Leandro A. F. Pereira [Sat, 11 Jan 2020 03:34:06 +0000 (19:34 -0800)]
gc, pal: Avoid linking libnuma.so on non-NUMA systems (#1468)
* gc, pal: Avoid linking libnuma.so on non-NUMA systems
In an effort to reduce the system calls performed during the
initialization, this seemed like another low-hanging fruit: only link
in libnuma.so if the system has more than one NUMA node.
* gc, pal: Remove unneeded dlsym() call
The result of dlsym() is never attributed to any variable, and this
function is never called, so it's safe to remove these dlsym() calls.
Levi Broderick [Sat, 11 Jan 2020 02:45:23 +0000 (18:45 -0800)]
Fixup orphaned references to GetRawSzArrayData (#1458)
Davis Goodin [Sat, 11 Jan 2020 01:56:43 +0000 (19:56 -0600)]
Add LibrariesAllConfigurationsOverridePath (#1620)
Dan Moseley [Sat, 11 Jan 2020 00:24:38 +0000 (16:24 -0800)]
Update workflow readme (#1512)
* Update workflow readme
* CR feedback and more detail
* Add daily workflow for libraries dev
* rebuild->build
* More updates
* typo:
* more
* More edits
* testing
* space:
Jan Vorlicek [Fri, 10 Jan 2020 23:58:43 +0000 (00:58 +0100)]
Fix all GC stress failures in Pri1 tests on Unix with crossgen2 (#1619)
* Fix all GC stress failures in Pri1 tests on Unix with crossgen2
Without this PR, 90% of the CoreCLR Pri1 tests are failing due to GC
holes. All of the errors are caused by various bugs in the GCRefMap
generation that this PR fixes. With this PR, all the tests that pass
without GC stress also pass with it.
1. GetCorElementType() was incorrectly casting from TypeFlags to
CoreElementType. It was checking for a range of TypeFlags, but that
range didn't map to CorElementType 1:1. That resulted e.g. in getting
ELEMENT_TYPE_R4 for TypeFlags.IntPtr.
2. Besides that, the GetCorElementType() was also incorrectly mapping
Enum to ELEMENT_TYPE_CLASS.
3. The ReportPointersFromStructInRegisters function was incorrectly
assuming that the false stack entries correspond to stack slots while
they in fact correspond to bytes.
4. Arguments that are passed on stack were incorrectly reported as being
passed in registers (if there were still any registers available)
5. Members of structs embedded in other structs were reporting wrong
offsets (we were missing updating the offset by the delta parameter in
the GcScanValueType when calling GCScanRoots)
6. The code that determines if return values are passed in registers or
on stack was completely missing handling of structs returned in
registers.
* Fix typo in comment
Vivek Mishra [Fri, 10 Jan 2020 23:50:59 +0000 (15:50 -0800)]
Merge pull request #497 from dotnet/31GA-snap
Snap updates for 3.1 GA
Vivek Mishra [Fri, 10 Jan 2020 23:45:04 +0000 (15:45 -0800)]
Fixed stage-packages as per comments
Tanner Gooding [Fri, 10 Jan 2020 22:29:30 +0000 (14:29 -0800)]
Updating genCodeForBinary to be VEX aware (#1344)
* Updating genCodeForBinary to be VEX aware
* Resolving some PR feedback
Davis Goodin [Fri, 10 Jan 2020 19:11:28 +0000 (13:11 -0600)]
Generate full platform manifest in official build (#1594)
* Generate full platform manifest in official build
Generates a full platform manifest in the official build by downloading all platforms' artifacts from the CoreCLR and Libraries jobs for each Installer job.
Andy Ayers [Fri, 10 Jan 2020 18:31:04 +0000 (10:31 -0800)]
JIT: block local constant propagation of handles when compiling with relocs (#1593)
This gives local constant prop the same behavior as value-number-based constant
prop; both now block handle propagation when the jit must generate relocs (R2R),
to ensure that handle references appear only in simple instruction forms like movs.
