platform/upstream/coreclr.git
8 years agoMerge pull request #5121 from dotnet/revert-5075-Fix4940
Jan Kotas [Fri, 20 May 2016 06:39:22 +0000 (23:39 -0700)]
Merge pull request #5121 from dotnet/revert-5075-Fix4940

Revert "Change to reject varag methods"

8 years agoMerge pull request #5030 from kouvel/NamedMutex
Koundinya Veluri [Fri, 20 May 2016 04:57:09 +0000 (21:57 -0700)]
Merge pull request #5030 from kouvel/NamedMutex

Add named mutex for cross-process synchronization

8 years agoMerge pull request #5115 from brianrob/corelib_perfmap
Brian Robbins [Fri, 20 May 2016 04:05:43 +0000 (21:05 -0700)]
Merge pull request #5115 from brianrob/corelib_perfmap

Generate perfmap symbol file for System.Private.CoreLib

8 years agoRevert "Change to reject varag methods"
Gaurav Khanna [Fri, 20 May 2016 04:03:33 +0000 (21:03 -0700)]
Revert "Change to reject varag methods"

8 years agoMerge pull request #5075 from gkhanna79/Fix4940
Gaurav Khanna [Fri, 20 May 2016 03:54:08 +0000 (20:54 -0700)]
Merge pull request #5075 from gkhanna79/Fix4940

Change to reject varag methods

8 years agoMerge pull request #5118 from ramarag/fixexchangeTStest
Rama krishnan Raghupathy [Fri, 20 May 2016 03:06:31 +0000 (20:06 -0700)]
Merge pull request #5118 from ramarag/fixexchangeTStest

Removing ' from bash command line

8 years agoMerge pull request #5092 from yizhang82/icastable-fix2
Yi Zhang (CLR) [Fri, 20 May 2016 03:01:43 +0000 (20:01 -0700)]
Merge pull request #5092 from yizhang82/icastable-fix2

Fix x86 only ICastable feature bug when we pass RuntimeTypeHandles

8 years agoMerge pull request #5085 from sivarv/tcfix
Sivarv [Fri, 20 May 2016 02:13:49 +0000 (19:13 -0700)]
Merge pull request #5085 from sivarv/tcfix

Tail call test failure fixes.

8 years agoRemoving ' from bash command line
Rama Krishnan Raghupathy [Fri, 20 May 2016 01:45:16 +0000 (18:45 -0700)]
Removing ' from bash command line

8 years agoGenerate perfmap symbol file for System.Private.CoreLib.
Brian Robbins [Thu, 19 May 2016 23:08:32 +0000 (16:08 -0700)]
Generate perfmap symbol file for System.Private.CoreLib.

8 years agoMerge pull request #5067 from AndyAyersMS/CalleeHashInXml
Andy Ayers [Fri, 20 May 2016 00:33:10 +0000 (17:33 -0700)]
Merge pull request #5067 from AndyAyersMS/CalleeHashInXml

Inliner: small changes to inline xml

8 years agoBaseline more tests
Gaurav Khanna [Fri, 20 May 2016 00:21:40 +0000 (17:21 -0700)]
Baseline more tests

8 years agoFix case-sensitive issue on Linux (#5086)
Jiyoung Giuliana Yun [Fri, 20 May 2016 00:01:53 +0000 (09:01 +0900)]
Fix case-sensitive issue on Linux (#5086)

Fix #5035

The file paths should be case-sensitive on Linux.

Signed-off-by: Jiyoung Yun <jy910.yun@samsung.com>
8 years agoAdd named mutex for cross-process synchronization
Koundinya Veluri [Wed, 20 Apr 2016 13:57:19 +0000 (06:57 -0700)]
Add named mutex for cross-process synchronization

Fixes #3422

- On systems that support pthread process-shared robust recursive mutexes, they will be used
- On other systems, file locks are used. File locks unfortunately don't have a timeout in the blocking wait call, and I didn't find any other sync object with a timed wait with the necessary properties, so polling is done for timed waits.

Shared memory files:
- Session-scoped mutexes (name not prefixed, or prefixed with Local\) go in /tmp/coreclr/shm/session<sessionId>/<mutexName>
- Globally-scoped mutexes (name prefixed with Global\) go in /tmp/coreclr/shm/global/<mutexName>
- Contains shared state, and is mmap'ped into the process, see SharedMemorySharedDataHeader and NamedMutexSharedData for data stored
- Creation and deletion is synchronized using an exclusive file lock on the shm directory
- Any process using the shared memory file holds a shared file lock on the shared memory file
- Upon creation, if the shared memory file already exists, an exclusive file lock is attempted on it, to see if the file data is valid. If no other processes have the mutex open, the file is reinitialized.
- Upon releasing the last reference to a mutex in a process, it will try to get an exclusive lock on the shared memory file to see if any other processes have the mutex opened. If not, the file is deleted, along with the session directory if it's empty. The coreclr and shm directories are not deleted.
- This allows managing the lifetime of mutex state based on active processes that have the mutex open. Depending on how the process terminated, the file may still be left over in the tmp directory, I haven't found anything that can be done about that.

