Aditya Mandaleeka [Sat, 27 Feb 2016 22:34:51 +0000 (14:34 -0800)]
Merge pull request #3140 from adityamandaleeka/env_pal
Fix usage of environment in PAL
Aditya Mandaleeka [Sat, 27 Feb 2016 21:12:17 +0000 (13:12 -0800)]
Merge pull request #3261 from krytarowski/netbsd-support-51
Fix NetBSD build: Define PAL_STDCPP_COMPAT=1 for twowaypipe.cpp
Aditya Mandaleeka [Fri, 12 Feb 2016 03:08:06 +0000 (19:08 -0800)]
Further cleanup of environment code.
Jan Kotas [Sat, 27 Feb 2016 16:38:31 +0000 (08:38 -0800)]
Merge pull request #3406 from BruceForstall/RemoveDupGS2
Remove duplicate GS compiler switch setting
Jan Kotas [Sat, 27 Feb 2016 16:33:58 +0000 (08:33 -0800)]
Merge pull request #3395 from dotnet-bot/from-tfs
Merge changes from TFS
Bruce Forstall [Sat, 27 Feb 2016 16:25:50 +0000 (08:25 -0800)]
Merge pull request #3397 from BruceForstall/AddRuntimeInformationReference
Add reference to System.Runtime.InteropServices.RuntimeInformation
Bruce Forstall [Sat, 27 Feb 2016 16:21:48 +0000 (08:21 -0800)]
Merge pull request #3403 from BruceForstall/FixLinuxStandalone
Add linker magic to non-Windows standalone RyuJIT build
Jan Kotas [Sat, 27 Feb 2016 16:09:41 +0000 (08:09 -0800)]
Merge pull request #3409 from stephentoub/stop_masking_job_signals
Stop masking SIGTSTP and SIGCONT
Pat Gavlin [Sat, 27 Feb 2016 15:52:07 +0000 (07:52 -0800)]
Merge pull request #3407 from mikedn/nra-comments
Update JIT's ArenaAllocator comments
Stephen Toub [Sat, 27 Feb 2016 13:51:58 +0000 (08:51 -0500)]
Stop masking SIGTSTP and SIGCONT
These are needed for job management, e.g. with SIGTSTP masked, ctrl-z doesn't work to suspend a process, and with SIGCONT masked, when the process is resumed it doesn't get notified.
Mike Danes [Sat, 27 Feb 2016 11:34:28 +0000 (13:34 +0200)]
Update JIT's ArenaAllocator comments
A recent allocator refactoring changed function and class names but the old names were still referenced in a comment.
Pat Gavlin [Sat, 27 Feb 2016 03:58:34 +0000 (19:58 -0800)]
Merge pull request #3400 from pgavlin/TruncationCleanup
Clean up some truncation warnings in the JIT.
Mike McLaughlin [Sat, 27 Feb 2016 03:29:35 +0000 (19:29 -0800)]
Merge pull request #3401 from mikem8361/sosdocs
Add the "sosdocsunix.txt" sos help file to the nuget packages.
Bruce Forstall [Sat, 27 Feb 2016 01:40:23 +0000 (17:40 -0800)]
Add linker magic to non-Windows standalone RyuJIT build to make PAL work better when dynamically loading it.
Bruce Forstall [Sat, 27 Feb 2016 03:08:24 +0000 (19:08 -0800)]
Remove duplicate /GS switch
Also, fix a typo.
Matt Ellis [Sat, 27 Feb 2016 02:22:11 +0000 (18:22 -0800)]
Merge pull request #3398 from mikem8361/fixstrike
Fix sos/strike build.
Brian Sullivan [Sat, 27 Feb 2016 01:59:44 +0000 (17:59 -0800)]
Merge pull request #3390 from briansull/mod-fixes
Fixes GT_MOD codegen and three other issues for ARM64
Brian Sullivan [Fri, 26 Feb 2016 22:21:51 +0000 (14:21 -0800)]
Fixes GT_MOD codegen and three other issues for ARM64
Fix Issue 3263 - Incorrect codegen for rem.un - Unisgned Remainder
Fix Issue 3317 - Improve the codegen for divide so that we omit checks when we have a constant divisor
Fix Issue 3326 - assert EA_4BYTE or EA_8BYTE
Fix an assert when using alloca with allocation between 65 and 4095 bytes
Also fixed the name for GT_UDIV and GT_UDIV to contain a "un-" prefix.
