Matt Ellis [Wed, 7 Oct 2015 06:47:20 +0000 (23:47 -0700)]
Merge pull request #1694 from janvorli/fix-exception-handling-issue
Fix issue in the new exception handling
Jan Kotas [Tue, 6 Oct 2015 22:40:08 +0000 (15:40 -0700)]
Merge pull request #1666 from jasonwilliams200OK/master
cmake: cleanup FreeBSD system include path
Jan Kotas [Tue, 6 Oct 2015 22:39:44 +0000 (15:39 -0700)]
Merge pull request #1687 from jasonwilliams200OK/__fastcall-Wmacro-redefined
code: Fix redefined macro warning
Jan Vorlicek [Tue, 6 Oct 2015 21:46:49 +0000 (23:46 +0200)]
Fix issue in the new exception handling
There was a problem with scanned stack range stored in an exception tracker.
After unwinding a sequence of native frames during the 2nd pass of unwinding,
the scanned stack range was not cleared and referenced stack frames that
were unwound. In some rare cases related to throwing an exception from
a catch handler, this was causing the exception handling to consider stack
frames that accidentally had the same range of addresses as already visited,
which resulted in a wrong exception handler being executed.
Peter Jas [Tue, 6 Oct 2015 18:13:59 +0000 (18:13 +0000)]
code: Fix redefined macro warning.
For example, in this log:
https://dotnet-ci.cloudapp.net/job/dotnet_coreclr_freebsd_debug_prtest/1082/consoleFull
search for '-Wmacro-redefined'. All 26 occurences are pointing to this line.
Sergiy Kuryata [Tue, 6 Oct 2015 19:38:49 +0000 (12:38 -0700)]
Merge pull request #1675 from wtgodbe/clrBugFixes
Fixed small bugs in decodemd.cpp, gcrecord.h, gc.cpp, Strike/util.cpp…
William Godbe [Tue, 29 Sep 2015 21:06:58 +0000 (14:06 -0700)]
Fixed small bugs in decodemd.cpp, gcrecord.h, gc.cpp, Strike/util.cpp, and strike.cpp, as outlined in this article: viva64.com/en/b/0310/#ID0E5BJM
Matt Ellis [Mon, 5 Oct 2015 22:39:52 +0000 (15:39 -0700)]
Merge pull request #1662 from steveharter/master
fix ICU compile error on CentOS
Jan Vorlicek [Mon, 5 Oct 2015 21:21:54 +0000 (23:21 +0200)]
Merge pull request #1677 from janvorli/refactor-exception-handling
Implement refactored Unix managed exception handling
Koundinya Veluri [Mon, 5 Oct 2015 18:36:55 +0000 (11:36 -0700)]
Merge pull request #1681 from kouvel/FixPathSeparators
Fix loading native libraries from native search paths provided by the…
Koundinya Veluri [Sat, 3 Oct 2015 18:45:17 +0000 (11:45 -0700)]
Fix loading native libraries from native search paths provided by the host
Fixed a couple of issues outside Windows regarding the directory separator char, path separator char, and absolute path determination.
Fixes #1680
Steve Harter [Thu, 1 Oct 2015 22:18:00 +0000 (17:18 -0500)]
Fix ICU compile error on CentOS by using DateFormatSymbols::NARROW insead of SHORT as SHORT is not defined in the version 50 of ICU being used by Centos. SHORT was added in ICU 51.
Jan Vorlicek [Mon, 5 Oct 2015 15:49:28 +0000 (17:49 +0200)]
Few fixes
This change fixes few issues that I've found:
- The frame popping in the Frame destructor is still needed, so I've put it back.
- There was a problem (I've found that it was there before this change) with the "ex"
variable passed to the StartUnwindingNativeFrames. The local was in frames that were
obsoleted by the StartUnwindingNativeFrames before the ThrowExceptionHelper was called.