Lock files when using file locks:
- In addition to the shared memory file, we need another file for the actual synchronization file lock, since a file lock on the shared memory file is used for lifetime purposes.
- These files go in /tmp/coreclr/lockfiles/session<sessionId>|global/<mutexName>
- The file is empty, and is only used for file locks

Process data
- See SharedMemoryProcessDataHeader and NamedMutexProcessData for data stored
- Per mutex name, there is only one instance of process data that is ref-counted. They are currently stored in a linked list in SharedMemoryManager. It should use a hash table, but of the many hash table implementations that are already there, none seem to be easily usable in the PAL. I'll look into that and will fix later.
- Refers to the associated shared memory, and knows how to clean up both the process data and shared data
- When using file locks for synchronization, a process-local mutex is also created for synchronizing threads, since file locks are owned at the file descriptor level and there is only one open file descriptor in the process per mutex name. The process-local mutex is locked around the file lock, so that only one thread per process is ever trying to flock on a given file descriptor.

Abandon detection
- When a lock is acquired, the process data is added to a linked list on the owning thread
- When a thread exits, the list is walked, each mutex is flagged as abandoned and released
- For detecting process abruptly terminating, pthread robust mutexes give us that. When using file locks, the file lock is automatically released by the system. Upon acquiring a lock, the lock owner info in the shared memory is checked to see if the mutex was abandoned.

Miscellaneous
- CreateMutex and OpenMutex both create new handles for each mutex opened. Each handle just refers to the process data header for the mutex name.
- Some of the above features are already available in the PAL, but not quite in a way that I can use for this purpose. The existing shared memory, naming, and waiting infrastructure is not suitable for this purpose, and is not used.

8 years agoMerge pull request #4963 from sejongoh/vector3_clear_upperbits
Sejong Oh [Thu, 19 May 2016 23:21:42 +0000 (16:21 -0700)]
Merge pull request #4963 from sejongoh/vector3_clear_upperbits

Vector3 clear upperbits

8 years agoRemove the wrong optimization from the string initialization (#5107)
Tarek Mahmoud Sayed [Thu, 19 May 2016 23:09:03 +0000 (16:09 -0700)]
Remove the wrong optimization from the string initialization (#5107)

The string initilaization misteknly setting some flag for characters under 0x80
the fix here is just remove this part (per Jan Kotas recommendation)

8 years agoMerge pull request #5106 from dotnet-bot/from-tfs
Jan Kotas [Thu, 19 May 2016 23:08:39 +0000 (16:08 -0700)]
Merge pull request #5106 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoFix x86 only ICastable feature bug.
Yi Zhang [Thu, 19 May 2016 06:05:53 +0000 (23:05 -0700)]
Fix x86 only ICastable feature bug.
When we call ICastable.IsInstanceOfInterface, we treat RuntimeTypeHandle as a OBJECTREF, which is incorrect as-per x86 calling convention since RuntimeTypehandle is a struct that contains a RuntimeType ref field and needs to be passed in stack. Our VM simple call helpers CALL_MANAGED_METHOD doesn't handle this correctly (it does the simple thing that always assume all the arguments are passed in register first). I'm fixing this by using a static method that takes RuntimeType instead of RuntimeTypeHandle, then convert it to RuntimeTypehandle.
Also switch to use PREPARE_NONVIRTUAL_CALLSITE(METHODID) as per Jan's suggestion.
ICastable test is now enabled in x86 for both JITs.

8 years agoMerge pull request #5074 from jhendrixMSFT/master
Joel Hendrix [Thu, 19 May 2016 22:18:56 +0000 (15:18 -0700)]
Merge pull request #5074 from jhendrixMSFT/master

Add sign.builds projects to facilitate real signing in official builds.

8 years agoTail call test failure fixes.
sivarv [Thu, 19 May 2016 00:15:53 +0000 (17:15 -0700)]
Tail call test failure fixes.

8 years agoThrow InvalidProgramException when a method with varargs is encountered.
Gaurav Khanna [Wed, 18 May 2016 04:57:21 +0000 (21:57 -0700)]
Throw InvalidProgramException when a method with varargs is encountered.

8 years agoMerge pull request #5099 from kyulee1/fixr2r
Kyungwoo Lee [Thu, 19 May 2016 21:48:42 +0000 (14:48 -0700)]
Merge pull request #5099 from kyulee1/fixr2r

ARM64: Fix R2R EntryPoint for Intrinsic

8 years agoMerge pull request #5098 from RussKeldorph/objects
Russ Keldorph [Thu, 19 May 2016 21:39:49 +0000 (14:39 -0700)]
Merge pull request #5098 from RussKeldorph/objects

Mark doublealign test as incompatible with GCStress

8 years agoMerge pull request #5100 from RussKeldorph/gcstress
Russ Keldorph [Thu, 19 May 2016 21:39:09 +0000 (14:39 -0700)]
Merge pull request #5100 from RussKeldorph/gcstress

Fix GCStress=0xC when build(tests).cmd called from outside root

8 years agoMerge pull request #5091 from gkhanna79/Fix4531
Gaurav Khanna [Thu, 19 May 2016 19:54:30 +0000 (12:54 -0700)]
Merge pull request #5091 from gkhanna79/Fix4531

Make Microsoft.NetCore.JIT package dependency of CoreCLR package

8 years agoInitialize unused upper bits of Vector3 arguments and return registers for
Sejong OH [Fri, 13 May 2016 00:06:45 +0000 (17:06 -0700)]
Initialize unused upper bits of Vector3 arguments and return registers for
Linux

JIT requires the unused upper bits to be cleared but native compiler
doesn't clear the upper bits. This changes clear those bits for Reverse
PInvoke arguments and PInvoke return.

