platform/upstream/coreclr.git
5 years agoMake Import of Helix.Sdk conditional in tests/helixpublishwitharcade.proj (#22360)
Egor Chesakov [Sat, 2 Feb 2019 05:01:53 +0000 (21:01 -0800)]
Make Import of Helix.Sdk conditional in tests/helixpublishwitharcade.proj (#22360)

5 years agoMove Marshal::UnsafeAddrOfPinnedArrayElement to shared partition (#22359)
Marek Safar [Sat, 2 Feb 2019 03:48:58 +0000 (04:48 +0100)]
Move Marshal::UnsafeAddrOfPinnedArrayElement to shared partition (#22359)

5 years agoMake Exception._HResult private (#22354)
Jan Kotas [Sat, 2 Feb 2019 03:47:52 +0000 (19:47 -0800)]
Make Exception._HResult private (#22354)

5 years agoFix the scratch register used for arm64 stack probing (#22313)
Bruce Forstall [Sat, 2 Feb 2019 00:09:10 +0000 (16:09 -0800)]
Fix the scratch register used for arm64 stack probing (#22313)

* Fix the scratch register used for arm64 stack probing

In some circumstances, genFnProlog might pick a callee-saved
register to be the initReg. This doesn't work for stack probing
since we might need a scratch register before we save the
callee-save regs. So, always use R9 (REG_SCRATCH) instead.

Fixes #22284

* Formatting

5 years agoPreserve Vector Arg registers on Arm64 (#22257)
Carol Eidt [Fri, 1 Feb 2019 23:35:28 +0000 (15:35 -0800)]
Preserve Vector Arg registers on Arm64 (#22257)

* Preserve Vector Arg registers on Arm64

Fix #14371

5 years agoRemove stale warning 420 pragmas (#22352)
Stephen Toub [Fri, 1 Feb 2019 22:26:36 +0000 (17:26 -0500)]
Remove stale warning 420 pragmas (#22352)

It used to be that the CS0420 warning ("a reference to a volatile field will not be treated as volatile") would fire when a volatile was used with an Interlocked.* operation.  That warning was unnecessary, as Interlocked.* would itself provide the relevant barriers, and these functions were special cased in Roslyn a long time ago.  But there are still lots of places where pragmas disabling the warning have stuck around.  I'm deleting them all.

5 years agoMove ResourceManager to shared CoreLib. (#22249)
Filip Navara [Fri, 1 Feb 2019 22:25:56 +0000 (23:25 +0100)]
Move ResourceManager to shared CoreLib. (#22249)

* Move ResourceManager to shared CoreLib.

* Split UAP code out of ResourceManager.cs into separate files.

* Remove ResourceManager.NonUap.cs.

* Do not access internal field (CoreRT build fix).

* Fix non-Windows builds (CoreRT).

* Remove obsolete Debug code.

5 years agoExclude RecursiveException and MultipleException from GCStress runs. (#22340)
Sergey Andreenko [Fri, 1 Feb 2019 20:20:51 +0000 (12:20 -0800)]
Exclude RecursiveException and MultipleException from GCStress runs. (#22340)

Fixes #20460.

See issue #22339.

5 years agoUpdate the double/float formatters to return the shortest roundtrippable string....
Tanner Gooding [Fri, 1 Feb 2019 19:58:35 +0000 (11:58 -0800)]
Update the double/float formatters to return the shortest roundtrippable string. (#22040)

* Updating Number.Formatting to always compute a round-trippable number by default.

* Adding a third-party notice entry for the Grisu3 reference implementation

* Fixing up the Grisu3 algorithm to better match the reference implementation (including comments, etc)

* Porting the Grisu3 implementation that generates the shortest roundtrippable sequence.

* Updating the Dragon4 algorithm to support printing the shortest roundtrippable string.

* Fixing the double/float formatters to ignore precision specifiers for 'R'

* Fixing the double/float formatters to handle +0.0 and -0.0

* Fix the casing of `point` in THIRD-PARTY-NOTICES

Co-Authored-By: tannergooding <tagoo@outlook.com>
* Fixing the double/float formatting code to consider a precision specifier of 0 to be the same as default.

* Fixing the double/float formatter so that nMaxDigits is set appropriately.

* Changing the double/float formatting logic to account for the format when determining the correct precision to request.

* Updating the double/float formatter to take the format specifier into account when determining the number of digits to request.

* Fixing the double/float formatting code to continue handling zero correctly.

* Disabling some outdated CoreFX tests.

* Responding to various feedback from the PR

5 years agoRefactor helixpublishwitharcade.proj (#22346)
Egor Chesakov [Fri, 1 Feb 2019 19:49:39 +0000 (11:49 -0800)]
Refactor helixpublishwitharcade.proj (#22346)

5 years agoEnsure method is active in PrepareMethod (#22351)
Andy Ayers [Fri, 1 Feb 2019 19:34:31 +0000 (11:34 -0800)]
Ensure method is active in PrepareMethod (#22351)

Fixes an assert when first call to PrepareMethod for an assembly
is an interop method.

5 years agoAnnotate types moved from S.R.Extensions
Michal Strehovský [Thu, 31 Jan 2019 21:51:45 +0000 (13:51 -0800)]
Annotate types moved from S.R.Extensions

Testing how much hate this is going to bring me.

These annotations let ProjectN deal with the fact that we now have two
definitions of System.Environment in the system. This is needed as long
as Project N is stuck on 2.1-level version of the framework.