Prerequisite to #1309, which can enable such propagation.
Tomáš Rylek [Fri, 10 Jan 2020 17:53:52 +0000 (18:53 +0100)]
R2RDump improvements w.r.t. GC info reporting (#1539)
While I haven't yet managed to nail down the cause of hangs
in the r426654 regression test, I have at least used the opportunity
to improve R2RDump GC info reporting. I have made the reporting
much more compact and more descriptive at the same time by translating
register names and showing stack-relative addresses in a much more
straightforward manner. I'm also proposing to ass a VS solution
for R2RDump - as it now depends on the ILCompiler.Reflection.ReadyToRun
project, it just sucks not to be able to work with both in the VS.
Thanks
Tomas
Carol Eidt [Fri, 10 Jan 2020 17:09:00 +0000 (09:09 -0800)]
Fix and Test case for #27924 (#1059)
Fix and Test case for #27924
Jeremy Barton [Fri, 10 Jan 2020 16:13:53 +0000 (08:13 -0800)]
Always build the OpenSSL shim in portable mode on macOS.
Aaron Robinson [Fri, 10 Jan 2020 02:19:41 +0000 (18:19 -0800)]
Merge pull request #1226 from AaronRobinsonMSFT/sdk_4099
Fix parsing of metadata when looking for COM related CustomAttributes.
buyaa-n [Fri, 10 Jan 2020 01:33:58 +0000 (17:33 -0800)]
Annotate System.Linq.Expressions for nullable (#1064)
* Annotate System.Linq.Expressions for nullable
Egor Chesakov [Thu, 9 Jan 2020 21:11:05 +0000 (13:11 -0800)]
[Arm64] Implement AbsoluteDifference hardware intrinsics (#1334)
And use VecBinOpTest.template in AbsoluteCompare* tests to remove duplicated logic in validation
Elinor Fung [Thu, 9 Jan 2020 19:30:46 +0000 (11:30 -0800)]
Add binder tracing test for ResolveSatelliteAssembly stage (#1455)
Tamar Christina [Thu, 9 Jan 2020 18:32:17 +0000 (18:32 +0000)]
[runtime] Arm64: Add TransposeEven and TransposeOdd intrinsics (#1220)
Tomas Weinfurt [Thu, 9 Jan 2020 18:00:27 +0000 (10:00 -0800)]
remove unneeded code (#1521)
* check
* cleanup
Jan Vorlicek [Thu, 9 Jan 2020 17:58:34 +0000 (18:58 +0100)]
Add gcstress option to SuperILC (#1446)
* Add gcstress option to SuperILC
This option allows executing tests with gcstress mode set to a passed
in value.
I have also added additional fixes for race issues with _logWriter.
Justin Kotalik [Thu, 9 Jan 2020 17:55:04 +0000 (09:55 -0800)]
Add MsQuic implementation of System.Net.Quic (#427)
Josh Groves [Thu, 9 Jan 2020 17:49:31 +0000 (14:19 -0330)]
Fix typo (#1528)
Eirik Tsarpalis [Thu, 9 Jan 2020 17:30:40 +0000 (19:30 +0200)]
Nullable annotate System.Reflection.Metadata (#1056)
* Nullable annotate System.Reflection.Metadata
* remove unused declaration
* update System.Diagnostics.Stacktrace for System.Reflection.Metadata annotations
* address feedback
* reinstate annotations erroneously removed by generator
* Apply feedback
* address feedback
* address feedback
Eirik Tsarpalis [Thu, 9 Jan 2020 17:29:38 +0000 (19:29 +0200)]
fix stress script issue (#1452)
Alexander Köplinger [Thu, 9 Jan 2020 17:17:46 +0000 (18:17 +0100)]
Rename file extension of RunnerTemplate.Windows.txt to .cmd (#1470)
So it automatically gets the EOL treatment from https://github.com/dotnet/runtime/pull/1336.