8 years agoChecking in on behalf of JKuhne
Alex Ghiondea [Thu, 19 May 2016 19:17:41 +0000 (12:17 -0700)]
Checking in on behalf of JKuhne

Fixing build break in K caused by the long file path work.

[tfs-changeset: 1606374]

8 years agoAdd sign.builds projects to facilitate real signing in official builds.
Joel Hendrix [Thu, 19 May 2016 18:33:21 +0000 (11:33 -0700)]
Add sign.builds projects to facilitate real signing in official builds.

New build project, sign.builds, is to be built post building of binaries
so that they can be real-signed in the official build.
Updated some project files with the correct strong name certificate.

8 years agoFix GCStress=0xC when build(tests).cmd called from outside root
Russ Keldorph [Thu, 19 May 2016 17:44:38 +0000 (10:44 -0700)]
Fix GCStress=0xC when build(tests).cmd called from outside root

Use absolute path intead of relative path when invoking
setup-runtime-dependencies.cmd, which is required for GCStress=0xC to
work.

8 years agoMerge pull request #4991 from swgillespie/gcsimulator-buildandrun
Sean Gillespie [Thu, 19 May 2016 17:17:17 +0000 (10:17 -0700)]
Merge pull request #4991 from swgillespie/gcsimulator-buildandrun

Switch GCSimulator to BuildAndRun so that a shell wrapper gets generated

8 years agoImplement source/line number support for Exception.StackTrace. (#5078)
Mike McLaughlin [Thu, 19 May 2016 17:14:56 +0000 (10:14 -0700)]
Implement source/line number support for Exception.StackTrace. (#5078)

* Implement source/line number support for Exception.StackTrace.

Uses a new helper class (StackTraceSymbols) in System.Diagnostics.StackTrace assembly
to get source/line info for portable PDBs. On Windows both types of pdbs are support;
on xplat only portable PDBs.

The first part of the work is in the GetStackFramesInternal pinvoke to return the neccessary
info needed by the portable pdb reader: MethodToken, assembly path, loaded PE image address
and if in-memory symbol file, the address.

The second part is in the StackFrameHelper in mscorlib. It attempts to load the System.Diagnostics.StackTrace,
gets the new type and creates a delegate to the helper function which is called for each stack
frame.

* Code review feedback.

8 years agoMerge pull request #5031 from wtgodbe/pipelineFix
William Godbe [Thu, 19 May 2016 17:13:10 +0000 (10:13 -0700)]
Merge pull request #5031 from wtgodbe/pipelineFix

Update build scripts to fix small errors

8 years agoARM64: Fix R2R EntryPoint for Intrinsic
Kyungwoo Lee [Thu, 19 May 2016 16:51:56 +0000 (09:51 -0700)]
ARM64: Fix R2R EntryPoint for Intrinsic

This is fix for missing case that we should handle call that needs to pass
indirect cell address. The call appears in rationalizer, which is
originally an intrinsic when we import it.
I also refactor the code so that when we set entryPoint we ensure
tagging such information on the tree.

8 years agoMark doublealign test as incompatible with GCStress
Russ Keldorph [Thu, 19 May 2016 17:09:22 +0000 (10:09 -0700)]
Mark doublealign test as incompatible with GCStress

The test makes assumptions about allocation alignment that cannot be
expected to hold under GCStress, particularly on 32-bit platforms.

8 years agoARM, SOS: correct IP and SP value in ClrStack command (#5077)
SaeHie Park [Thu, 19 May 2016 16:25:00 +0000 (01:25 +0900)]
ARM, SOS: correct IP and SP value in ClrStack command (#5077)

This fixes casting 64bit pointer value in 32bit machines
that show wrong IP and SP values in ClrStack command.

Related Issue: #5037

8 years agoMerge pull request #5057 from RussKeldorph/smallframe
Russ Keldorph [Thu, 19 May 2016 14:55:38 +0000 (07:55 -0700)]
Merge pull request #5057 from RussKeldorph/smallframe

Exclude smallFrame for JIT32

8 years agoMerge pull request #5001 from AlexGhiondea/ReadAppContextOverrides
Senthil [Thu, 19 May 2016 06:48:34 +0000 (23:48 -0700)]
Merge pull request #5001 from AlexGhiondea/ReadAppContextOverrides

Introduce a way to override AppContext switches from the project.json

8 years agoMerge pull request #5076 from dotnet-bot/from-tfs
Jan Kotas [Thu, 19 May 2016 04:47:22 +0000 (21:47 -0700)]
Merge pull request #5076 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoSkip configuration with skipnative option. (#5059)
mkborg [Thu, 19 May 2016 04:44:54 +0000 (08:44 +0400)]
Skip configuration with skipnative option. (#5059)

Fixes #4022.