Added 10 additional passing tests with tag REL_PASS
Mike McLaughlin [Sat, 27 Feb 2016 01:15:13 +0000 (17:15 -0800)]
Add the "sosdocsunix.txt" sos help file to the nuget packages.
This allows the sos help command to work when installed from our packages.
Jan Vorlicek [Sat, 27 Feb 2016 01:06:01 +0000 (02:06 +0100)]
Merge pull request #3393 from janvorli/make-hw-exception-test-pri-0
Change avtest priority to 0
Pat Gavlin [Sat, 27 Feb 2016 00:48:47 +0000 (16:48 -0800)]
Clean up some truncation warnings in the JIT.
These changes address casts that trigger warnings C4242, C4254, or
C4302 (all of which are truncation-related). Most of the warnings
turned out to be innocuous, but there does seem to be some fishiness
when truncating pointer values to 32-bit integers for the purpose
of generating hash codes.
Mike McLaughlin [Sat, 27 Feb 2016 00:35:19 +0000 (16:35 -0800)]
Fix sos/strike build.
The previous sos commit added some xplat specific files in a "inc" directory that already exists on the tfs side containing the actual dbgeng version of the files. Renamed the directory to "xplat".
Bruce Forstall [Sat, 27 Feb 2016 00:29:57 +0000 (16:29 -0800)]
Add reference to System.Runtime.InteropServices.RuntimeInformation.dll for some JIT tests
Bruce Forstall [Sat, 27 Feb 2016 00:23:38 +0000 (16:23 -0800)]
Merge pull request #3383 from BruceForstall/FixJitSystemIOFilesystemBinding
Change binding of System.IO.FileSystem
Matt Ellis [Fri, 26 Feb 2016 23:57:40 +0000 (15:57 -0800)]
Fix directory name casing in TFS to match what is in git.
[tfs-changeset: 1579410]
Jan Vorlicek [Fri, 26 Feb 2016 23:39:00 +0000 (00:39 +0100)]
Change avtest priority to 0
This change modifies the priority of the coreclr regression test avtest
to 0 so that it is run in the lab as part of the CI. This fulfills the
Matt Ellis [Fri, 26 Feb 2016 22:05:51 +0000 (14:05 -0800)]
Add missing .gitmirror file.
[tfs-changeset: 1579391]
Aditya Mandaleeka [Thu, 11 Feb 2016 02:06:38 +0000 (18:06 -0800)]
Add documentation to environ functions.
Aditya Mandaleeka [Wed, 10 Feb 2016 03:14:28 +0000 (19:14 -0800)]
Fix all callers of environment functions.
Aditya Mandaleeka [Tue, 9 Feb 2016 03:17:09 +0000 (19:17 -0800)]
Remove usage of PALC versions of critsec functions in environ.
Aditya Mandaleeka [Tue, 9 Feb 2016 02:11:15 +0000 (18:11 -0800)]
Modify environment code to avoid unsafe environ usage.
Aditya Mandaleeka [Tue, 9 Feb 2016 00:14:57 +0000 (16:14 -0800)]
Consolidate environment code in environ.cpp.
Chris McKinsey [Fri, 26 Feb 2016 20:37:32 +0000 (12:37 -0800)]
Merge pull request #3381 from cmckinsey/DevDiv_168744
Fix bug in optOptimizeBools for equality of 0/1.
Kyungwoo Lee [Fri, 26 Feb 2016 19:52:22 +0000 (11:52 -0800)]
Merge pull request #3376 from kyulee1/fixcast
ARM64: Fix for casting long to ulong
Mike McLaughlin [Fri, 26 Feb 2016 19:43:57 +0000 (11:43 -0800)]
Merge pull request #3365 from mikem8361/osxexception
Fix the osx exception port forwarding limitations.
Mike McLaughlin [Wed, 27 Jan 2016 00:37:14 +0000 (16:37 -0800)]
Fix the osx exception port forwarding limitations.
Cleanup CThreadMachExceptionHandlerNode/CThreadMachExceptionHandlers.
Send forward exception, save thread info, restore thread back to faulting context and restart faulting instruction and forward.
Add MachExceptionInfo struct used to pass all the exception state around. Bump the IP back one when restarting breakpoint exceptions.