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoRevert "Re-point Jenkins to new Ubuntu ARM64 queue"
Russ Keldorph [Fri, 1 Feb 2019 04:23:21 +0000 (20:23 -0800)]
Revert "Re-point Jenkins to new Ubuntu ARM64 queue"

This reverts commit 193d5da27fecb7bf95e7342f1c4b9e2b63ce5313.

5 years agoCoreclr gnuport (#22129)
Sinan Kaya [Fri, 1 Feb 2019 15:27:39 +0000 (10:27 -0500)]
Coreclr gnuport (#22129)

* Abstract away NOINLINE statement

MSVC and GNU compilers use different attributes for noinline.
Abstract away compiler differences.

* Replace __sync_swap with __atomic_exchange_n

__sync_swap doesn't exist on GNU. Replacing with __atomic_exchange_n
which is universally available.

* Define CDECL for GNUC

__cdecl is not defined by default on GNU compilers.

* Define gcc version of __declspec(thread)

* Correct pointer casting

A pointer value is usually unsigned long on most platforms.
Casting it to integer causes signedness issues. Use size_t
to be efficient on all 32 and 64 bit architectures.

* Put quotes around the error string

Correct error statement. GNU G++ is picky about the string
following the error statement with ' character in it. It needs
to be enclosed with double quotes.

* Fix casting problem

Seeing these warnings with GNU G++ compiler

src/pal/src/sync/cs.cpp: In function ‘void CorUnix::InternalInitializeCriticalSectionAndSpinCount(PCRITICAL_SECTION, DWORD, bool)’:
src/pal/src/sync/cs.cpp:630:48: warning: converting to non-pointer type ‘SIZE_T {aka long unsigned int}’ from NULL [-Wconversion-null]
         pPalCriticalSection->OwningThread      = NULL;
                                                ^
src/pal/src/sync/cs.cpp: In function ‘void CorUnix::InternalLeaveCriticalSection(CorUnix::CPalThread*, _CRITICAL_SECTION*)’:
src/pal/src/sync/cs.cpp:880:43: warning: converting to non-pointer type ‘SIZE_T {aka long unsigned int}’ from NULL [-Wconversion-null]
         pPalCriticalSection->OwningThread = NULL;
                                           ^

* Abstract optnone compiler attribute

GNU compiler doesn't support optnone attribute.

pal/src/exception/seh-unwind.cpp:449:77: warning: ‘optnone’ attribute directive ignored [-Wattributes]

* Set the aligned attribute for GNU compiler

* Make __rotl and __rotr functions portable

GNU compiler doesn't have an intrinsic for these. Open code them
using the provided implementation.

* Define deprecated attribute for gcc

* Add throw specifier for GCC

/usr/include/string.h:43:28: error: declaration of ‘void* memcpy(void*, const void*, size_t) throw ()’ has a different exception specifier
        size_t __n) __THROW __nonnull ((1, 2));

5 years agoFixing a GC hole in Number.BigInteger (#22334)
Tanner Gooding [Fri, 1 Feb 2019 06:52:01 +0000 (22:52 -0800)]
Fixing a GC hole in Number.BigInteger (#22334)

5 years agoAdd -skipcrossarchnative option to build.cmd (#22091)
Bruce Forstall [Fri, 1 Feb 2019 05:13:00 +0000 (21:13 -0800)]
Add -skipcrossarchnative option to build.cmd (#22091)

* Add -skipcrossarchnative option to build.cmd

This allows you to skip building the cross-architecture native components.

* Add -skipcrossarchnative to build.sh

5 years agoReplace several Nullable<T>.Value with .GetValueOrDefault() (#22297)
Stephen Toub [Fri, 1 Feb 2019 04:20:14 +0000 (23:20 -0500)]
Replace several Nullable<T>.Value with .GetValueOrDefault() (#22297)

The former does extra work that the latter doesn't do, and they're equivalent when we know it contains a value, such as immediately after a HasValue check.

5 years agoAsyncSteppingFix - Eagerly initialize the AsyncStateMachineBox to avoid expression...
Andrew Au [Thu, 31 Jan 2019 17:07:09 +0000 (09:07 -0800)]
AsyncSteppingFix - Eagerly initialize the AsyncStateMachineBox to avoid expression evaluation in ProjectN async stepping.

[tfs-changeset: 1732745]

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoIncrease arm64 timeout. (#22298)
Sergey Andreenko [Fri, 1 Feb 2019 02:55:58 +0000 (18:55 -0800)]
Increase arm64 timeout. (#22298)

* increase arm64 debug/checked timeout.

* Increase timeout only for corefx jobs.

5 years ago[Arm64/arm32] GCStress failure in tracing/eventsourcetrace/eventsourcetrace/eventsour...
José Rivero [Fri, 1 Feb 2019 02:19:09 +0000 (18:19 -0800)]
[Arm64/arm32] GCStress failure in tracing/eventsourcetrace/eventsourcetrace/eventsourcetrace.sh (#22328)

Marking this test `GCStressIncompatible` = True for now.

5 years agoRe-point Jenkins to new Ubuntu ARM64 queue
Russ Keldorph [Thu, 31 Jan 2019 23:20:41 +0000 (15:20 -0800)]
Re-point Jenkins to new Ubuntu ARM64 queue

Move Jenkins off the Ubuntu.1604.Arm64.Open queue so it can be repurposed for Azure DevOps.  This "Iron" queue should only be used by Jenkins.