And when runtime called malloc to allocate its internal exception data, the stack
of the malloc has overwritten part of the PAL_SEHException. It was not visible before
since that part was in the end of the ContextRecord. But now that I've put the
TargetFrameSp to the end of the PAL_SEHException, it was sometimes getting overwritten.
The fix is to create a copy of the "ex" out of stack.
Jan Vorlicek [Fri, 2 Oct 2015 12:55:07 +0000 (14:55 +0200)]
Implement refactored Unix managed exception handling
This change refactors managed exception handling so that we perform full first
pass over both managed and native frames and only after we find target frame,
we perform the second pass.
This helps in debugging of unhandled managed exceptions since the debugger can
kick in at the end of the first pass when the stack frames are still intact and
it can show the context where the exception was fired.
Matt Mitchell [Fri, 2 Oct 2015 23:17:49 +0000 (16:17 -0700)]
Merge pull request #1672 from kouvel/IncludeTestsInBuildSh
On builds outside Windows, build native test components by default
Koundinya Veluri [Fri, 2 Oct 2015 18:05:54 +0000 (11:05 -0700)]
On builds outside Windows, build native test components by default
Mike McLaughlin [Fri, 2 Oct 2015 17:23:17 +0000 (10:23 -0700)]
Merge pull request #1663 from mikem8361/excepthan
Add native exception filter holder.
Peter Jas [Fri, 2 Oct 2015 10:02:28 +0000 (10:02 +0000)]
cmake: cleanup FreeBSD system include path.
Include once and DRY.
Similar change made at dotnet/corefx#3487.
Jan Vorlicek [Fri, 2 Oct 2015 08:37:53 +0000 (10:37 +0200)]
Merge pull request #1665 from ursine/patch-1
Fix comment variable name
Gary Coulbourne [Fri, 2 Oct 2015 02:04:44 +0000 (22:04 -0400)]
Fix comment variable name
Mike McLaughlin [Sun, 27 Sep 2015 23:11:05 +0000 (16:11 -0700)]
Add native exception filter holder.
The problem is that the debugger unhandled managed exception notification is sent
because the proper filter function (InternalUnhandledExcpetionFilter) isn't called
and the filter can't be called currently during the first pass of managed exception
dispatch that it requires.
The NativeExceptionHolder is used to hold the filter handler for the
PAL_TRY/PAL_EXCEPT/PAL_EXCEPT_FILTER macros so managed exception dispatcher
can call them during the first pass like real SEH on Windows.
Jan Kotas [Thu, 1 Oct 2015 21:47:55 +0000 (14:47 -0700)]
Merge pull request #1661 from janvorli/fix-crossgen
Fix wrong register with module address in crossgen on Unix
Jan Vorlicek [Thu, 1 Oct 2015 20:13:09 +0000 (22:13 +0200)]
Fix wrong register with module address in crossgen on Unix
This change fixes a crossgen bug when the module pointer as a parameter to
JIT_StrCns was incorrectly placed into RDX both on Windows and on Unix.
Since it is the second parameter to the function, the calling convention
on Unix requires it to be in RSI.