Stephen Halter [Thu, 9 Jan 2020 14:18:17 +0000 (06:18 -0800)]
Don't call AdvanceTo(default) in PipeReader.CopyToAsync (#1437)
* Don't call AdvanceTo(default) in PipeReader.CopyToAsync
* Address PR feedback
Elinor Fung [Thu, 9 Jan 2020 14:15:39 +0000 (06:15 -0800)]
Remove BaseDomain::IsSharedDomain (always false now) (#1456)
Stephen Toub [Thu, 9 Jan 2020 13:56:42 +0000 (08:56 -0500)]
Expose CreateLinkedTokenSource(CancellationToken) (#1379)
Erhan Atesoglu [Thu, 9 Jan 2020 13:41:23 +0000 (05:41 -0800)]
CoreFX renames (#1440)
Stephen Toub [Thu, 9 Jan 2020 10:42:17 +0000 (05:42 -0500)]
Merge pull request #1348 from stephentoub/moreregexperf
More Regex perf improvements
Tomas Weinfurt [Thu, 9 Jan 2020 07:41:46 +0000 (23:41 -0800)]
improve ALPN handling (#1347)
Michal Strehovský [Thu, 9 Jan 2020 07:11:21 +0000 (08:11 +0100)]
Fix fat call transform for x86 (#1451)
`AddArgumentToTail` with a null first argument will AV.
I *think* this also works with methods that take a return buffer (methods that return large structs), but might require a pair of eyes more experienced in this codebase.
Stephen Toub [Thu, 9 Jan 2020 02:57:55 +0000 (21:57 -0500)]
Add a few more tests / improve code coverage
Stephen Toub [Wed, 8 Jan 2020 20:28:26 +0000 (15:28 -0500)]
Address PR feedback
Stephen Toub [Wed, 8 Jan 2020 19:36:27 +0000 (14:36 -0500)]
Additional code gen improvements
- In Go, Improve code generation from several constructs to reduce code size and increase bounds reduction in the non-backtracking compiler
- In FindFirstChars, use IndexOf instead of Boyer-Moore if the found prefix is only one character in length.
- Hide some debug-only functions from code coverage
Stephen Toub [Mon, 6 Jan 2020 21:43:05 +0000 (16:43 -0500)]
Rename "Greedy" subexpressions to "Atomic"
The greedy naming is really confusing (we should fix it in the framework documentation as well).
In general regular expression nomenclature, "greedy" is used as the opposite of "lazy", indicating how much a loop initially consumes, e.g. does `a*` first try consuming as many 'a's as possible or does it first try consuming as few 'a's as possible... it's "greedy" (denoted as `a*`) if it consumes as many as possible and "lazy" (denoted as `a*?` if it consumes as few.
How aggressively it consumes, however, is orthogonal to whether it backtracks. Whereas `a*` is greedy and backtracking and `a*?` is lazy and backtracking, `(?>a*)` is greedy and non-backtracking and `(?>a*?) is lazy and non-backtracking.
Unfortunately, the nomenclature in the implementation and the documentation describes the `(?> ... )` as being a "greedy subexpression", which then conflates the meaning of "greedy".
The rest of the industry refers to these instead as "atomic", so I've changed it to that in the implementation.
Stephen Toub [Mon, 6 Jan 2020 20:46:25 +0000 (15:46 -0500)]
Fix CharInClass handling of invariant in lookup table
The Boolean logic was previously wrong, leading to characters being included in the set even if they weren't meant to be, e.g. given the set [^bc], this should include everything other than b and c, but if invariant culture was being used, it incorrectly included B and C in the set.
Stephen Toub [Mon, 6 Jan 2020 19:31:42 +0000 (14:31 -0500)]
Add more tests, and fix some bugs found
Stephen Toub [Sun, 5 Jan 2020 20:20:45 +0000 (15:20 -0500)]
Restore netfx test configuration to serve as oracle
With all the changes being made in the regex implementation in this release, we want to make sure we don't regress, and a good way to help doing that is add a bunch of tests and ensure they pass on netfx as well, where appropriate.