8 years agoMake Microsoft.NetCore.JIT package dependency of CoreCLR package
Gaurav Khanna [Thu, 19 May 2016 04:38:48 +0000 (21:38 -0700)]
Make Microsoft.NetCore.JIT package dependency of CoreCLR package

8 years agoMerge pull request #5073 from kyulee1/newtest
Kyungwoo Lee [Thu, 19 May 2016 03:36:50 +0000 (20:36 -0700)]
Merge pull request #5073 from kyulee1/newtest

ARM64: New Test Binary Update

8 years agoMerge pull request #5020 from kyulee1/imp
Kyungwoo Lee [Thu, 19 May 2016 01:30:55 +0000 (18:30 -0700)]
Merge pull request #5020 from kyulee1/imp

ARM64: Enable End-To-End ReadyToRun (R2R) Crossgen

8 years agoARM64: New Test Binary Update
Kyungwoo Lee [Wed, 18 May 2016 21:10:57 +0000 (14:10 -0700)]
ARM64: New Test Binary Update

Currently 135 of 9661 are failing.
Those are marked as EXPECTED_FAIL.
~55: mcc failed due to differnet struct calling convetion
~13: regression from the prior tests: marked as EXPECTED_FAIL;REGRESS
~22: failures from new tests: marked as NEW;EXPECTED_FAIL
~45: HFA

Tests.lst head now contains git hash number which tells us what version
we use to build such binaries. This is also used to dispatch different test folder in the lab.

8 years agoPort from tfs: onload exception debugger crash fix (#4868)
sbomer [Wed, 18 May 2016 23:40:12 +0000 (16:40 -0700)]
Port from tfs: onload exception debugger crash fix (#4868)

The ExceptionHijackPersonalityRoutine was retrieving a context from a
particular stack frame, but the calling conventions for the function
executing in that frame allow it to overwrite the context. This was
causing the debugger to crash for exceptions thrown from the onload
method in winforms on x64.

The fix is to instead retrieve the context from the previous stack
frame, whose layout is set up explicitly in ExceptionHijack.

8 years agoMerge pull request #5070 from dotnet-bot/from-tfs
Bruce Forstall [Wed, 18 May 2016 23:18:38 +0000 (16:18 -0700)]
Merge pull request #5070 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoSwitch GCSimulator to BuildAndRun so that a shell wrapper gets generated
Sean Gillespie [Mon, 16 May 2016 18:53:07 +0000 (11:53 -0700)]
Switch GCSimulator to BuildAndRun so that a shell wrapper gets generated

8 years agoUpdate build scripts to fix small errors
wtgodbe [Tue, 17 May 2016 23:27:35 +0000 (16:27 -0700)]
Update build scripts to fix small errors

8 years agoExplicitly reference the file containing the mscorlib reference assembly.
Alex Ghiondea [Wed, 18 May 2016 21:57:00 +0000 (14:57 -0700)]
Explicitly reference the file containing the mscorlib reference assembly.

[tfs-changeset: 1606007]

8 years agoMerge pull request #5039 from adityamandaleeka/sigterm_chaining
Aditya Mandaleeka [Wed, 18 May 2016 21:51:05 +0000 (14:51 -0700)]
Merge pull request #5039 from adityamandaleeka/sigterm_chaining

Put SIGTERM registration behind a PAL initialization flag

8 years agoCache GetLowerBound in Array.Reverse and Array.Sort (#5019)
Hugh Bellamy [Wed, 18 May 2016 21:27:37 +0000 (22:27 +0100)]
Cache GetLowerBound in Array.Reverse and Array.Sort (#5019)

- Avoids calling internal native method 4 times per sort: 2 times is
enough
- Avoids calling internal native method 2 times per reverse: 1 time is
enough

8 years agoARM64: Enable End-To-End ReadyToRun (R2R) Crossgen
Kyungwoo Lee [Sat, 14 May 2016 20:06:57 +0000 (13:06 -0700)]
ARM64: Enable End-To-End ReadyToRun (R2R) Crossgen

Fixes https://github.com/dotnet/coreclr/issues/4649
The immediate issues was NYI on genEmitHelperCalls. The initial
implementation for the missing part was enough to just crossgen System.dll.
But running tests revealed various issues in crossgened binaries (R2R).
Most common user/helper calls in R2R are represented as indirect calls
similar to interface call using virtual stub dispatch cell --
thunk/helper needs a indirect cell address to update the final target
address on the data location. `IsDelayLoadHelper` and `IsLazyHelper` belong to
this case.
Instead of passing such parameter, x64/x86 uses an encoding trick -- it
assumes the call is dispatched like `call [addr]`. So from the return
address, runtime could extract indirect cell address. Unfortunately this is not
an option for arm64 (actually arm as well but I haven't fixed it in this
change) where indirect call on memory is not encodable.
So, I made the following changes:

1. For the call requiring that needs to pass indirect cell address, I
tagged the call tree via `setR2RRelativeIndir`. Tried to be comprehensive,
but I may miss something. Currently, it includes a regular call and
various helpers for (virtual) load function pointer/static data access, etc.
Hopely we change JIT/EE interface somehow that gives us such explicit information.