Jan Kotas [Thu, 1 Oct 2015 15:47:09 +0000 (08:47 -0700)]
Merge pull request #1620 from justinvp/dictionary
Avoid the enumerator allocation in Dictionary's constructor
Jan Kotas [Thu, 1 Oct 2015 15:41:50 +0000 (08:41 -0700)]
Merge pull request #1650 from Dmitry-Me/fixLeftShiftIssue1526
Fix left shift of signed integer - issue 1526
Jan Kotas [Thu, 1 Oct 2015 15:36:12 +0000 (08:36 -0700)]
Merge pull request #1658 from dotnet-bot/from-tfs
Merge changes from TFS
Jan Vorlicek [Thu, 1 Oct 2015 08:58:00 +0000 (10:58 +0200)]
Merge pull request #1654 from janvorli/fix-isipinepilog-assert
Fix assert in IsIPInEpilog
Koundinya Veluri [Thu, 1 Oct 2015 06:50:41 +0000 (23:50 -0700)]
Merge pull request #1638 from kouvel/CoreClrTests
Improve CoreCLR test runner outside Windows
John Chen [Thu, 1 Oct 2015 04:08:57 +0000 (21:08 -0700)]
Fix SigPointer::PeekValueTypeTokenClosed to handle generic instantiation represented by ELEMENT_TYPE_INTERNAL
Integrate change #1529861 from NetFXDev1
[tfs-changeset: 1532127]
Jan Kotas [Thu, 1 Oct 2015 04:01:01 +0000 (21:01 -0700)]
Merge pull request #1657 from Djuffin/disable_lu_ptrace
Remove references to libunwind-ptrace to fix build on RPM distros
Jan Kotas [Thu, 1 Oct 2015 03:59:48 +0000 (20:59 -0700)]
Merge pull request #1656 from janvorli/exception-handling-perf
Improve Unix exception handling performance
Eugene [Thu, 1 Oct 2015 01:09:12 +0000 (18:09 -0700)]
Remove references to libunwind-ptrace to fix build on RPM distros
Jan Vorlicek [Thu, 1 Oct 2015 00:55:30 +0000 (02:55 +0200)]
Improve Unix exception handling performance
This change adds caching of thread base / limit in PAL. Profiling has shown
extraction of the base / limit using pthread APIs to be the main contributors
to the bad performance.
With this fix, a simple loop with throw / catch is about 40 times faster than
before.
Jan Vorlicek [Wed, 30 Sep 2015 22:10:41 +0000 (00:10 +0200)]
Fix assert in IsIPInEpilog
The assert was incorrect for managed functions without body. Since there
is no easy way to find out that function has no body at that place,
the fix is to remove the assert.
Lakshmi Priya [Wed, 30 Sep 2015 21:57:41 +0000 (14:57 -0700)]
Merge pull request #1524 from Priya91/stackstring
Add StackString class for stack/heap allocation for MAX_LONGPATH.
Koundinya Veluri [Wed, 30 Sep 2015 21:48:32 +0000 (14:48 -0700)]
Merge pull request #1651 from kouvel/TpaPktFix
Remove the need for the PKT to match when binding to an assembly on t…
Koundinya Veluri [Wed, 30 Sep 2015 20:22:13 +0000 (13:22 -0700)]
Merge pull request #1646 from kouvel/HeapAllocationFix
Heap allocation fix
Koundinya Veluri [Wed, 30 Sep 2015 19:20:52 +0000 (12:20 -0700)]
Remove the need for the PKT to match when binding to an assembly on the TPA list
There is some inconsistency in how binding works with and without a custom AssemblyLoadContext being active. By default, the binder allows binding to assemblies on the TPA list despite a PKT mismatch while searching app paths. This part is skipped when a custom AssemblyLoadContext is active.
It has been determined that generally, the binder shouldn't require the PKT to match, so disabling that check.
Fixes #1640
Lakshmi Priya Sekar [Thu, 10 Sep 2015 06:18:32 +0000 (23:18 -0700)]
Add StackString class for stack/heap allocation for MAX_LONGPATH.
Add stackstring.cpp to build.
Dmitry-Me [Wed, 30 Sep 2015 09:16:30 +0000 (12:16 +0300)]
Fix left shift of signed integer - issue 1526
Justin Van Patten [Thu, 24 Sep 2015 19:38:32 +0000 (12:38 -0700)]
Dictionary constructor perf optimization
Regarding the constructor overloads that take an
IDictionary<TKey, TValue>: It is likely that the passed-in dictionary
is Dictionary<TKey, TValue>. When this is the case, avoid the
enumerator allocation and overhead by looping through the entries
array directly. This is only done when the dictionary is
Dictionary<TKey, TValue> and not a subclass, to maintain back-compat
with subclasses that may have overridden the enumerator behavior.