Match lldb exception port options. Found and fix bug with MACH_EXCEPTION_CODES behavior. Needed pack(4) around the exception message structs.
Cleaned up the _ASSERT, CHECK_MACH, FATAL_ERROR macros.
Dump the task level exception ports.
Add more checks to PAL SXS tests. Validate that the fault addr (ExceptionRecord.ExceptionInformation[1]) is correct. Check that a third party signal handler will be properly chained to.
Bruce Forstall [Fri, 26 Feb 2016 19:18:07 +0000 (11:18 -0800)]
Change binding of System.IO.FileSystem
Change the binding from 4.0.1-beta-* to 4.0.0-beta-23302 to match all other
JIT configs. This allows a not-yet-committed test to successfully run, and
doesn't seem to negatively affect existing tests.
Jose Perez Rodriguez [Fri, 26 Feb 2016 18:36:08 +0000 (10:36 -0800)]
Merge pull request #3329 from joperezr/CleanToolRuntime
Correctly cleaning up the ToolRuntime folder once a new version is required
Chris McKinsey [Fri, 26 Feb 2016 18:04:18 +0000 (10:04 -0800)]
Fix bug in optOptimizeBools for equality of 0/1.
The method optIsBoolCond had a clever way to test for a constant
of value 0/1 by casting the resulting gtIconVal to (unsigned) and
comparing for <= 1. Unfortunately, for 64-bit targets gtIconVal as
ssize_t holds 64-bits and the cast to (unsigned) results in truncating
significant bits and results in the optimization falsely recognizing
legal patterns of bit-tests comparing against 0/1. Because 32-bit
targets represent 64-bit long constants as GT_CNS_LNG they do not
suffer from this problem as the code already had exclusions for them.
Andy Ayers [Fri, 26 Feb 2016 16:10:41 +0000 (08:10 -0800)]
Merge pull request #3362 from AndyAyersMS/InlineRefactor5a2
InlineRefactoring: capturing bits of legacy policy
Jan Kotas [Fri, 26 Feb 2016 15:28:42 +0000 (07:28 -0800)]
Merge pull request #3374 from dotnet-bot/from-tfs
Merge changes from TFS
Jan Kotas [Fri, 26 Feb 2016 14:55:55 +0000 (06:55 -0800)]
Merge pull request #3375 from BruceForstall/FixAltJitLoading
Fix CoreCLR altjit loading that was broken by LongFile work
Jan Vorlicek [Fri, 26 Feb 2016 10:21:47 +0000 (11:21 +0100)]
Merge pull request #3357 from janvorli/fix-stackwalk
Fix stack walker on Unix
Kyungwoo Lee [Fri, 26 Feb 2016 07:28:35 +0000 (23:28 -0800)]
ARM64: Fix for casting long to ulong
When converting long (i8) to ulong (u8) with overflow check,
JIT inadvertently inserts 4 byte mov which clears the upper bits.
This explicit conversion is only needed when converting from i4 to u8.
Before
IN0002: 000024 cmp x0, #0
IN0003: 000028 bge G_M62889_IG03
IN0004: 00002C bl CORINFO_HELP_OVERFLOW
IN0005: 000030 mov w0, w0 --> Incorrect conversion clearing the
upper 32 bits of x0.
After
IN0005: 000028 cmp x0, #0
IN0006: 00002C bge G_M62889_IG03
IN0007: 000030 bl CORINFO_HELP_OVERFLOW
Bruce Forstall [Fri, 26 Feb 2016 07:39:39 +0000 (23:39 -0800)]
Fix CoreCLR altjit loading that was broken by LongFile work
Matt Ellis [Fri, 26 Feb 2016 06:40:34 +0000 (22:40 -0800)]
Remove need for mono resgen
The last remaining dependency on mono was to use the resgen
implementation since cross platform we don't have a version of
ResGen.exe. While MSBuild and dotnet-resgen can both convert txt files
into resources, neither of them understand the #if syntax we use in
mscorlib.txt to allow us to remove resources if features of the runtime
are not enabled.
Since we already have a dependency on Python, I've added a dead simple
script that understands the limited preprocessor features we support
and pre-processes the text file. We can then pass that preprocessed file
to MSBuild which will handle converting it to a .resources file.