Stephen Toub [Sun, 5 Jan 2020 19:54:32 +0000 (14:54 -0500)]
Add non-backtracing support for greedy loops and lazyloops
Stephen Toub [Sun, 5 Jan 2020 19:06:03 +0000 (14:06 -0500)]
Auto-greedy for last Alternate/Loop/Lazyloop in regex
Stephen Toub [Sun, 5 Jan 2020 19:00:53 +0000 (14:00 -0500)]
Avoid unnecessary operations when textSpanPos == 0
Stephen Toub [Sat, 4 Jan 2020 04:10:41 +0000 (23:10 -0500)]
Add non-backtracking go support for anchors, alternation, lookahead, repeaters
Stephen Toub [Sat, 4 Jan 2020 02:08:41 +0000 (21:08 -0500)]
Use vectorized IndeOf for some notoneloopgreedy loops
Stephen Toub [Fri, 3 Jan 2020 17:13:59 +0000 (12:13 -0500)]
Add Notoneloopgreedy to go with One/Setloopgreedy
I initially mistakenly thought this wouldn't be useful, but it is. For example, the expression ".*\n" can be made non-backtracing.
Stephen Toub [Fri, 3 Jan 2020 16:38:07 +0000 (11:38 -0500)]
Auto-greedy last set/one loop in regex
Stephen Toub [Thu, 2 Jan 2020 19:46:42 +0000 (14:46 -0500)]
Upgrade some loops followed by boundaries to be greedy
Stephen Toub [Thu, 2 Jan 2020 01:51:07 +0000 (20:51 -0500)]
Add optimized non-backtracking code gen for limited subset of regexes
When we can determine from the regex that backtracking isn't needed, we
can generate much smaller / better code in RegexCompiler.GenerateGo.
Stephen Toub [Mon, 30 Dec 2019 02:52:30 +0000 (21:52 -0500)]
Improve StringCode lookup in RegexWriter to avoid unnecessary list
Stephen Toub [Sun, 29 Dec 2019 20:26:30 +0000 (15:26 -0500)]
Optimize allocation for child lists of RegexNodes
Stephen Toub [Sun, 29 Dec 2019 19:29:51 +0000 (14:29 -0500)]
Optimize {Lazy}loops that directly wrap Set/One/Notone
When written as a* these are already transformed appropriately by the parser, but other reductions can expose them in a way that the parser won't see. For example, (?:a)* ends up being a much more expensive regex even though it's identical functionally. This reduces the latter to the former, for cases where a {Lazy}loop is found to directly wrap a Set/One/Notone.
Stephen Toub [Thu, 26 Dec 2019 20:25:18 +0000 (15:25 -0500)]
Automatically upgrade some loops to be greedy
In an expression like A*B, backtracing may be involved even if it's impossible that it'll do any good, namely if there's no overlap between the A and B sets. This backtracking can be avoided by using a greedy subexpression, e.g. (?>A*)B, but a) few developers think to do that, and b) even when they do the implementation has more overhead than is desirable.
This change does three things:
1. It introduces primitives for "oneloopgreedy" and "setloopgreedy" that correspond to the existing "oneloop" (a single character in a loop) and "setloop" (a char class in a loop) primitives.
2. It reduces a Greedy node that contains a one/setloop to instead just be a one/setloopgreedy node.
3. It finds cases of one/setloop concatenated with other things, and automatically updates those to be one/setloopgreedy when it can provie that the backtracking would have no benefit.
(2) removes a small amount of overhead, in particular in compiled, because it avoids needing to spit additional tracking state for backtracking.
(3) can likely be expanded in the future to catch more cases, but it handles the common cases now. This has a very measurable impact on non-matches that would otherwise trigger backtracking into the subexpression.
Stephen Toub [Sun, 22 Dec 2019 03:45:35 +0000 (22:45 -0500)]
Code spit EnsureStorage replacement to avoid spilling locals
Only spill if we have to grow the relevant arrays.