5 years agoSupport for dumping GC info in R2RDump (#22216)
Tomáš Rylek [Thu, 31 Jan 2019 23:10:58 +0000 (00:10 +0100)]
Support for dumping GC info in R2RDump (#22216)

Thanks to tons of help from Jan Vorlicek I found out that a huge
failure bucket in the CoreCLR Pri#1 tests is caused by incorrect
GC info for generic methods causing failures in catch clauses
where the exception type is a generic referring to a generic
parameter of the method.

As part of investigation of this issue I'm adding provisions for
dumping the GC info alongside methods to make the differences
between CPAOT and Crossgen easier to analyze.

For now I have only implemented the AMD64 variant of the GC info.
Thankfully it turns out that the same GC info is used by both ARM32
and ARM64 so that the only remaining debt is supplying a variant
of the new code for X86.

Thanks

Tomas

5 years agoUpdate python lookup in CMakeLists.txt to match build.(sh|cmd) (#22145)
Omair Majid [Thu, 31 Jan 2019 21:09:35 +0000 (16:09 -0500)]
Update python lookup in CMakeLists.txt to match build.(sh|cmd) (#22145)

Use the same logic that's used in build.sh/build.cmd to lookup python:
first search for `python3`, then fall back to `python2` and finally to
`python`.

5 years agoFix ProjectN build breaks (dotnet/corert#6924)
Michal Strehovský [Thu, 31 Jan 2019 17:18:36 +0000 (18:18 +0100)]
Fix ProjectN build breaks (dotnet/corert#6924)

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoFix for missing GC ref map record on ARM64 for the X8 register (#22309)
Tomáš Rylek [Thu, 31 Jan 2019 19:23:38 +0000 (20:23 +0100)]
Fix for missing GC ref map record on ARM64 for the X8 register (#22309)

GC stress testing revealed a deficiency in GC ref map on ARM64 - we weren't
emitting the entry corresponding to the X8 return buffer address because
a recent refactoring changed its location in the transition block. This change
generalizes transition block by exposing a new function

GetOffsetOfFirstGCRefMapSlot()

that returns the same value as

GetOffsetOfArgumentRegisters()

except on ARM64 where it returns the offset of the X8 register in the
transition block (above the callee-saved registers and 8-byte padding i.e.
at position 13 corresponding to offset 0x68, or
GetOffsetOfArgumentRegisters() - 8).

Thanks

Tomas

5 years agoAdd design document for unloadability (#22166)
Jan Vorlicek [Thu, 31 Jan 2019 19:17:11 +0000 (20:17 +0100)]
Add design document for unloadability (#22166)

* Add design document for the unloadability

5 years agoReplace Environment.s_macOS with use of PLATFORM_OSX
Stephen Toub [Thu, 31 Jan 2019 14:35:33 +0000 (09:35 -0500)]
Replace Environment.s_macOS with use of PLATFORM_OSX

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoAdapt corert to Environment moving from corefx
Stephen Toub [Wed, 30 Jan 2019 17:18:01 +0000 (12:18 -0500)]
Adapt corert to Environment moving from corefx

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoStop collecting source line information during a stacktrace for the E… (#22305)
Aaron Robinson [Thu, 31 Jan 2019 13:04:43 +0000 (05:04 -0800)]
Stop collecting source line information during a stacktrace for the E… (#22305)

* Stop collecting source line information during a stacktrace for the Exception.Source property.

* Code review feedback

5 years agoAdd managed implementation of GetSystemMaxDBCSCharSize (#22290)
Marek Safar [Thu, 31 Jan 2019 06:57:39 +0000 (07:57 +0100)]
Add managed implementation of GetSystemMaxDBCSCharSize (#22290)

5 years agoAdd timeoutPerTestCollection and use this to control Helix workitem timeout (#22312)
Egor Chesakov [Thu, 31 Jan 2019 06:42:49 +0000 (22:42 -0800)]
Add timeoutPerTestCollection and use this to control Helix workitem timeout (#22312)

5 years agoUpdate CoreFx, IbcData, PgoData to preview.19080.2, master-20190130.3, master-2019013...
dotnet-maestro-bot [Thu, 31 Jan 2019 05:09:06 +0000 (21:09 -0800)]
Update CoreFx, IbcData, PgoData to preview.19080.2, master-20190130.3, master-20190130.3, respectively (#22310)

5 years agoRemove run.exe and config.json (#21608)
Sven Boemer [Thu, 31 Jan 2019 00:24:33 +0000 (16:24 -0800)]
Remove run.exe and config.json (#21608)

This moves us one step away from buildtools and towards arcade.

This replaces run.sh and run.cmd (and all invocations) with "dotnet.sh msbuild" and "msbuild.cmd".
- I'm using these wrapper scripts for now instead of those in eng/common in order to retain the very helpful "Running <command>" output in the build logs.
- I'm using msbuild.cmd instead of dotnet.sh to match the current behavior that uses desktop msbuild on windows, instead of dotnet.

All of the arguments that used to be implicitly generated by run.exe and config.json are now explicit, resulting in longer (but easier to copy+paste) commands. Some of these arguments are likely unnecessary, but in this change my goal is just to match the run.exe behavior. Later, I would like to go through and clean up parameters that don't need to be passed in every invocation. I might also consider moving more of the common arguments out into variables in a later change.

Some of the wrapper scripts now have limited support for parsing "-Argument=Value" style parameters, to support our existing buildpipeline infrastructure, since I thought this was easier to test than changing our buildpipeline definitions. We can remove that parsing logic once we stop using buildpipeline (which has happened at this point).

Some subtle parts of the change:

* Add msbuild.cmd

This simulates the behavior of Tools\msbuild.cmd, which calls desktop msbuild.