[tfs-changeset: 1579180]
Jan Kotas [Fri, 26 Feb 2016 04:09:25 +0000 (20:09 -0800)]
Merge pull request #3350 from arnmac/patch-1
Update to freebsd-instructions.md
Jan Kotas [Fri, 26 Feb 2016 04:07:28 +0000 (20:07 -0800)]
Merge pull request #3361 from dotnet-bot/from-tfs
Merge changes from TFS
Stephen Toub [Fri, 26 Feb 2016 04:04:14 +0000 (23:04 -0500)]
Merge pull request #3370 from stephentoub/upgrade_buildtools
Upgrade buildtools to 177
Jan Kotas [Fri, 26 Feb 2016 03:24:36 +0000 (19:24 -0800)]
Merge pull request #3363 from pallavit/UpdateReflectionSA
Remove Assembly.EscapeCodeBase property from CoreCLR
Andy Ayers [Thu, 25 Feb 2016 00:14:32 +0000 (16:14 -0800)]
InlineRefactoring: capturing bits of legacy policy
Rework the logic for force inline, basic block count, il size,
and maxstack so that the policy decides when these values
should inhibit inlining.
Pat Gavlin [Fri, 26 Feb 2016 02:39:03 +0000 (18:39 -0800)]
Merge remote-tracking branch 'upstream/master' into from-tfs
Conflicts:
src/jit/CMakeLists.txt
src/jit/alloc.cpp
src/jit/alloc.h
src/jit/jit.settings.targets
stephentoub [Fri, 26 Feb 2016 02:19:17 +0000 (21:19 -0500)]
Upgrade buildtools to 177
Pat Gavlin [Fri, 26 Feb 2016 02:07:05 +0000 (18:07 -0800)]
Merge pull request #3359 from pgavlin/AllocatorCleanup2
Clean up norls_allocator.
Lubomir Litchev [Fri, 26 Feb 2016 01:37:18 +0000 (17:37 -0800)]
Merge pull request #3319 from LLITCHEV/implicit-RetBuf-return-type
Explicitly change the type of a method with implicit RetBuf to BYREF for System V OSs.
Pat Gavlin [Wed, 24 Feb 2016 18:36:00 +0000 (10:36 -0800)]
Update norls_allocator to conform to the JIT CC.
There are no functional changes aside from the removal of a couple
of fields from norls_allocator and norls_pagdesc. As part of this
change, norls_allocator has been renamed to ArenaAllocator to better
reflect its behavior.
Jan Vorlicek [Wed, 24 Feb 2016 10:51:43 +0000 (11:51 +0100)]
Fix stack walker on Unix
This change fixes the stack walker on Unix to properly account for the cases
when funclet frames were reclaimed due to native frames unwinding and so
they are not on the stack anymore. The stack walker needs to know that to
properly skip reporting GC references for the parent frame of the funclet.
While there was already code attempting to do that, it was incorrectly
skipping some exception trackers and not skipping the current tracker
in case the catch handler was not called yet.
This problem was discovered while running stress tests with GCStress 3.
Aditya Mandaleeka [Fri, 26 Feb 2016 00:41:41 +0000 (16:41 -0800)]
Merge pull request #3364 from adityamandaleeka/semaphore_name_length
Reduce length of debugger semaphore names.
Mike McLaughlin [Thu, 25 Feb 2016 23:57:03 +0000 (15:57 -0800)]
Merge pull request #3336 from mikem8361/sosstack
Enabled more SOS commands and fixed various issues.
Mike McLaughlin [Sat, 20 Feb 2016 08:27:33 +0000 (00:27 -0800)]
Create a version resilient interface between sos (libsos.so) and the lldb plugin
(libsosplugin.so). There is a ILLDBServices interface between the two with a proper
QI/Addref/Release interface so we can add new interfaces later and still be backwards
and forward compatible. Internally in sos there is a DebugClient wrapper around this
interface that makes it look like all the individual dbgeng interfaces (IDebugControl4,
IDebugSymbols, IDebugDataSpaces, etc) even with QI, etc.
Enable the sos the u (clru), DumpStack (dumpstack), EEStack (eestack), DumpIL (dumpil), DumpSig and
DumpSigElem commands (aliases in parans).
Add the clrstack -f option that displays the intermixed native and managed frames
when managed assembly and offsets.