2. Use the X11 to record indirect cell address for such call tree in
lower similar to VSD.

3. Fixed encodings `ZapIndirectHelperThunk`. In particular the immediate
value/offset for `ldr` should be scaled down 4 times since HW will scale
it 4 times.

4. Implement `genEmitHelperCalls` for indirect case. This is not the case requiring indirect
cell address. This is the case we inlined the indirect helper thunk for
the speed. I'm seeing the case for size opt helper call, we invoke a
direct call to such thunk which actually uses x12 to dispatch the final
target. Likewise, I used x12 for this expansion which seems a trash register that is not
overlapped with arugments with jit helpers like writer barriers.

With this change, I've tested various cases/scenraios locally.
Also I've verified all tests are passed against mscorlib.ni.dll and System.ni.dll.

8 years agoMerge pull request #5069 from ramarag/R2RCleanup
Rama krishnan Raghupathy [Wed, 18 May 2016 21:17:54 +0000 (14:17 -0700)]
Merge pull request #5069 from ramarag/R2RCleanup

Removing hard coded path Fixes #5065

8 years agoMerge pull request #5048 from yizhang82/hostpath-fix
Yi Zhang (CLR) [Wed, 18 May 2016 20:36:13 +0000 (13:36 -0700)]
Merge pull request #5048 from yizhang82/hostpath-fix

Move NATIVE_DLL_SEARCH_DIRECTORIES probing earlier right before we pr…

8 years agoMerge pull request #5044 from RussKeldorph/lifetime2
Russ Keldorph [Wed, 18 May 2016 20:10:28 +0000 (13:10 -0700)]
Merge pull request #5044 from RussKeldorph/lifetime2

Fix and reenable lifetime2

8 years agoRemove Encoding FEATURE_* defines from mscorlib (#5026)
James Ko [Wed, 18 May 2016 20:09:33 +0000 (16:09 -0400)]
Remove Encoding FEATURE_* defines from mscorlib (#5026)

* Remove Encoding FEATURE_* defines from mscorlib

* Remove strange character from Encoding.cs

8 years agoRemoving hard coded path
Rama Krishnan Raghupathy [Wed, 18 May 2016 19:28:20 +0000 (12:28 -0700)]
Removing hard coded path

8 years agoMerge pull request #5062 from dotnet-bot/from-tfs
Bruce Forstall [Wed, 18 May 2016 19:33:06 +0000 (12:33 -0700)]
Merge pull request #5062 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoMerge pull request #5036 from rahku/fixes
Rahul Kumar [Wed, 18 May 2016 19:16:50 +0000 (12:16 -0700)]
Merge pull request #5036 from rahku/fixes

Fix for issue# 3651

8 years agoInliner: small changes to inline xml
Andy Ayers [Fri, 13 May 2016 07:39:14 +0000 (00:39 -0700)]
Inliner: small changes to inline xml

Emit the hash for callees, and check it during replay.

Capture the root method name, and do a simple Xml-safe conversion.

Add a missing newline to the banner message announcing that the replay
policy is in effect.

8 years agoFix to Jit32 test b46659.exe failure.
Venkata Sivaramakrishna Ramadugu [Wed, 18 May 2016 17:59:02 +0000 (10:59 -0700)]
Fix to Jit32 test b46659.exe  failure.

Here is the git hub issue: https://github.com/dotnet/coreclr/issues/4547

The repro case is

     i = -111
     while ( i > 0)
     {
         // another nested loop goes here
     }
     return 100

     After assertionProp, due to while loop condition lot of blocks/stmts get deleted as unreachable but optLoopCount is never decremented.  Later when optLoopCodeMotion() gets called, it iterates through optLoopTable[] and tries to access head

     if  ((loopRmv & loopBit)     ||
            tail == 0               ||
            head->bbTreeList == 0    )

     Since entire method reduces to a single basic block, head will be nullptr and accessing head->bbTreeList is leading to an AV.

     Fix: As a safe fix for RTM, also check whether head is nullptr.