Koundinya Veluri [Fri, 18 Sep 2015 18:08:28 +0000 (11:08 -0700)]
Improve CoreCLR test runner outside Windows
CoreCLR tests will need to be built on Windows and copied over to a non-Windows machine. Then, coreclr/tests/runtest.sh along
with some switches can be used to run the tests.
Changes:
- Run tests in parallel by default (use --sequential to force sequential mode)
- Update Windows test build to produce the Windows core overlay that includes test dependencies that are not built
- Add a number of switches (see coreclr/tests/runtest.sh for details) to provide locations of:
- The Windows test build
- The CoreCLR native test build
- Either:
- This:
- The core layout (including all of the below)
- Or:
- The CoreCLR build
- mscorlib.dll
- The CoreFX build
- The CoreFX native build
- Using information from above, build or use the core layout, and copy native test binaries to the respective test directories
- Add --testDir to specify an inclusive list of test directories to run
- Add text files to exclude tests:
- coreclr/tests/testsUnsupportedOutsideWindows.txt - Tests that are not supported outside Windows
- coreclr/tests/testsFailingOutsideWindows.txt - Tests that are known to fail outside Windows and are temporarily disabled while they are investigated
- Add --runFailingTestsOnly to run only the tests listed in coreclr/tests/testsFailingOutsideWindows.txt to verify known failures
- Produce xUnit-style output for the CI, output goes into <Windows test build folder>/coreclrtests.xml
Koundinya Veluri [Tue, 29 Sep 2015 21:01:57 +0000 (14:01 -0700)]
Fix HeapReAlloc to return a non-null pointer for a requested size of zero bytes
Jan Kotas [Tue, 29 Sep 2015 06:58:33 +0000 (23:58 -0700)]
Merge pull request #1632 from bbowyersmyth/StringStartsWith
String.StartsWith ordinal optimization
Lakshmi Priya [Mon, 28 Sep 2015 22:56:03 +0000 (15:56 -0700)]
Merge pull request #1589 from Priya91/pathstring
Use SString type - PathString for path allocations in binder.
Koundinya Veluri [Mon, 28 Sep 2015 21:54:46 +0000 (14:54 -0700)]
Merge pull request #1628 from kouvel/CppCheckFixes
Fix a few issues reported by cppcheck
Bruce Bowyer-Smyth [Sun, 27 Sep 2015 20:47:43 +0000 (06:47 +1000)]
String.StartsWith ordinal optimization
Eric Eilebrecht [Sat, 26 Sep 2015 17:54:11 +0000 (10:54 -0700)]
Merge pull request #1629 from ericeil/AsyncLocalPerf
Reduce allocations in AsyncLocal/ExecutionContext
1: Allocate a new AsyncLocal change notification list only if we need to add an item to the list.
2: Pre-size any collections we allocate, and manually copy into them to avoid IEnumerator allocations, etc.
3: Store the change notification list in an array, rather than a List, to avoid the extra List allocation.
Eric Eilebrecht [Sat, 26 Sep 2015 17:50:10 +0000 (10:50 -0700)]
Merge branch 'AsyncLocalPerf' of https://github.com/ericeil/coreclr into AsyncLocalPerf
Eric Eilebrecht [Fri, 25 Sep 2015 21:03:08 +0000 (14:03 -0700)]
Reduce allocations when setting AsyncLocal values
1: Allocate a new AsyncLocal change notification list only if we need to add an item to the list.
2: Pre-size any collections we allocate, and manually copy into them to avoid IEnumerator allocations, etc.
3: Store the change notification list in an array, rather than a List, to avoid the extra List allocation.
Jan Kotas [Sat, 26 Sep 2015 13:42:29 +0000 (06:42 -0700)]
Merge pull request #1617 from jamesqo/patch-1
Assign directly to this in Decimal constructor
Matt Ellis [Sat, 26 Sep 2015 01:15:07 +0000 (18:15 -0700)]
Merge pull request #1630 from dotnet-bot/from-tfs
Merge changes from TFS
Eric Eilebrecht [Sat, 26 Sep 2015 00:30:53 +0000 (17:30 -0700)]
Store AsyncLocal notification list in an array, rather than a List.