Implement GetContextFromFrame. Enable the _EFN_* exported functions.
Add source file/line number support (native only) via GetLineByOffset.
Fix bug when executed just "sos". Displays help now.
Disable the U -gcinfo option since it isn't implemented on xplat because it uses Windows fibers.
Fixed a problem where some HelperMethodFrames were not unwinding; clrstack would stop without displaying
managed functions on the stack. The HelperMethodFrames were not sometimes unwinding because the lldb
VirtualUnwind used by the DAC was using (via the data target) to do the out of context unwind has some
limitations. lldb doesn't have a way to unwind an arbitrary register context so the VirtualUnwind
implementation searches through the thread's frames until it finds a match and returns the next one. The
match was an exact match and in some cases it didn't find a frame. Changed it to check if the incoming
context's SP is in between frames and return the next one if so.
Only add the special internal "corerun" lldb plugin command in debug builds.
Lubomir Litchev [Tue, 23 Feb 2016 18:24:21 +0000 (10:24 -0800)]
Explicitly change the type of a method with implicit RetBuf to BYREF for
System V.
The System V ABI requires the address implicit RetBuf to be returned in
RAX.
Instead of moving the address of the buffer to RAX in codegenxarch.cpp,
change such methods return type to TYP_BYREF and generate the necessary
code to put the address in RAX.
Added/cleaned some comments.
Fixes issue #3299.
Aditya Mandaleeka [Thu, 25 Feb 2016 21:51:32 +0000 (13:51 -0800)]
Reduce debugger semaphore name length.
Pallavi Taneja [Thu, 25 Feb 2016 21:33:35 +0000 (13:33 -0800)]
Remove Assembly.EscapeCodeBase property from CoreCLR
Andy Ayers [Thu, 25 Feb 2016 21:07:43 +0000 (13:07 -0800)]
Merge pull request #3346 from AndyAyersMS/InlineRefactor5a1
Inline Refactoring: set up policy for bad inlinees
tijoytom [Thu, 25 Feb 2016 20:04:10 +0000 (12:04 -0800)]
Merge pull request #3314 from tijoytom/master
Adding more interop tests.
dotnet-bot [Thu, 25 Feb 2016 19:41:31 +0000 (11:41 -0800)]
Roll back change 1578859 on behalf of pagavlin. Change broke SuperPMI.
[tfs-changeset: 1578925]
Andy Ayers [Tue, 23 Feb 2016 23:21:22 +0000 (15:21 -0800)]
Inline Refactoring: set up policy for bad inlinees
Move inline policies to their own header and cpp file.
Add a method to the policy class to indicate if the policy wants newly
discovered `Never` inline cases to change the callee method attributes
to Noinline. This is an existing optimization that saves time when the
jit sees calls to this callee elsewhere as a possible inline candidates.
For example, in the trace below, at for the call at offset 31, the jit
determines that `ToInt32` is too large to be inlined and so marks it as
noinline. Then when it sees another call to `ToInt31` at offset 44 it
immediately fails the inline attempt.
```
Inlines into RegistryTimeZoneInformation:.ctor(ref):this
...
[IL=0031 TR=000040] [FAILED: too many il bytes] System.BitConverter:ToInt32(ref,int):int
[IL=0044 TR=000049] [FAILED: noinline per IL/cached result] System.BitConverter:ToInt32(ref,int):int
[IL=0057 TR=000058] [FAILED: noinline per IL/cached result] System.BitConverter:ToInt32(ref,int):int
```
Diagnostic and experimental policies may choose to disable this
optimization to make it easier to locally reason about failed inlines.
There were 5 calls to `setMethodAttribs` passing `CORINFO_FLG_BAD_INLINEE`.
This change consolidates 4 of them into the inlining code. The remaining
call is for a method with verification errors. I've left it as is.
Jan Kotas [Thu, 25 Feb 2016 17:40:59 +0000 (09:40 -0800)]
Merge pull request #3356 from bendono/PowerShell
Validate PowerShell prerequisite. Update documentation. -- Issue #3355
Jan Kotas [Thu, 25 Feb 2016 17:09:09 +0000 (09:09 -0800)]
Merge pull request #3358 from dotnet-bot/from-tfs
Merge changes from TFS
Ben Monroe [Thu, 25 Feb 2016 12:51:35 +0000 (21:51 +0900)]
Validate PowerShell prerequisite. Update documentation.
