
constraint violations like Linux and Windows do. Map this to
InvalidNameConstraint.
Jan Kotas [Wed, 9 Sep 2020 14:05:33 +0000 (07:05 -0700)]
Integrate minor fixes from dotnet/runtimelab:NativeAOT (#41999)
Mitchell Hwang [Wed, 9 Sep 2020 13:36:02 +0000 (09:36 -0400)]
Mark System.IO.Pipes Unsupported at assembly level (#41976)
Co-authored-by: Mitchell Hwang <mitchell.hwang@microsoft.com>
Marie Píchová [Wed, 9 Sep 2020 12:27:08 +0000 (14:27 +0200)]
Fix exception mapping during expect 100 cancellation. (#41800)
Eirik Tsarpalis [Wed, 9 Sep 2020 09:10:42 +0000 (10:10 +0100)]
fix assert ordering (#42007)
Ryan Lucia [Wed, 9 Sep 2020 06:26:39 +0000 (02:26 -0400)]
[mono] Minor fixes to the wasm browser sample (#41992)
Use python 3, fix typo, prevent inlining
Wraith [Wed, 9 Sep 2020 00:52:10 +0000 (01:52 +0100)]
Update xml char type (#41756)
Mitchell Hwang [Tue, 8 Sep 2020 23:24:33 +0000 (19:24 -0400)]
[wasm] Mark System.Net.NameResolution Unsupported at assembly level (#41985)
Co-authored-by: Mitchell Hwang <mitchell.hwang@microsoft.com>
Maxim Lipnin [Tue, 8 Sep 2020 23:24:22 +0000 (02:24 +0300)]
Mark System.Net.WebSockets.ClientWebSocketOptions APIs as unsupported on Browser (#41963)
* Mark System.Net.WebSockets.ClientWebSocketOptions APIs as unsupported on Browser
* Add the attributes to non-browser version of ClientWebSocketOptions class to avoid build error
* Add using
* Include platform attributes
Mitchell Hwang [Tue, 8 Sep 2020 23:24:09 +0000 (19:24 -0400)]
[wasm] Mark System.Threading.ThreadPool unsupported on Browser (#41891)
* System.Threading.ThreadPool enable platform attributes
* Mark System.Threading.ThreadPool APIs unsupported on browser
* System.Threading.ThreadPool Add Unsupported attribute to other ThreadPool files
* Remove Unsupported attributes from BindHandle
* Add windows Supported Attribute to BindHandle
Co-authored-by: Mitchell Hwang <mitchell.hwang@microsoft.com>
Tomáš Rylek [Tue, 8 Sep 2020 22:55:25 +0000 (00:55 +0200)]
Add initial release testing for outerloop and jitstress pipelines (#41856)
I have fixed the CoreCLR outerloop pipeline to include the release
branches. For jitstress and libraries_jitstress I have split the
schedules to run every other day for master vs. release as with
the growing number of Mono runs there's not much space for maneuver
w.r.t. allocating new lab HW.
Thanks
Tomas
monojenkins [Tue, 8 Sep 2020 21:52:54 +0000 (17:52 -0400)]
Remove netcore/ directory from the mono/mono repository. (#41972)
Co-authored-by: vargaz <vargaz@users.noreply.github.com>
Kunal Pathak [Tue, 8 Sep 2020 20:36:56 +0000 (13:36 -0700)]
Do not address expose a struct having 1 slot (#40957)
Kunal Pathak [Tue, 8 Sep 2020 20:36:06 +0000 (13:36 -0700)]
ARM64: Eliminate redundant ldr/str instruction (#39222)
* IsRedundantLdStr
Added method to check if we can eliminate ld/str instruction
* jit format
* Fixed IsRedundantLdStr and call it from emitIns_R_S() and emitIns_S_R()
* Added check for lastEmitInstr->opSize
* Add some comments
Santiago Fernandez Madero [Tue, 8 Sep 2020 19:25:12 +0000 (12:25 -0700)]
[WASM] Add satellite assemblies bundle and hook to register them (#41358)
* [WASM] Add satellite assemblies bundle and hook to register them
* Fix test failures
* PR Feedback
* PR Feedback and add test to make sure load from path respects culture
* Add binding so that blazor can register bundled assemblies
* Remove blank line.