This should further reduce allocations, as we don't need the intermediate List<IAsyncLocal> object.
Matt Ellis [Sat, 26 Sep 2015 00:18:34 +0000 (17:18 -0700)]
Fix clang only build breaks
Some new code from TFS was triggering this error in clang:
`error: suggest braces around initialization of subobject`
I just made the obvious fix.
Rahul Kumar [Fri, 25 Sep 2015 21:03:54 +0000 (14:03 -0700)]
Fix for 134453: fix prefast warnings
[tfs-changeset: 1529946]
Gaurav Khanna [Fri, 25 Sep 2015 19:31:07 +0000 (12:31 -0700)]
VS has mentioned that when they have an AV come through their reflection invoked code, they endup getting TargetInvocationException that is not actionable and also does not help in the Watson bucketing for the issue. To address this, they would like an opt-in flag to force a failfast when a CSE remains unhandled within reflection invocation.
This change adds support for an opt-in flag that will trigger failfast in Reflection codepath inface of an unhandled CSE.
[tfs-changeset: 1529878]
Koundinya Veluri [Sat, 26 Sep 2015 00:23:43 +0000 (17:23 -0700)]
Add AssemblyLoadContext.LoadUnmanagedDllFromPathfor use by overriders of AssemblyLoadContext.LoadUnmanagedDll
LoadUnmanagedDllFromPath needs to call LoadLibrary or dlopen, and return the system handle to the library through LoadUnmanagedDll. Outside Windows, when LoadUnmanagedDll returns a system handle to a library, the handle needs to be registered with PAL's module list for lifetime management. From that point on, the system handle is tracked as part of the PAL handle.
To handle both of the above, I have refactored module.cpp!LOADLoadLibrary into a LoadLibraryDirect portion and a RegisterLibrary component.
LoadLibraryDirect loads the specified library directly using the system call, without appending or prepending anything to the library name
RegisterLibrary registers a system library handle with PAL to get a PAL handle
This patch contains the necessary changes to coreclr and mscorlib. Tests will be added separately, after the new APIs are published and can be consumed.
Fixes dotnet/coreclr#935
Part of dotnet/coreclr#937 and dotnet/corefx#3054
See https://github.com/dotnet/coreclr/pull/1500
[tfs-changeset: 1529692]
Eric Eilebrecht [Sat, 26 Sep 2015 00:02:52 +0000 (17:02 -0700)]
Fix typo in my previous typo fix.
Eric Eilebrecht [Sat, 26 Sep 2015 00:01:40 +0000 (17:01 -0700)]
Fix typo
Eric Eilebrecht [Fri, 25 Sep 2015 23:45:16 +0000 (16:45 -0700)]
Reduce AsyncLocal setter allocations further
Now we pre-size any collections we allocate, and manually copy into them to avoid IEnumerator allocations, etc.
Matt Ellis [Fri, 25 Sep 2015 23:23:24 +0000 (16:23 -0700)]
Merge pull request #1627 from ellismg/icu-collation
Implement basic collation on top of ICU
Matt Ellis [Fri, 25 Sep 2015 23:03:36 +0000 (16:03 -0700)]
Merge pull request #1621 from dotnet-bot/from-tfs
Merge changes from TFS
Matt Ellis [Wed, 15 Jul 2015 18:22:39 +0000 (11:22 -0700)]
Implement basic collation on top of ICU
This change adds support for basic Unicode collation support, built on
top of ICU. Windows and ICU have different collation models, and the
windows model does not correspond 1:1 with the ICU model, so in addition
to differences in sort weights between the two platforms, the
CompareOptions enum does not map nicely to ICU options.
For now, we only map CompareOptions.None, CompareOptions.IgnoreCase as
well as CompareOptions.Ordinal and CompareOptions.OrdinalIngoreCase,
other CompareOptions are ignored during collation.