Stephen Toub [Fri, 20 Dec 2019 16:05:26 +0000 (11:05 -0500)]
Clean up minor style issues in RegexCompiler
Stephen Toub [Fri, 20 Dec 2019 15:25:55 +0000 (10:25 -0500)]
Minor RegexCompiler codegen tweak for text[pos++]
The JIT generates slightly better code for:
```C#
text[pos] = ....;
pos++;
```
instead of:
```C#
text[pos++] = ...;
```
so use the IL equivalent of the former.
Stephen Toub [Wed, 18 Dec 2019 21:41:06 +0000 (16:41 -0500)]
Remove stale generated local in RegexCompiler
Stephen Toub [Sat, 28 Dec 2019 02:49:23 +0000 (21:49 -0500)]
Fix inadvertent closure in EmitCallCharInClass
Stephen Toub [Mon, 16 Dec 2019 23:29:26 +0000 (18:29 -0500)]
Add EmitCallCharInClass optimization for singleton
Stephen Toub [Mon, 16 Dec 2019 23:19:49 +0000 (18:19 -0500)]
Use IndexOf{Any} for case-sensitive singletons in FindFirstChar
Takes advantage of vectorization in string.IndexOf.
Stephen Toub [Mon, 16 Dec 2019 20:03:36 +0000 (15:03 -0500)]
Clean up more code in RegexCompiler
- Only use ilg from "macro" helpers
- Move some blocks of code around
Stephen Toub [Mon, 16 Dec 2019 19:45:08 +0000 (14:45 -0500)]
Put back Regex singleton optimization
I had removed it because it seemed illogical that anyone would write such character classes. But it turns out the implementation/parser itself generates them for certain forms, e.g. "hello|hithere" will create a character class containing just 'h' in order to find the start of the alternation.
Stephen Toub [Tue, 31 Dec 2019 19:43:19 +0000 (14:43 -0500)]
Delay JIT'ing until first execution
I'd previously removed multiple delegate allocations that were being incurred every time we needed to run a compiled regex concurrently. In doing so, however, I inadvertently forced JIT'ing to occur when the Regex was created rather than on first use; that can in turn lead to more start-up overheads for regexes that are created on startup (e.g. to initialize a static field) but that aren't used at start-up, as well as reduce potential parallelism that could otherwise occur for distinct regexes being run for the first time concurrently. This addresses the issue by still caching delegates, but in addition to rather than instead of the DynamicMethod objects.
Santiago Fernandez Madero [Thu, 9 Jan 2020 01:32:03 +0000 (17:32 -0800)]
Remove dead code in build.ps1 after -help option was added (#1430)
buyaa-n [Thu, 9 Jan 2020 01:09:53 +0000 (17:09 -0800)]
Forgot remove a ? from ref assembly (#1467)
Bruce Forstall [Thu, 9 Jan 2020 01:03:01 +0000 (17:03 -0800)]
Update superpmi.py helper script (#1474)
Mostly, updated documentation. Added some additional logging
output for improved clarity. Tried to simplify the code so
it is easier to understand, in places.