* Fix BuildOS processing in package build and publish

Previously, config.json had its own processing that would set
OSName. Instead, we now pass it in explicitly where it's
needed (building packages), or not at all (publishing them).

* Handle "=" in publish-packages.cmd and other scripts

This seems necessary to properly handle the azure access token

* Set __BuildOS in PublishPackages

Required for cases where the build OS isn't detected during the build
such as freebsd.

* Use dotnet msbuild in runtest.py

This prevents us from having to deal with different quote escape
behavior on windows and linux. Previously, arguments like
fileloggerparameters and the logger were given quotes to escape
semicolons in the argument. On unix, this prevented the argument from
being split up by bash. On windows, it seems that the run.cmd/run.exe
tools would prevent the extra quotes from being passed to
msbuild.exe (desktop msbuild would choke on the quotes if they were
passed along).

Unlike desktop msbuild, dotnet msbuild is able to parse the quoted
strings, so we simply psas the quoted arguments directly to it on
windows. We may be able to do the same on unix.

* Fix build-test.sh problem with BuildOS

When copying native files during the unix test build, we rely on
__BuildOS being set. Fixing the import order and always setting
__BuildOS fixes this. We should eventually fix the inconsistent use of
BuildOS vs __BuildOS.

5 years agoReverting "Fixing up the UTF8Formatter and some tests to properly handle -0.0" (...
Tanner Gooding [Wed, 30 Jan 2019 23:35:07 +0000 (15:35 -0800)]
Reverting "Fixing up the UTF8Formatter and some tests to properly handle -0.0" (#22300)

5 years agoDisable outdated test until it is updated
Russ Keldorph [Wed, 30 Jan 2019 20:55:30 +0000 (12:55 -0800)]
Disable outdated test until it is updated

5 years agoUpdate CoreFx, PgoData to preview.19080.1, master-20190130.1, respectively
dotnet-maestro-bot [Wed, 30 Jan 2019 15:14:33 +0000 (07:14 -0800)]
Update CoreFx, PgoData to preview.19080.1, master-20190130.1, respectively

5 years agoDisable tests consistently failing in AzDO CI (#22304)
Egor Chesakov [Wed, 30 Jan 2019 22:59:29 +0000 (14:59 -0800)]
Disable tests consistently failing in AzDO CI (#22304)

* JIT\jit64\opt\cse\hugeexpr1 on Linux/arm

Related issue: https://github.com/dotnet/coreclr/issues/22260

* baseservices\threading\interlocked\compareexchange\CompareExchangeTClass
* baseservices\threading\interlocked\compareexchange\CompareExchangeTClass_1

Related issue: https://github.com/dotnet/coreclr/issues/22303

5 years agoRemove GTF_ADDR_ONSTACK and IsVarAddr.
Eugene Rozenfeld [Tue, 29 Jan 2019 19:26:49 +0000 (11:26 -0800)]
Remove GTF_ADDR_ONSTACK and IsVarAddr.

IsVarAddr was checking GTF_ADDR_ONSTACK  to determine if
the GT_ADDR node is an address of a local. This change removes both
GTF_ADDR_ONSTACK and  IsVarAddr and uses IsLocalAdrExpr instead.
IsLocalAddrExpr uses opcodes to determine if GT_ADDR node is
a local address.

GTF_ADDR_ONSTACK flag is ancient, added before 2002 so I couldn't find
the checkin that introduced it.

I changed the assert to a check and an assignment since simplifications
inside fgMorphArgs between
https://github.com/dotnet/coreclr/blob/1a1e4c4d5a8030cb8d82a2e5b06c2ab357b92534/src/jit/morph.cpp#L3709
(which causes https://github.com/dotnet/coreclr/blob/1a1e4c4d5a8030cb8d82a2e5b06c2ab357b92534/src/jit/morph.cpp#L3057)
and
https://github.com/dotnet/coreclr/blob/1a1e4c4d5a8030cb8d82a2e5b06c2ab357b92534/src/jit/morph.cpp#L3790
may result in more GT_ADDR nodes recognized by IsLocalAdrExpr.

x86 and x64 pmi frameworks had no code diffs and some gcinfo reductions
(15 methods with gcinfo diffs in x86).

Fixes #22190.

5 years agoSwitch CoreLib over to using shared Environment from corefx (#22106)
Stephen Toub [Wed, 30 Jan 2019 21:22:27 +0000 (16:22 -0500)]
Switch CoreLib over to using shared Environment from corefx (#22106)

This requires merging/adapting the implementation with EnvironmentAugments (which goes away completely), the shared files, what corert has, etc.

5 years agoFix CorError header file generation (#22294)
Michal Strehovský [Wed, 30 Jan 2019 20:16:13 +0000 (21:16 +0100)]
Fix CorError header file generation (#22294)

The checked in batch script was only usable if you have a Razzle environment and even then it didn't update the right files.

5 years agofixed pdf link in dotnet-standards.md (#22292)
Meir Blachman [Wed, 30 Jan 2019 15:44:57 +0000 (17:44 +0200)]
fixed pdf link in dotnet-standards.md (#22292)

5 years agoFix namespace of the AmbiguousImplementationException (#22291)
Michal Strehovský [Wed, 30 Jan 2019 12:41:44 +0000 (13:41 +0100)]
Fix namespace of the AmbiguousImplementationException (#22291)

Moving to the namespace that was approved in dotnet/corefx#34124.

5 years agoRecover some of the size lost with AsyncMethodBuilder/Task unification (#6913)
Michal Strehovský [Tue, 29 Jan 2019 21:38:31 +0000 (22:38 +0100)]
Recover some of the size lost with AsyncMethodBuilder/Task unification (#6913)

When we unified the implementations of these across all of our runtimes, we lost all size optimizations people have been doing on the Project N side over the past six years.