[tfs-changeset: 1605919]

8 years agoMerge pull request #4902 from JeremyKuhne/WindowsLongPath
Jeremy Kuhne [Wed, 18 May 2016 17:41:27 +0000 (10:41 -0700)]
Merge pull request #4902 from JeremyKuhne/WindowsLongPath

Allow long paths in CoreCLR managed code

8 years agoRollback CS#1603899 that led to a JIT assert ngen'ing System.Windows.Forms.dll
Bruce Forstall [Wed, 18 May 2016 17:33:28 +0000 (10:33 -0700)]
Rollback CS#1603899 that led to a JIT assert ngen'ing System.Windows.Forms.dll

====================
007551: Merge pull request #1241 from mikedn/modopt

Extend the DIV/MOD dividend into RDX:RAX only if needed
====================

Assert failure(PID 33656 [0x00008378], Thread: 17792 [0x4580]): Assertion failed 'addr->OperIsAddrMode() || (addr->IsCnsIntOrI() && addr->isContained()) || !addr->isContained()' in 'System.Windows.Forms.CheckedListBox:OnDrawItem(ref):this' (IL size 1216)

    File: e:\dd\projectk\src\ndp\clr\src\jit\emitxarch.cpp Line: 2698
    Image: C:\Windows\Microsoft.NET\Framework64\v4.0.rb1605209\mscorsvw.exe

The tree:

***** BB41, stmt 82 (embedded)
     (  6,  8) [003723] ------------             *  stmtExpr  void  (embedded) (IL 0x109...  ???)
N1045 (  3,  2) [000115] ------------             |        /--*  lclVar    ref    V00 this         u:2 REG rcx $80
N1047 (  1,  4) [002642] ------------             |        +--*  const     long   344 field offset Fseq[idealCheckSize] REG NA $10b
N1049 (  4,  6) [002643] -------N----             |     /--*  +         byref  REG NA $356
N1051 (  6,  8) [000116] ----GO------             |  /--*  indir     int    REG rcx <l:$685, c:$2ef>
N1053 (  6,  8) [003669] DA--GO------             \--*  st.lclVar int    V172 cse1         rcx REG rcx RV

During codegen:

Generating BB41, stmt 71        Holding variables: [rbx rsi rdi r12-r15]

Generating: N1043 (  3,  2) [000114] ------------             *  lclVar    int    V05 loc3         u:3 r12 (last use) REG r12 RV $31a
Generating: N1045 (  3,  2) [000115] ------------             *  lclVar    ref    V00 this         u:2 REG rcx $80
IN00db:        mov      rcx, gword ptr [V00 rbp+10H]
                            GC regs: 00000040 {rsi} => 00000042 {rcx rsi}
Generating: N1047 (  1,  4) [002642] ------------             *  const     long   344 field offset Fseq[idealCheckSize] REG NA $10b
Generating: N1049 (  4,  6) [002643] -------N----             *  +         byref  REG NA $356
Generating: N1051 (  6,  8) [000116] ----GO------             *  indir     int    REG rcx <l:$685, c:$2ef>
... assert ...

(This is rollback #2: the TFS/GitHub mirror unfortunately undid rollback CS#1605814 with CS#1605840. This change should avoid that problem.)

[tfs-changeset: 1605917]

8 years agoExclude smallframe for JIT32
Russ Keldorph [Wed, 18 May 2016 17:22:45 +0000 (10:22 -0700)]
Exclude smallframe for JIT32

This test was flagged as incompatible with COMPlus_JITMinOpts in the
previous test harness.  It doesn't fail for RyuJIT, though, even under
MinOpts, so, rather than disabling it everywhere under MinOpts, I'm only
disabling it for JIT32.

Fixes #4548

8 years agoPorting the changes to support generic dictionaries lookups in R2R images to the...
Fadi Hanna [Wed, 18 May 2016 17:18:53 +0000 (10:18 -0700)]
Porting the changes to support generic dictionaries lookups in R2R images to the JIT32 implementation, to support the x86 architecture.

All of the changes in importer.cpp were ported from JIT\importer.cpp, which was merged to the coreclr master branch last Friday evening (reviewed and signed-off in a PR on GitHub).

The remaining changes include:
         1) Adjusting #ifdefs to enable the x86 scenario
         2) Implementation of the R2R stub for generic dictionary lookups
         3) Small change in prestub.cpp required for x86 to work correctly

[tfs-changeset: 1605906]

8 years agoExplicitly specify the location from where to pick up the model.corelib.xml
Alex Ghiondea [Wed, 18 May 2016 16:21:25 +0000 (09:21 -0700)]
Explicitly specify the location from where to pick up the model.corelib.xml

[tfs-changeset: 1605881]

8 years agoFix Interop TCs to use long long instead of long (#5054)
Dongyun Jin [Wed, 18 May 2016 15:44:36 +0000 (00:44 +0900)]
Fix Interop TCs to use long long instead of long (#5054)

Some TCs were using long for 64bit variable.
While it is correct on x86, long is 32bit on ARM.
Fix TCs to use long long instead so that they can pass on ARM as well.

Fix #5053

Signed-off-by: Dongyun Jin <dongyun.jin@samsung.com>
8 years agoMerge pull request #5055 from dotnet-bot/from-tfs
Bruce Forstall [Wed, 18 May 2016 14:19:23 +0000 (07:19 -0700)]
Merge pull request #5055 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoFix crossgen /createpdb when input filename is given without path (#5045)
John Chen [Wed, 18 May 2016 09:46:23 +0000 (02:46 -0700)]
Fix crossgen /createpdb when input filename is given without path (#5045)

Fix a bug in "crossgen /createpdb" command that can cause it to fail
if the input filename is given without a path.