In addition to collation support, I have enabed the randomized string
hashing code (using Marvin32 + a per app domain seed) so that string
hashcodes are not predictable across runs.
Eric Eilebrecht [Fri, 25 Sep 2015 21:03:08 +0000 (14:03 -0700)]
Allocate a new AsyncLocal change notification list only if we need to add an item to the list.
Mike McLaughlin [Fri, 25 Sep 2015 21:00:05 +0000 (14:00 -0700)]
Merge pull request #1626 from caslan/exportsforee
Add new exports to get expression evaluator working in clrdbg
Lakshmi Priya Sekar [Tue, 22 Sep 2015 06:32:54 +0000 (23:32 -0700)]
Use PathString type to allocate large path strings on linux scenario.
Respond to PR feedback. Fix test failures in mscoree.
Koundinya Veluri [Wed, 23 Sep 2015 19:19:33 +0000 (12:19 -0700)]
Fix a few issues reported by cppcheck
Cagri [Fri, 25 Sep 2015 17:35:57 +0000 (10:35 -0700)]
Add new exports to get expression evaluator working in clrdbg
Rama Krishnan Raghupathy [Thu, 24 Sep 2015 23:48:04 +0000 (16:48 -0700)]
Fixing SAL anotation for VM
[tfs-changeset: 1529342]
Jan Kotas [Thu, 24 Sep 2015 18:08:51 +0000 (11:08 -0700)]
Merge pull request #1612 from jkotas/readytorun-directcall
Handle direct call ReadyToRun helpers
Jan Kotas [Thu, 24 Sep 2015 16:37:53 +0000 (09:37 -0700)]
Merge pull request #1616 from dotnet-bot/from-tfs
Merge changes from TFS
James Ko [Thu, 24 Sep 2015 16:05:21 +0000 (12:05 -0400)]
Assign directly to this in Decimal.cs
dotnet-bot [Thu, 24 Sep 2015 15:04:08 +0000 (08:04 -0700)]
Porting EventCounter from Roxel Branch to ProjectK
[tfs-changeset: 1529077]
Jan Kotas [Thu, 24 Sep 2015 05:26:08 +0000 (22:26 -0700)]
Merge pull request #1613 from jkotas/crossgen-commandline
Fix crossgen command line handling on Unix
Jan Kotas [Thu, 24 Sep 2015 05:07:45 +0000 (22:07 -0700)]
Delete incorrect AMD64 ifdef
ReadyToRun case of getCallInstruction was incorrectly ifdefed out
for AMD64. Delete the incorrect ifdef and make other adjustment
to make the code compile for X86 as well.
Jan Kotas [Thu, 24 Sep 2015 04:30:35 +0000 (21:30 -0700)]
Fix crossgen command line handling on Unix
Replace hardcoded Windows directory and path separators with symbolic
constants.
Jan Kotas [Thu, 24 Sep 2015 01:34:37 +0000 (18:34 -0700)]
Handle direct call ReadyToRun helpers
Add missing check in the JIT to generate correct code for direct call ReadyToRun helpers. I have run into this bug while experimenting with ReadyToRun optimizations. It is not observable bug today because of crossgen happens to always emit ReadyToRun helpers as indirect calls.
Matt Ellis [Wed, 23 Sep 2015 22:41:59 +0000 (15:41 -0700)]
Merge pull request #1609 from Dmitry-Me/duplicatedCheck
Fix erroneous check
Jan Vorlicek [Wed, 23 Sep 2015 22:18:14 +0000 (00:18 +0200)]
Merge pull request #1610 from janvorli/osx-thread-suspension-activation-2
Implement runtime suspension for OSX
Jan Vorlicek [Fri, 18 Sep 2015 13:57:49 +0000 (15:57 +0200)]
Implement runtime suspension for OSX
This change implements runtime suspension for OSX using activation
injection, the same mechanism as Linux uses.
I have modified the activation injection mechanism by adding a new
function that CoreCLR registers with PAL and PAL calls it to check
if an instruction address is safe for injection. The injection is
performed only if this new function returns true.