dotnet-maestro[bot] [Thu, 9 Jan 2020 00:08:17 +0000 (16:08 -0800)]
[master] Update dependencies from dotnet/arcade Microsoft/vstest (#1371)
* Update dependencies from https://github.com/dotnet/arcade build
20200107.1
- Microsoft.DotNet.XUnitExtensions - 5.0.0-beta.20057.1
- Microsoft.DotNet.VersionTools.Tasks - 5.0.0-beta.20057.1
- Microsoft.DotNet.ApiCompat - 5.0.0-beta.20057.1
- Microsoft.DotNet.Arcade.Sdk - 5.0.0-beta.20057.1
- Microsoft.DotNet.Build.Tasks.Configuration - 5.0.0-beta.20057.1
- Microsoft.DotNet.Build.Tasks.Feed - 5.0.0-beta.20057.1
- Microsoft.DotNet.Build.Tasks.Packaging - 5.0.0-beta.20057.1
- Microsoft.DotNet.Build.Tasks.SharedFramework.Sdk - 5.0.0-beta.20057.1
- Microsoft.DotNet.CodeAnalysis - 5.0.0-beta.20057.1
- Microsoft.DotNet.GenAPI - 5.0.0-beta.20057.1
- Microsoft.DotNet.XUnitConsoleRunner - 2.5.1-beta.20057.1
- Microsoft.DotNet.GenFacades - 5.0.0-beta.20057.1
- Microsoft.DotNet.Helix.Sdk - 5.0.0-beta.20057.1
- Microsoft.DotNet.RemoteExecutor - 5.0.0-beta.20057.1
* Update dependencies from https://github.com/dotnet/arcade build
20200107.5
- Microsoft.DotNet.XUnitExtensions - 5.0.0-beta.20057.5
- Microsoft.DotNet.VersionTools.Tasks - 5.0.0-beta.20057.5
- Microsoft.DotNet.ApiCompat - 5.0.0-beta.20057.5
- Microsoft.DotNet.Arcade.Sdk - 5.0.0-beta.20057.5
- Microsoft.DotNet.Build.Tasks.Configuration - 5.0.0-beta.20057.5
- Microsoft.DotNet.Build.Tasks.Feed - 5.0.0-beta.20057.5
- Microsoft.DotNet.Build.Tasks.Packaging - 5.0.0-beta.20057.5
- Microsoft.DotNet.Build.Tasks.SharedFramework.Sdk - 5.0.0-beta.20057.5
- Microsoft.DotNet.CodeAnalysis - 5.0.0-beta.20057.5
- Microsoft.DotNet.GenAPI - 5.0.0-beta.20057.5
- Microsoft.DotNet.XUnitConsoleRunner - 2.5.1-beta.20057.5
- Microsoft.DotNet.GenFacades - 5.0.0-beta.20057.5
- Microsoft.DotNet.Helix.Sdk - 5.0.0-beta.20057.5
- Microsoft.DotNet.RemoteExecutor - 5.0.0-beta.20057.5
* Update dependencies from https://github.com/microsoft/vstest build
20200103-01
- Microsoft.NET.Test.Sdk - 16.5.0-preview-
20200103-01
* Fix space in resource key to react to arcade resource generation fix
Co-authored-by: Santiago Fernandez Madero <safern@microsoft.com>
Drew Scoggins [Wed, 8 Jan 2020 20:01:14 +0000 (12:01 -0800)]
Merge pull request #1073 from DrewScoggins/dev/anscoggi/activateLiveBuilds
Fix passing of liveBuilds name string in perf job.
Levi Broderick [Wed, 8 Jan 2020 18:57:59 +0000 (10:57 -0800)]
Expose MemoryMarshal.GetArrayDataReference (#1036)
Given an SzArray, returns a reference to where the element at index 0 would be located, even if the array is empty. No variance checks or bounds checks are performed.
Next Turn [Wed, 8 Jan 2020 17:58:22 +0000 (01:58 +0800)]
Improve RawArrayData comments (#1450)
Adeel Mujahid [Wed, 8 Jan 2020 16:20:19 +0000 (18:20 +0200)]
Update URLs in BOTR (#1441)
- update links to sources from dotnet/coreclr -> dotnet/runtime/src/coreclr
- fix dead urls (github and non-github) under docs/design/coreclr.
- use https://software.intel.com/sites/default/files/article/402129/mpx-linux64-abi.pdf as a new home for x64 System V ABI, as http://x86-64.org is now abandoned. Update source files as well.
- normalize “, ” and ’ to ascii " and ' and remove trailing whitespaces in modified files.