This restores a bit of the loss. For one sample app with lots of async usage, this removes 2.1 MB of generic instantiations.

There is more we can do, but I can't spend time on that right now. These two things jumped out on me when I was looking at it back in December and were an easy fix I wanted to do for a while.

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoAdd AmbiguousImplementationException (#22280)
Markus Kitsinger [Wed, 30 Jan 2019 04:57:23 +0000 (22:57 -0600)]
Add AmbiguousImplementationException (#22280)

5 years agoDo not cache error string in SerializationException (#22281)
Markus Kitsinger [Wed, 30 Jan 2019 04:56:32 +0000 (22:56 -0600)]
Do not cache error string in SerializationException (#22281)

5 years agoRunCrossGen in ReadyToRun tests, set TimeoutPerTest, set FailOnWorkItemFailure (...
Egor Chesakov [Wed, 30 Jan 2019 01:17:18 +0000 (17:17 -0800)]
RunCrossGen in ReadyToRun tests, set TimeoutPerTest, set FailOnWorkItemFailure (#22261)

* For ReadyToRun tests in AzDO set RunCrossGen=true during Helix test running

* Set __TestTimeout and introduce TimeoutPerTestInMinutes for test jobs

* Set FailOnWorkItemFailure - fail the AzDO step on Helix work item "catastrophic failure"

5 years agoMove Marshal::StringTo* methods to shared partition (#22266)
Marek Safar [Wed, 30 Jan 2019 00:35:09 +0000 (01:35 +0100)]
Move Marshal::StringTo* methods to shared partition (#22266)

5 years agoFixing several of the HWIntrinsic SoftwareFallback functions to be captureless (...
Tanner Gooding [Tue, 29 Jan 2019 21:34:01 +0000 (13:34 -0800)]
Fixing several of the HWIntrinsic SoftwareFallback functions to be captureless (#22263)

5 years agoTo support container scenario, 2 HardLimit configs are added - (#22180)
Maoni Stephens [Tue, 29 Jan 2019 21:09:32 +0000 (13:09 -0800)]
To support container scenario, 2 HardLimit configs are added - (#22180)

GCHeapHardLimit - specifies a hard limit for the GC heap
GCHeapHardLimitPercent - specifies a percentage of the physical memory this process is allowed to use

If both are specified, GCHeapHardLimit is checked first and only when it's not specified
would we check GCHeapHardLimitPercent.

If neither is specified but the process is running inside a container with a memory
limit specified, we will take this as the hard limit:

max (20mb, 75% of the memory limit on the container)

If one of the HardLimit configs is specified, and the process is running inside a container
with a memory limit, the GC heap usage will not exceed the HardLimit but the total memory
is still the memory limit on the container so when we calculate the memory load it's based
off the container memory limit.

An example,

process is running inside a container with 200mb limit
user also specified GCHeapHardLimit as 100mb.

if 50mb out of the 100mb is used for GC, and 100mb is used for other things, the memory load
is (50 + 100)/200 = 75%.

Some notes on these configs -

+ The limit is the commit size.

+ This is only supported on 64-bit.

+ For Server GC the minimum *reserved* segment size is 16mb per heap, this is to avoid the
scenario where the hard limit is small but the process can use many procs and we end up
with tiny segments which doesn't make sense. We then keep track of the committed on the segments
so the total does not exceed the hard limit.

5 years agoAdd Marshal ansi conversion to shared partition (#22243)
Marek Safar [Tue, 29 Jan 2019 01:16:16 +0000 (02:16 +0100)]
Add Marshal ansi conversion to shared partition (#22243)

* Add Marshal ansi conversion to shared partition

* Ignore bestFit/throwOnUnmappableChar on Unix

* Use same assert in StringToAnsiString

* wideCharLen parameter rename

5 years agoFix various fgMorphInitBlock issues (#21820)
mikedn [Mon, 28 Jan 2019 19:38:19 +0000 (21:38 +0200)]
Fix various fgMorphInitBlock issues (#21820)

* Fix various fgMorphInitBlock issues

* Remove unnecessary destAddr variable

* Extend/fix comment headers

5 years agoFix finding entry point unwind record for a method with fragmented un… (#22202)
Andy Ayers [Mon, 28 Jan 2019 18:45:45 +0000 (10:45 -0800)]
Fix finding entry point unwind record for a method with fragmented un… (#22202)

On Arm and Arm64 unwind records can only cover a limited range of code
(512K and 1MB respectively). So for methods larger than this the jit
will emit multiple "fragment" unwind records to cover the full method code
range. Only the first of these describes the behavior of the method prolog.

When mapping an offset back to a method's entry point unwind, make sure to
find this "root" unwind record instead of one of the internal fragments.

Fixes #19209.

5 years agoUpdate PgoData to master-20190128.1 (#22237)
dotnet-maestro-bot [Mon, 28 Jan 2019 14:16:58 +0000 (06:16 -0800)]
Update PgoData to master-20190128.1 (#22237)

5 years agostring.EndsWith use SequenceEqual not SequenceCompareTo (#22207)
Ben Adams [Mon, 28 Jan 2019 02:29:19 +0000 (03:29 +0100)]
string.EndsWith use SequenceEqual not SequenceCompareTo (#22207)

5 years agoImprove BinaryReader.ReadUInt32() perf by 30% when using MemoryStream (#22102)
TW [Mon, 28 Jan 2019 02:26:53 +0000 (04:26 +0200)]
Improve BinaryReader.ReadUInt32() perf by 30% when using MemoryStream (#22102)

BinaryReader.ReadInt32() has nice optimization which was missing from the ReadUInt32() version. Now both implementations are aligned.