Pat Gavlin [Thu, 25 Feb 2016 15:19:33 +0000 (07:19 -0800)]
Add a new set of APIs for JIT configuration.
These APIs accommodate the retrieval of config values using the JIT
interface rather than the utilcode library. All configuration options
are now initialized upon the first call to compileMethod. The values
of configuration options are available off of an ambient JitConfig
object.
This also changed `JitHost::get*ConfigValue` to use the
`EEConfig_default` policy instead of `REGUTIL_default` in order to
avoid breaking a small set of JIT config options available in release
builds that were using the former. This change is exceedingly
unlikely to adversely affect the behavior of other JIT config options
that were originally fetched using `REGUTIL_default`, since values
for these options should not be present any locations searched
by `EEConfig_default` that are not searched by
`REGUTIL_default` (namely config files).
[tfs-changeset: 1578859]
Aaron Havens [Thu, 25 Feb 2016 14:38:53 +0000 (08:38 -0600)]
Update freebsd-instructions.md
Corrected the typo.
Jan Kotas [Thu, 25 Feb 2016 14:29:11 +0000 (06:29 -0800)]
Merge pull request #3354 from dotnet-bot/from-tfs
Merge changes from TFS
Jan Kotas [Thu, 25 Feb 2016 14:28:59 +0000 (06:28 -0800)]
Merge pull request #3353 from jkotas/gc-update
Update GC from CoreRT
Jan Kotas [Thu, 25 Feb 2016 07:59:25 +0000 (23:59 -0800)]
Merge pull request #3335 from dotnet-bot/from-tfs
Merge changes from TFS
Jan Kotas [Thu, 25 Feb 2016 07:53:11 +0000 (23:53 -0800)]
Fixes for GC update
Jan Kotas [Thu, 25 Feb 2016 07:01:24 +0000 (23:01 -0800)]
Workaround missing UrlEscape
Jan Kotas [Thu, 25 Feb 2016 06:41:19 +0000 (22:41 -0800)]
Update GC from CoreRT
https://github.com/dotnet/corert/tree/master/src/Native/gc
d18f7e5d9fea784b4531aa8988fdad9f3a9cffc3
Jan Kotas [Thu, 25 Feb 2016 06:24:59 +0000 (22:24 -0800)]
Fix build break - add precompiled header
Jan Kotas [Thu, 25 Feb 2016 06:07:57 +0000 (22:07 -0800)]
Merge pull request #3339 from RussKeldorph/msvcrt
Replace reference to msvcr110.dll with msvcrt.dll in test b286991
Jan Kotas [Thu, 25 Feb 2016 04:28:57 +0000 (20:28 -0800)]
Merge pull request #3334 from brianrob/perfmap_crossgen_r2r
Fix PerfMap::GetNativeImageSignature to work for ready to run images
Carol Eidt [Thu, 25 Feb 2016 04:18:43 +0000 (20:18 -0800)]
Merge pull request #3296 from CarolEidt/NonLeafLongOpnd3292
Handle long operators with non-leaf operands
Aaron Havens [Thu, 25 Feb 2016 03:34:49 +0000 (21:34 -0600)]
Update to freebsd-instructions.md
Added note specifically calling out the need to add the clang version to the build command.
Jan Kotas [Thu, 25 Feb 2016 03:06:19 +0000 (19:06 -0800)]
Merge pull request #3340 from stephentoub/debuggernotify_model
Ensure Debugger.NotifyOfCrossthreadDependency is exposed from mscorlib
Jan Kotas [Thu, 25 Feb 2016 03:06:04 +0000 (19:06 -0800)]
Merge pull request #3341 from jkotas/tokenkind-newobj
Add CORINFO_TOKENKIND_NewObj
Jan Kotas [Thu, 25 Feb 2016 02:15:11 +0000 (18:15 -0800)]
Merge pull request #3331 from brianrob/crossgen_perfmap
Update Crossgen /CreatePerfMap to clear NGENWORKER_FLAGS_READYTORUN
Jan Kotas [Thu, 25 Feb 2016 00:58:56 +0000 (16:58 -0800)]
Merge pull request #3348 from JohnChen0/master
Fix .gitignore so src/tools is not ignored
Pat Gavlin [Thu, 25 Feb 2016 00:29:55 +0000 (16:29 -0800)]
Merge pull request #3333 from pgavlin/JitAllocatorCleanup
Remove dead definitions in alloc.h.