Stephen Toub [Wed, 8 Jan 2020 15:40:03 +0000 (10:40 -0500)]
Disable failing BasicTest_AccessInstanceProperties_NoExceptions_Bsd test (#1447)
Stephen Toub [Wed, 8 Jan 2020 14:09:54 +0000 (09:09 -0500)]
Fix ReflectionTypeLoadException nullable annotations (#545)
We were previously using `Type[]?`, but the array can actually contain null, so it needs to be `Type?[]?`. Further, upon reviewing a bunch of call sites, it's clear that a) code expects to be able to pass null for the arrays into the ctor but also b) expects to be able to read Types/LoaderExceptions without doing null checks. As such, I modified the annotations accordingly, and fix the implementation to match.
Stephen Toub [Wed, 8 Jan 2020 14:09:28 +0000 (09:09 -0500)]
Fix nullable annotations on CollectionExtensions.GetValueOrDefault (#1364)
Stephen Toub [Wed, 8 Jan 2020 14:09:09 +0000 (09:09 -0500)]
Suppress ExecutionContext flow in SocketHttpHandler's ConnectHelper (#1381)
Anton Firszov [Wed, 8 Jan 2020 13:46:57 +0000 (14:46 +0100)]
Ensure LocalEndPoint being updated in SendToAsync(SocketAsyncEventArgs) (#808)
Fixes #915 by assigning _rightEndPoint like in other overloads, introducing a breaking change in the behavior of SendToAsync(SocketAsyncEventArgs).
Stephen Toub [Wed, 8 Jan 2020 10:49:14 +0000 (05:49 -0500)]
Augment several "bool IsXx" APIs with [NotNullWhen(true)] (#1372)
If these methods return true, we know the argument passed in wasn't null. Annotating it as such can help reduce warnings in callers.
Anton Lapounov [Wed, 8 Jan 2020 10:12:25 +0000 (02:12 -0800)]
Fix GC heap corruption on ARM. (#1389)
The allocate_in_free code path in allocate_in_expanded_heap incorrectly calculated the large (double) alignment padding size when limiting the plug size (SHORT_PLUGS) if set_padding_on_saved_p was true:
set_padding_in_expand (old_loc, set_padding_on_saved_p, pinned_plug_entry); // Sets the padding flag on the saved plug
...
pad += switch_alignment_size (is_plug_padded (old_loc)); // Reads the padding flag from the old (different!) plug
That caused access violation during a later heap walk since the g_gc_pFreeObjectMethodTable pointer marking the gap was not placed at the right address.
Zoltan Varga [Wed, 8 Jan 2020 06:17:31 +0000 (07:17 +0100)]
Merge pull request #1346 from vargaz/string-ctor-preserve-dep
Add PreserveDependency attributes to String ctors to mark their dependency on the corresponding String:Ctor methods.
Zoltan Varga [Wed, 8 Jan 2020 02:22:31 +0000 (03:22 +0100)]
Fix PreserveDependencyAttribute.cs location.
Elinor Fung [Wed, 8 Jan 2020 02:55:09 +0000 (18:55 -0800)]
Fix reported result in binder tracing for mismatched MVID when loading assembly by path / stream (#1339)
Zoltan Varga [Tue, 7 Jan 2020 03:23:54 +0000 (04:23 +0100)]
Add PreserveDependency attributes to String ctors to mark their dependency on the corresponding String:Ctor methods.