8 years agoARM64: Fix LR update in a few stubs (#5013)
Swaroop Sridhar [Wed, 18 May 2016 09:44:44 +0000 (02:44 -0700)]
ARM64: Fix LR update in a few stubs (#5013)

Fix LR capture/restore in a few stubs helpers.
This change fixed failures in 6 GC tests under GCStress.

8 years agoAdd private properties for debugger access to Task fields (#5024)
Stephen Toub [Wed, 18 May 2016 09:43:08 +0000 (05:43 -0400)]
Add private properties for debugger access to Task fields (#5024)

8 years agoAdd missing method contract
Jan Kotas [Wed, 18 May 2016 09:37:56 +0000 (02:37 -0700)]
Add missing method contract

[tfs-changeset: 1605842]

8 years agoString.Equals: Don't typecast if object refs are the same (#5021)
James Ko [Wed, 18 May 2016 09:19:30 +0000 (05:19 -0400)]
String.Equals: Don't typecast if object refs are the same (#5021)

8 years agoMerge pull request #5050 from dotnet-bot/from-tfs
Jan Kotas [Wed, 18 May 2016 09:18:20 +0000 (02:18 -0700)]
Merge pull request #5050 from dotnet-bot/from-tfs

Merge changes from TFS

8 years agoMerge pull request #5028 from ramarag/R2Rtests
Rama krishnan Raghupathy [Wed, 18 May 2016 07:59:10 +0000 (00:59 -0700)]
Merge pull request #5028 from ramarag/R2Rtests

Guard Crossgen Test Runtime logic with RunCrossGen Env
this fixes https://github.com/dotnet/coreclr/issues/3759

8 years agoAdd missing .gitmirrorall file
Matt Ellis [Wed, 18 May 2016 06:39:18 +0000 (23:39 -0700)]
Add missing .gitmirrorall file

[tfs-changeset: 1605815]

8 years agoMove NATIVE_DLL_SEARCH_DIRECTORIES probing earlier right before we probe for absolute...
Yi Zhang [Wed, 18 May 2016 06:01:51 +0000 (23:01 -0700)]
Move NATIVE_DLL_SEARCH_DIRECTORIES probing earlier right before we probe for absolute path and assembly path. This is needed to make sure servicing directories (set in NATIVE_DLL_SEARCH_DIRECTORIES by host) are honored correctly before everything else, along with rest of the NATIVE_DLL_SEARCH_DIRECTORIES probing path. In order to preserve compat with UWP host, I kept the rest of the interop probing logic even when NATIVE_DLL_SEARCH_DIRECTORIES probing path is specified.
To my horror I discovered the fact that we actually never set hmod when loading DLL from NATIVE_DLL_SEARCH_DIRECTORIES probing, and the probably the only saving grace is that our ultimate fallback LoadLibrary call after that would "just work" and return the HMODULE we just loaded... This is now fixed.

8 years agoMerge pull request #5034 from dotnet/revert-4969-RelaxALC
Gaurav Khanna [Wed, 18 May 2016 05:50:26 +0000 (22:50 -0700)]
Merge pull request #5034 from dotnet/revert-4969-RelaxALC

Revert "Enable overriding assemblies from TPA"

8 years agoMerge pull request #5016 from tijoytom/master
tijoytom [Wed, 18 May 2016 05:36:25 +0000 (22:36 -0700)]
Merge pull request #5016 from tijoytom/master

[Do not Merge]Renaming ZeroFreeAllocMemUTF8 to ZeroFreeCoTaskMemUTF8

8 years agoMerge branch 'master' of https://github.com/dotnet/coreclr
tijoytom [Wed, 18 May 2016 05:29:45 +0000 (22:29 -0700)]
Merge branch 'master' of https://github.com/dotnet/coreclr

8 years agoMerge pull request #5032 from kspawa/r2rngen
Kshama Pawar [Wed, 18 May 2016 04:50:56 +0000 (21:50 -0700)]
Merge pull request #5032 from kspawa/r2rngen

Treat R2R images as NI for profiling

8 years agoFix and reenable lifetime2
Russ Keldorph [Wed, 18 May 2016 04:33:51 +0000 (21:33 -0700)]
Fix and reenable lifetime2

lifetime2 is another questionable test case that relies on lifetime
guarantees that don't exist.  Per our convention, force optimization on
and mark it as "optimization sensitive" to make a best effort to get it to
pass.

Fixes #1037

8 years agoLinux/ARM: Change default clang version from 3.5 to 3.6 (#4928)
Geunsik Lim [Wed, 18 May 2016 02:44:23 +0000 (11:44 +0900)]
Linux/ARM: Change default clang version from 3.5 to 3.6 (#4928)

In case of the cross compilation on the ubuntu 14.04 x64
for Linux/ARM, the default version of clang/llvm that is
generated by "#> sudo ./cross/build-rootfs.sh arm" command
is 3.6 without 3.5 as following:

u14.04@x64> ls ./coreclr/cross/rootfs/arm/usr/lib/clang/3.6
include  lib
u14.04@x64> ls ./coreclr/cross/rootfs/arm/usr/lib/llvm-3.6/
bin  build  include  lib  share

Let's keep the clang/llvm version consistently by moving
the default version from 'clang3.5' to 'clang3.6' at build-time
for Linux/ARM.