ReadUInt32 and all the 64 bits types are now 30% to 50% faster than original implementation.

5 years agoKeep CoreCLR-specific temporary hack under ifdef
Jan Kotas [Sun, 27 Jan 2019 19:48:53 +0000 (11:48 -0800)]
Keep CoreCLR-specific temporary hack under ifdef

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoShrink Task.Delay when used without cancellation (#22233)
Ben Adams [Mon, 28 Jan 2019 02:01:22 +0000 (03:01 +0100)]
Shrink Task.Delay when used without cancellation (#22233)

5 years agoRemove several uses of VoidTaskResult (#22238)
Stephen Toub [Sun, 27 Jan 2019 19:15:08 +0000 (14:15 -0500)]
Remove several uses of VoidTaskResult  (#22238)

* Remove defunct netstandard code from ValueTask

* Remove several uses of VoidTaskResult

Currently TrySetResult/Canceled/Exception live on `Task<T>`.  There's no reason `TrySetCanceled` and `TrySetException` need to live there, as they only access state from the base `Task`, and so can be moved down.  `TrySetResult` needs the `TResult`, however in a variety of cases `Task<T>` is used with a `VoidTaskResult`, and for such cases we can just have a parameterless `TrySetResult()` on the base class as well, which can be used any time there is no `TResult` or when `default(TResult)` is the desired result.  This lets us switch several cases where we were using `Task<VoidTaskResult>` to just be `Task`, which saves 8 bytes on the task instance on 64-bit.  It also avoids an Interlocked.Exchange as part of the TrySetResult call.

This primarily affects Task.Delay and the non-generic variants of Task.WhenAll, ValueTask.AsTask(), Task.FromCanceled, and Task.FromException.

5 years agoMove Timer to shared CoreLib partition. (#22231)
Filip Navara [Sun, 27 Jan 2019 19:12:30 +0000 (20:12 +0100)]
Move Timer to shared CoreLib partition. (#22231)

* Move Timer to shared CoreLib partition.

* Move SignalNoCallbacksRunning back to shared code, add static EventWaitHandle.Set.

* Remove AppDomain references from shared Timer code, reshuffle some code out of SetTimer in non-shared code.

* Change m_ prefix to match code style.

5 years agoUpdate PgoData to master-20190126.1 (#22198)
dotnet-maestro-bot [Sun, 27 Jan 2019 04:26:13 +0000 (20:26 -0800)]
Update PgoData to master-20190126.1 (#22198)

5 years agoMake ResourceManager use Assembly instead of RuntimeAssembly internally. (#21979)
Filip Navara [Sun, 27 Jan 2019 04:25:06 +0000 (05:25 +0100)]
Make ResourceManager use Assembly instead of RuntimeAssembly internally. (#21979)

5 years agoMove Marshal::ZeroFree* to shared partition (#22230)
Marek Safar [Sun, 27 Jan 2019 01:08:53 +0000 (02:08 +0100)]
Move Marshal::ZeroFree* to shared partition (#22230)

* Move Marshal::ZeroFree* to shared partition

* Add managed version of SysStringByteLen

5 years agoCleanup stackoverflow handling leftovers (#22228)
Jan Kotas [Sat, 26 Jan 2019 17:26:16 +0000 (09:26 -0800)]
Cleanup stackoverflow handling leftovers (#22228)

5 years agoFactor WinRT GetFolderPath into static accessible via reflection (#34873)
Stephen Toub [Sat, 26 Jan 2019 11:26:35 +0000 (06:26 -0500)]
Factor WinRT GetFolderPath into static accessible via reflection (#34873)

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoMove more Marshal methods to shared partition (#22205)
Marek Safar [Sat, 26 Jan 2019 02:32:24 +0000 (03:32 +0100)]
Move more Marshal methods to shared partition (#22205)

* Move more Marshal methods to shared partition

* Implement ThrowExceptionForHR in managed

5 years agoSubmit to one Helix queue on pull requests and CI builds. Add CI trigger for master...
Egor Chesakov [Sat, 26 Jan 2019 01:05:33 +0000 (17:05 -0800)]
Submit to one Helix queue on pull requests and CI builds. Add CI trigger for master branch (#22211)

This limits pull request and CI (push) triggered jobs to running only on one *default* Helix queue:

Linux arm32 - Ubuntu.1404.Arm32.Open
Linux arm64 - Ubuntu.1804.Arm64.Open
Linux x64 - Ubuntu.1804.Amd64.Open

Windows_NT x86/x64 - Windows.10.Amd64.Open
Windows_NT arm32/arm64 - we don't run these in AzDO

OSX - OSX.1013.Amd64.Open

Scheduled and manually triggered jobs will use more than one Helix queues for more comprehensive checking.