Stephen Toub [Wed, 8 Jan 2020 02:22:46 +0000 (21:22 -0500)]
Fix stack overflow in PhysicalAddress.TryParse taking out CI (#1385)
Swaroop Sridhar [Wed, 8 Jan 2020 01:42:09 +0000 (17:42 -0800)]
Merge pull request #1102 from swaroop-sridhar/release-testing
Add Job for Release testing
Adeel Mujahid [Wed, 8 Jan 2020 00:27:29 +0000 (02:27 +0200)]
Change location of some source files (#1376)
* Move coreclr/src/coreclr/hosts to one level up
* Move native commons to :/eng/native
Tarek Mahmoud Sayed [Wed, 8 Jan 2020 00:10:26 +0000 (16:10 -0800)]
Fix Reading Time Formats with quotes from ICU (#1370)
Matt Galbraith [Tue, 7 Jan 2020 23:09:23 +0000 (15:09 -0800)]
Merge pull request #1373 from dotnet/alpine-3-11
Add Alpine 3.11 testing
Adeel Mujahid [Tue, 7 Jan 2020 22:12:46 +0000 (00:12 +0200)]
Fix musl condition for chroot case (#1369)
Matt Galbraith [Tue, 7 Jan 2020 20:50:17 +0000 (12:50 -0800)]
PR feedback
Toggle meaning of isFullMatrix (PR request)
Swaroop Sridhar [Fri, 20 Dec 2019 22:43:38 +0000 (14:43 -0800)]
Add Job Description for Release testing
This change adds a job description for testing release bits.
It runs outer loop tests on all platforms (with and without R2R) every night at 10pm.
Jan Kotas [Tue, 7 Jan 2020 20:15:13 +0000 (12:15 -0800)]
Delete Tests.props (#1236)
It creates a conficting defintiion of CommonPath that is not really needed for anything
Matt Galbraith [Tue, 7 Jan 2020 19:46:21 +0000 (11:46 -0800)]
Add Alpine 3.11 testing
Juan Hoyos [Tue, 7 Jan 2020 18:53:20 +0000 (10:53 -0800)]
Add createdump to the sharedFramework folder to be added to runtime packages (#1345)
Youssef Victor [Tue, 7 Jan 2020 16:39:01 +0000 (18:39 +0200)]
Add vbproj to .editorconfig (#1354)
Michal Strehovský [Tue, 7 Jan 2020 16:37:10 +0000 (17:37 +0100)]
Reduce confusion in vector layout algorithm (#1356)
We special case layouts of two kinds of vectors: `Vector<T>` and `Vector64/128/256<T>`. Each gets laid out a bit differently. When porting crossgen2 changes to CoreRT (I'm still on leave of absence), it felt odd that code handling vanilla Vector was added to a class responsible for the 64/128/256 variant.
Turns out this was forced by vector handling being really messy. This commit untangles it:
* Rename the class that handles `Vector<T>` so that the name doesn't conflict with the class that handles `Vector64/128/256` and require renaming to use. Use the same name as on the CoreRT side.
* Unnest the class and move the new method `IsVectorOfTType` there.
Carol Eidt [Tue, 7 Jan 2020 16:30:13 +0000 (08:30 -0800)]
Update LSRA document (#1181)
* Update LSRA document
Carol Eidt [Tue, 7 Jan 2020 16:24:19 +0000 (08:24 -0800)]
Kill local asserts for partial write in copyBlock (#1279)
* Kill local asserts for partial write in copyBlock
Fix #1241
Jan Vorlicek [Tue, 7 Jan 2020 16:21:34 +0000 (17:21 +0100)]
Another PAL cleanup (#1359)
* Make _fullpath private
* Make _open_osfhandle private
* Remove GetWriteWatch and ResetWriteWatch
* Remove atexit
* Remove div and div_t
* Remove PAL__getcwd
* Remove STACK_ALIGN_BITS and STACK_ALIGN_REQ
* Remove couple of LANG_* constants
Michal Strehovský [Tue, 7 Jan 2020 16:21:17 +0000 (17:21 +0100)]
Update shared parts of crossgen2 (#1358)
* Port changes from CoreRT.
* Stop sharing embedded object nodes. These are becoming increasingly difficult to share and there's little value.
Egor Bogatov [Tue, 7 Jan 2020 16:12:19 +0000 (19:12 +0300)]
Fix TypeIntrinsics_il tests (#1271)
* Fix TypeIntrinsics_il.il
* Remove CLRTestPriority
Carol Eidt [Tue, 7 Jan 2020 15:43:01 +0000 (07:43 -0800)]
Fix LSRA dump of GC Kills (#1288)