Signed-off-by: Geunsik Lim <geunsik.lim@samsung.com>
8 years agoFix for issue# 3651
Rahul Kumar [Wed, 18 May 2016 02:26:06 +0000 (19:26 -0700)]
Fix for issue# 3651

8 years agoRevert "Enable overriding assemblies from TPA"
Gaurav Khanna [Wed, 18 May 2016 01:46:21 +0000 (18:46 -0700)]
Revert "Enable overriding assemblies from TPA"

8 years agoGuard Crossgen Test Runtime logic with RunCrossGen Env
Rama Krishnan Raghupathy [Sat, 14 May 2016 02:41:46 +0000 (19:41 -0700)]
Guard Crossgen Test Runtime logic with RunCrossGen Env
Adds Semaphores to prevent race conditions

8 years agoMerge pull request #4994 from pgavlin/gh4912.PAGE_SIZE.jit
Pat Gavlin [Wed, 18 May 2016 00:19:37 +0000 (17:19 -0700)]
Merge pull request #4994 from pgavlin/gh4912.PAGE_SIZE.jit

Encapsulate or replace references to CORINFO_PAGE_SIZE.

8 years agoMerge pull request #5015 from pgavlin/jitOnDllHooks
Pat Gavlin [Wed, 18 May 2016 00:18:07 +0000 (17:18 -0700)]
Merge pull request #5015 from pgavlin/jitOnDllHooks

Remove jitOnDllProcess{Attach,Detach}.

8 years agoAdd more help options to corerun (#4973)
James Ko [Wed, 18 May 2016 00:16:32 +0000 (20:16 -0400)]
Add more help options to corerun (#4973)

8 years agoRemove redundant Rank checks from SZArrayHelper/Enumerator (#5022)
James Ko [Wed, 18 May 2016 00:12:15 +0000 (20:12 -0400)]
Remove redundant Rank checks from SZArrayHelper/Enumerator (#5022)

8 years agoPut SIGTERM registration behind a PAL initialization flag.
Aditya Mandaleeka [Wed, 18 May 2016 00:06:28 +0000 (17:06 -0700)]
Put SIGTERM registration behind a PAL initialization flag.

8 years agoSome test cases require more than 2GB RAM (#4778)
Dongyun Jin [Wed, 18 May 2016 00:01:05 +0000 (09:01 +0900)]
Some test cases require more than 2GB RAM (#4778)

Some test cases try to allocate 2GB, which is huge
for some testing environments.

Fix #4714

Signed-off-by: Dongyun Jin <dongyun.jin@samsung.com>
8 years agoTreat R2R images as NI for profiling
Kshama Pawar [Tue, 17 May 2016 23:42:35 +0000 (16:42 -0700)]
Treat R2R images as NI for profiling

8 years agoMerge pull request #4995 from adityamandaleeka/reenable_native_callable
Aditya Mandaleeka [Tue, 17 May 2016 23:37:35 +0000 (16:37 -0700)]
Merge pull request #4995 from adityamandaleeka/reenable_native_callable

Reenable NativeCallable test on x86 legacy backend

8 years agoUpdate contributing-workflow.md
Aditya Mandaleeka [Tue, 17 May 2016 23:35:45 +0000 (16:35 -0700)]
Update contributing-workflow.md

Fix broken link to CODE_OWNERS file.

8 years agoFix breakpoints on jitted code for Linux/ARM (#5014)
Dmitri-Botcharnikov [Tue, 17 May 2016 23:21:36 +0000 (03:21 +0400)]
Fix breakpoints on jitted code for Linux/ARM (#5014)

8 years agoStart using CODE_OWNERS.TXT (#5027)
noahfalk [Tue, 17 May 2016 23:18:15 +0000 (16:18 -0700)]
Start using CODE_OWNERS.TXT (#5027)

8 years agoEnable Checksum csc param for mscorlib (#4996)
Jose Perez Rodriguez [Tue, 17 May 2016 22:21:08 +0000 (17:21 -0500)]
Enable Checksum csc param for mscorlib (#4996)

Enable Checksum csc param for mscorlib

8 years agoMerge pull request #4954 from kvochko/sos_tests
noahfalk [Tue, 17 May 2016 22:10:20 +0000 (15:10 -0700)]
Merge pull request #4954 from kvochko/sos_tests

Added tests for libsosplugin

8 years agoupdating model.corlib.xml
tijoytom [Tue, 17 May 2016 22:03:35 +0000 (15:03 -0700)]
updating model.corlib.xml

8 years agoMerge branch 'master' of https://github.com/dotnet/coreclr
tijoytom [Tue, 17 May 2016 22:01:38 +0000 (15:01 -0700)]
Merge branch 'master' of https://github.com/dotnet/coreclr