* PR Feedback
Vitek Karas [Tue, 8 Sep 2020 18:41:46 +0000 (11:41 -0700)]
Exclude .a files from the runtime native files when packaging (#41966)
There's no use for the .a files in the shared framework or runtime pack, they just increase disk size.
Also not that we didn't include them in 3.1, so it's effectively a regression in 5.0.
Alexander Nikolaev [Tue, 8 Sep 2020 18:22:24 +0000 (20:22 +0200)]
Disable Http2_MultipleConnectionsEnabled_ConnectionLimitNotReached_ConcurrentRequestsSuccessfullyHandled (#41974)
Wei Zheng [Tue, 8 Sep 2020 16:23:35 +0000 (00:23 +0800)]
Fix Configuration Save Full setting default value (#41912)
* Fix Configuration Save Full setting default value
* Cleaned out untrimmed whitespaces
Jiri Cincura ↹ [Tue, 8 Sep 2020 16:18:39 +0000 (18:18 +0200)]
Various Regexes for more testing. (#41935)
Layomi Akinrinade [Tue, 8 Sep 2020 15:39:46 +0000 (08:39 -0700)]
Honor custom number handling only when property/type is a number/collection of numbers (#41679)
* Honor custom number handling only when property/type is a number/collection of numbers
* Verify typeof(object) + custom number handling behavior
* Simplify setting of per-property number handling
* Clean up number-handling-applicable check
* Fix issue with number-handling + boxed non-number types
Maxim Lipnin [Tue, 8 Sep 2020 15:21:59 +0000 (18:21 +0300)]
Disable System.Xml.Tests.AsyncReaderLateInitTests.InitializationWithUriOnNonAsyncReaderTrows test on Browser WASM due to PNSE (#41936)
Marie Píchová [Tue, 8 Sep 2020 13:36:43 +0000 (15:36 +0200)]
Fixed stress test pipeline and other issues. (#41513)
Jan Kotas [Tue, 8 Sep 2020 04:51:03 +0000 (21:51 -0700)]
Move Marshal memory allocation methods into CoreLib shared partition (#41911)
* Move Marshal memory allocation methods into CoreLib shared partition
* Unify behavior across platforms
Eric StJohn [Tue, 8 Sep 2020 04:33:51 +0000 (21:33 -0700)]
Force concurrent LdapConnection in new process (#41880)
* Ensure an initial non-concurrent call to OpenLDAP
OpenLDAP requires a single call for initialization before any other concurrent call.
This fixes asserts and segfaults we were seeing when calling OpenLDAP concurrently.
Tomáš Rylek [Mon, 7 Sep 2020 22:18:14 +0000 (00:18 +0200)]
September infra rollout - continued CoreCLR / Mono test cleanup (minimal) (#40345)
Similar to the previous occasion, I have identified this set as the minimum meaningful set of scripts to move without causing too much disruption apart from purely mechanical renames. This change moves the developer-facing test-related CoreCLR scripts under the src/tests folder in accordance with the previous move of the tests themselves:
src/coreclr/build-test.cmd to src/tests/build.cmd
src/coreclr/build-test.sh to src/tests/build.sh
src/coreclr/tests/build.proj to src/tests/build.proj
src/coreclr/tests/runtest.cmd to src/tests/run.cmd
src/coreclr/tests/runtest.sh to src/tests/run.sh
src/coreclr/tests/src/runtest.proj to src/tests/run.proj
src/coreclr/tests/issues.targets to src/tests/issues.targets
The rest of the change constitutes mere path adjustments to match the primary rename.
Fixes: https://github.com/dotnet/runtime/issues/38919
Thanks
Tomas
Jan Vorlicek [Mon, 7 Sep 2020 22:16:19 +0000 (00:16 +0200)]
Cleanup stack limits usage (#41946)
There were couple of members / functions for stack limits checking that
were not used anymore.
Viktor Hofer [Mon, 7 Sep 2020 20:18:28 +0000 (22:18 +0200)]
Add transitive CompilerServices.Unsafe dependency (#41616)
* Add transitive CompilerServices.Unsafe dependency
for System.Memory and System.Threading.Tasks.Extensions package
references. Adding the dependency as a P2P to prefer the live asset
over the prebuilt.