This enables CI (push) trigger for master branch. During CI build the following scenarios for Pri1 tests are going to run:
- normal
- no_tiered_compilation

5 years agoSwitch to dnceng-linux-external-temp pool (#22220)
Egor Chesakov [Sat, 26 Jan 2019 01:02:12 +0000 (17:02 -0800)]
Switch to dnceng-linux-external-temp pool (#22220)

5 years agoFix eventsource metadata exception (#21993)
Sung Yoon Whang [Fri, 25 Jan 2019 21:35:37 +0000 (13:35 -0800)]
Fix eventsource metadata exception (#21993)

* Do not throw an exception when we fail to generate metadata for eventsource

* some build failures

* return null directly here

* remove FAILURE_UNSUPPORTED_TYPE

* address PR comments

* address more pr comments

* Fix potential incorrect generation of metadata blob when property metadata generation fails

5 years agothis test should not be run under gcstress (#22214)
Sergey Andreenko [Fri, 25 Jan 2019 20:21:09 +0000 (12:21 -0800)]
this test should not be run under gcstress (#22214)

5 years agoSkip rangecheck in string.EndsWith(char) (#22206)
Ben Adams [Fri, 25 Jan 2019 19:03:33 +0000 (20:03 +0100)]
Skip rangecheck in string.EndsWith(char) (#22206)

5 years agoRemove reflection-based wraper for AssemblyDependencyResolver. (#22195)
Jeremy Koritzinsky [Fri, 25 Jan 2019 16:37:43 +0000 (08:37 -0800)]
Remove reflection-based wraper for AssemblyDependencyResolver. (#22195)

Now that AssemblyDependencyResolver is exposed in CoreFX and has propogated back to CoreCLR, we can remove the reflection-based wrapper for the unit tests for it.

5 years agoUse AggressiveOptimization for intrinsics-based SpanHelpers (#22191)
Ben Adams [Fri, 25 Jan 2019 16:17:58 +0000 (17:17 +0100)]
Use AggressiveOptimization for intrinsics-based SpanHelpers (#22191)

* Use AggressiveOptimization for intrinsics-based SpanHelpers

* Remove from LastIndexOfAny

5 years agoReduce Task's static ctor dependencies (#22172)
Marek Safar [Fri, 25 Jan 2019 15:58:14 +0000 (16:58 +0100)]
Reduce Task's static ctor dependencies (#22172)

* Reduce Task's static ctor dependencies

* Remove defensive check

* Revert "Remove defensive check"

This reverts commit df15ab0350e8f44fc5cbb1a18c5211cd1de11989.

5 years agoCancel CTS immediately if delay is zero. (#18098)
Elliot Prior [Fri, 25 Jan 2019 13:59:30 +0000 (13:59 +0000)]
Cancel CTS immediately if delay is zero. (#18098)

The intended behaviour on specifying a delay of zero is that the CancellationTokenSource is cancelled immediately - before this change, it would depend on the scheduling of the timer callback which could lead to seeing a non-cancelled token. Now, it marks itself as cancelled immediately, without invoking any callbacks (as this method is called only from within the constructor of the object, before any callbacks could have been registered.

5 years agoFixes the desktop build (#22197)
Brian Sullivan [Fri, 25 Jan 2019 11:20:23 +0000 (03:20 -0800)]
Fixes the desktop build (#22197)

* Fixes the desktop build
warning C4311: 'reinterpret_cast': pointer truncation from 'BYTE *' to 'uint32_t'  f:\j6\codegenmirror\src\ndp\clr\src\jit\emit.cpp(5636): warning C4302: 'reinterpret_cast': truncation from 'BYTE *' to 'uint32_t'

* Fix formatting

5 years agoSpeedup .SequenceCompareTo(byte, ...) (#22127)
Ben Adams [Fri, 25 Jan 2019 02:48:43 +0000 (03:48 +0100)]
Speedup .SequenceCompareTo(byte, ...) (#22127)

* Speedup .SequenceCompareTo(byte, ...)

* Rename jump location

* Better annotations for clarity

5 years agoFix collectible AssemblyLoadContext weak handle (#22193)
Jan Vorlicek [Fri, 25 Jan 2019 01:34:39 +0000 (02:34 +0100)]
Fix collectible AssemblyLoadContext weak handle (#22193)

Native runtime keeps short weak handle for collectible AssemblyLoadContext
until the unload time when it gets converted to a strong one. A test
running with gc stress has recently hit a problem caused by the fact
that the weak handle is "short weak" and thus doesn't track
resurrection. Here is what happened:
* The test code has lost reference to the AssemblyLoadContext
* Finalizer was just invoked on that context due to that and so the weak
  short handle was destroyed.
* Native runtime called AssemblyLoadContext.Resolve with the now dead
  handle, the function tried to get the managed object and failed.

The fix is to change the handle type to long weak so that it doesn't get
destroyed before the finalization.

5 years agoUpdating MicrosoftDiagnosticsTracingTraceEventPackageVersion (#22071)
José Rivero [Fri, 25 Jan 2019 01:03:47 +0000 (17:03 -0800)]
Updating MicrosoftDiagnosticsTracingTraceEventPackageVersion (#22071)

5 years agoFix argument checks in ThreadPool.RegisterWaitForSingleObject. (#6887) (#22196)
Dotnet-GitSync-Bot [Fri, 25 Jan 2019 01:02:36 +0000 (17:02 -0800)]
Fix argument checks in ThreadPool.RegisterWaitForSingleObject. (#6887) (#22196)

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoRename TplEtwProvider to TplEventSource (#22194)
Jan Kotas [Fri, 25 Jan 2019 01:01:09 +0000 (17:01 -0800)]
Rename TplEtwProvider to TplEventSource (#22194)

* Rename TplEtwProvider to TplEventSource

* Rename etwLog to log

5 years agoMove Marshal to shared partition (#22146)
Marek Safar [Fri, 25 Jan 2019 01:00:33 +0000 (02:00 +0100)]
Move Marshal to shared partition (#22146)

* Move Marshal to shared partition

* Moved code optimizations

* Cosmetic changes

* Avoid duplicate arguments checks bys shipping Encoding.UTF8.GetString

5 years agoFix SOS GCInfo for Arm/Arm64 (#22188)
Andy Ayers [Fri, 25 Jan 2019 00:18:23 +0000 (16:18 -0800)]
Fix SOS GCInfo for Arm/Arm64 (#22188)

Add missing DECODE_HAS_TAILCALLS flag.