Sivarv [Wed, 24 Feb 2016 23:56:25 +0000 (15:56 -0800)]
Merge pull request #3345 from sivarv/vector3fix
Fix to second CoreFX SIMD test failure mentioned in issue #2886.
John Chen (CLR) [Wed, 24 Feb 2016 22:31:38 +0000 (14:31 -0800)]
Fix .gitignore so src/tools is not ignored
Commit 3079b40 modified .gitignore to ignore top-level Tools directory.
However the change was too broad and caused src/tools to be ignored as well.
This commit fixes it.
Rahul Kumar [Wed, 24 Feb 2016 21:30:48 +0000 (13:30 -0800)]
Merge pull request #3343 from rahku/Helix
Helix testing : Add xunit.performance.core to core_root. required for few jit tests
sivarv [Wed, 24 Feb 2016 19:26:15 +0000 (11:26 -0800)]
Fix to second CoreFX SIMD test failure mentioned in issue #2886.
Root Cause:
The following managed method is executed by the test
Vector3 Vecor3.Normalize(Vector3)
Vector3 on Unix gets passed in two registers by caller of Normalize()
method. Within prolog of Normalie(), RyuJIt homes Vector3 arg by
writing only 12 bytes of arg regs xmm0/xmm1. As a result the upper
4-bytes could end up garbage. Further down test performs dot product
and the codegen of which makes the assumption that Vector3 types
when loaded in regs will have upper 4-bytes zero'ed out. Since
that assumption is violated, incorrect results gets computed and
hence CoreFx test failure not fiding expected result.
Fix: While homing Vector3 arg treat it as TYP_SIMD16 and home
16-bytes of arg regs xmm0/xmm1.
Also added a TODO comment on the assumptions RyuJIT is making on
Vector3 type. Will be opening a Git issue for the same.
Andy Ayers [Wed, 24 Feb 2016 18:41:21 +0000 (10:41 -0800)]
Merge pull request #3315 from AndyAyersMS/InlineRefactor5
InlineRefactoring: introducing InlinePolicy
Jan Kotas [Wed, 24 Feb 2016 17:23:01 +0000 (09:23 -0800)]
Merge pull request #3233 from vancem/RemoveSecurityCheck
Revert "Revert "Remove useless security checks""
Kyungwoo Lee [Wed, 24 Feb 2016 17:14:48 +0000 (09:14 -0800)]
Merge pull request #3316 from kyulee1/fixmustexpand
Fix for pMustExpand early fail.
dotnet-bot [Wed, 24 Feb 2016 16:51:21 +0000 (08:51 -0800)]
Fix for Bug 192726:[amd64][ilGen] gtVNPair_GetLiberal assert
The previously added asserts in fgMorph about valid value numbers are not always true. In the failing case, it turned out a GT_COMMA came out of fgMorph after value numbering did have a NoVN value number. The fix here is to simply remove the asserts since the code there does not really care about what value numbers the child expressions have.
[tfs-changeset: 1578470]
Jan Kotas [Wed, 24 Feb 2016 16:07:23 +0000 (08:07 -0800)]
Add CORINFO_TOKENKIND_NewObj
CoreRT implementation of the JIT-EE interface needs to know whether the token is coming from newobj instruction
stephentoub [Wed, 24 Feb 2016 16:03:50 +0000 (11:03 -0500)]
Ensure Debugger.NotifyOfCrossthreadDependency is exposed from mscorlib
Russ Keldorph [Wed, 24 Feb 2016 15:40:46 +0000 (07:40 -0800)]
Replace reference to msvcr110.dll with msvcrt.dll in test b286991
Removes unnecessary (and unsupported) dependency and makes the test
consistent with its peers.
Pallavi Taneja [Tue, 23 Feb 2016 23:54:09 +0000 (15:54 -0800)]
[Fix build break] In the previous change I exposed some APIs in mscorlib.dll to unblock System.Reflection change. However, crossgen.exe trips on this mscorlib.dll since some of the methods it uses is not part of the FEATURE_CORECLR. This change simply exposes these APIs for now.
[tfs-changeset: 1578199]