Closes #22163

5 years agoFixing some issues with the decimal parser/formatter (#22070)
Tanner Gooding [Thu, 24 Jan 2019 21:58:28 +0000 (13:58 -0800)]
Fixing some issues with the decimal parser/formatter (#22070)

* Fixing the decimal formatting so that -0 prints as 0

* Fixing the decimal parser to consider digits still in the buffer as part of the zero/non-zero tail check

5 years agoDisable exception handling on MacOS when DLL like DAC, etc. (#22171)
Mike McLaughlin [Thu, 24 Jan 2019 21:37:47 +0000 (13:37 -0800)]
Disable exception handling on MacOS when DLL like DAC, etc. (#22171)

5 years agoMove TaskContinuation to shared partition (#22167)
Marek Safar [Thu, 24 Jan 2019 20:56:41 +0000 (21:56 +0100)]
Move TaskContinuation to shared partition (#22167)

* Move TaskContinuation to shared partition

* Remove TASK_STATE_THREAD_WAS_ABORTED from Task

* Use already existing ThrowAsync

5 years agoMerge pull request #14393 from mikedn/float-neg-abs
Carol Eidt [Thu, 24 Jan 2019 20:23:51 +0000 (12:23 -0800)]
Merge pull request #14393 from mikedn/float-neg-abs

Improve `new Vector2/3/4(c1, c2, c3, c4)` codegen

5 years agoUpdate SSE2 test
Mike Danes [Thu, 24 Jan 2019 09:28:28 +0000 (11:28 +0200)]
Update SSE2 test

5 years agoFix importer spilling in the presence of assignment side effects
Mike Danes [Thu, 13 Sep 2018 19:24:04 +0000 (22:24 +0300)]
Fix importer spilling in the presence of assignment side effects

Atomic ops like GT_CMPXCHG and some HW intrinsic nodes act like assignements so impAppendStmt has to account for them. They can be top level nodes or they can appear in the RHS of a GT_ASG node that perhaps is not considered to have an assignment side effect itself.

5 years agoUse .asArray/.asString syntax for scenarios and helixQueues (#22173)
Egor Chesakov [Thu, 24 Jan 2019 18:25:46 +0000 (10:25 -0800)]
Use .asArray/.asString syntax for scenarios and helixQueues (#22173)

5 years agoFix test build warning. (#22168)
Eugene Rozenfeld [Thu, 24 Jan 2019 17:26:53 +0000 (09:26 -0800)]
Fix test build warning. (#22168)

I decided to suppress the warning for the test instead of
removing the test for test matrix completeness.

Fixes #21519.

5 years agoUpdate PgoData to master-20190124.1 (#22181)
dotnet-maestro-bot [Thu, 24 Jan 2019 17:25:59 +0000 (09:25 -0800)]
Update PgoData to master-20190124.1 (#22181)

5 years agoFix UWP VirtualAlloc import (dotnet/corert#6882)
Michal Strehovský [Thu, 24 Jan 2019 15:25:00 +0000 (16:25 +0100)]
Fix UWP VirtualAlloc import (dotnet/corert#6882)

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoRemove obsolete thread abortion flags. (#22185)
Filip Navara [Thu, 24 Jan 2019 17:25:33 +0000 (18:25 +0100)]
Remove obsolete thread abortion flags. (#22185)

5 years agoDisable PSPSym in filter prolog for CoreRT (#22184)
Andrey Kvochko [Thu, 24 Jan 2019 16:09:34 +0000 (19:09 +0300)]
Disable PSPSym in filter prolog for CoreRT (#22184)

5 years agoWork around build failure (#22138)
Koundinya Veluri [Thu, 24 Jan 2019 15:30:31 +0000 (07:30 -0800)]
Work around build failure (#22138)

Fixes https://github.com/dotnet/coreclr/issues/22103
- There were reports of build failure from https://github.com/dotnet/coreclr/pull/21292, worked around it for now with a todo

5 years agoIOCompletionCallback moved to shared partition
Jan Kotas [Thu, 24 Jan 2019 06:29:26 +0000 (22:29 -0800)]
IOCompletionCallback moved to shared partition

5 years agoMove IOCompletionCallback to shared partition (dotnet/corert#6875)
Marek Safar [Thu, 24 Jan 2019 06:18:48 +0000 (07:18 +0100)]
Move IOCompletionCallback to shared partition (dotnet/corert#6875)

Signed-off-by: dotnet-bot <dotnet-bot@microsoft.com>
5 years agoRemove no-op holder stack validation. (#22182)
Filip Navara [Thu, 24 Jan 2019 12:34:09 +0000 (13:34 +0100)]
Remove no-op holder stack validation. (#22182)

5 years agoFix incorrect usage of wcscpy_s/wcscat_s. (#22153)
Filip Navara [Thu, 24 Jan 2019 07:39:36 +0000 (08:39 +0100)]
Fix incorrect usage of wcscpy_s/wcscat_s. (#22153)

5 years agoFix accidental assignment inside an assert. (#22148)
Calum Grant [Thu, 24 Jan 2019 07:38:46 +0000 (07:38 +0000)]
Fix accidental assignment inside an assert. (#22148)