The activation on OSX is performed by suspending the target thread,
checking the instruction address in the suspended thread's context
using the above mentioned function and modifying the context to
perform the injection only at injection safe place.
Dmitry-Me [Wed, 23 Sep 2015 15:32:54 +0000 (18:32 +0300)]
Fix erroneous check
Matt Ellis [Wed, 23 Sep 2015 07:39:33 +0000 (00:39 -0700)]
Merge pull request #1608 from ellismg/fix-infinite-recursion-in-culture-init
Use Ordinal when searching for collation keyword
Matt Ellis [Wed, 23 Sep 2015 06:47:59 +0000 (23:47 -0700)]
Use Ordinal when searching for collation keyword
IndexOf by default is culture specific, which could lead to us doing
the wrong thing depending on the current culture and also can cause
problems where we need to access CultureData while we are still building
up culture information, causing infinite recursion.
Some collection tests were triggering the latter case and failing after
we merged in the ICU changes.
Matt Ellis [Wed, 23 Sep 2015 01:24:23 +0000 (18:24 -0700)]
Merge pull request #1604 from ellismg/icu-merge-staging
Merge in initial round of ICU Changes
Jan Kotas [Tue, 22 Sep 2015 22:15:48 +0000 (15:15 -0700)]
Merge pull request #1603 from dotnet-bot/from-tfs
Merge changes from TFS
Matt Ellis [Tue, 22 Sep 2015 21:47:20 +0000 (14:47 -0700)]
Respond to PR feedback
Matt Ellis [Tue, 22 Sep 2015 19:08:34 +0000 (12:08 -0700)]
Update documentation to list ICU dependency
Provide some hints on how to obtain the required ICU components for
Linux, OSX and FreeBSD
Steve Harter [Fri, 18 Sep 2015 19:01:16 +0000 (14:01 -0500)]
complete core implementation of CultureData for Unix
Eric Erhardt [Thu, 17 Sep 2015 15:50:37 +0000 (10:50 -0500)]
Use 'readlink' of the /etc/localtime file to determine the TimeZoneInfo.Local.Id on Linux.
Fix https://github.com/dotnet/corefx/issues/2489.
Eric Erhardt [Fri, 11 Sep 2015 21:02:46 +0000 (16:02 -0500)]
Fixing date patterns on Linux.
1. Getting more LongDatePatterns by getting both full and long ICU patterns.
2. Getting more ShortDatePatterns by getting both medium and short ICU patterns. Still keeping the "yMd" pattern, since this closely matches what is used on Windows.
3. Removing any duplicates in the date patterns.
4. "Normalizing" the date patterns from ICU format to .NET format.
a. "EEEE", "eeee" and "cccc" is replaced with "dddd"
b. "LLLL" is replaced with "MMMM"
c. "G" is replaced with "g"
d. A single 'y' is replaced with 'yyyy'
Matt Ellis [Tue, 25 Aug 2015 00:44:16 +0000 (17:44 -0700)]
Build System.Native.Globalization on OSX
This requires the 'icu4c' package from homebrew, which can be installed
with `brew install icu4c`.
Eric Erhardt [Wed, 9 Sep 2015 21:29:13 +0000 (16:29 -0500)]
Adding an empty string to the end of month names if ICU only returns 12.
This bug was caught by a test case in System.Globalization.Tests.
Matt Ellis [Wed, 2 Sep 2015 01:40:04 +0000 (18:40 -0700)]
Don't use StringBuilderCache for casing
During code review, Steve proposed we moved to using StringBuilderCache
to the allocation of the temporary char[] array during casing
operations. I made the change but later realized that this caused a few
issues.
- The native layer operates on UChar/length pairs, not null terminated
strings. Because of this, we don't actually write a terminating null
into the destination buffer (since we just do a 1:1 casing operation
on the input and never see the null terminator). However, the
marshalling code for StringBuilder assumes the buffer will be null
terminated and uses that to compute the new length value after a
marshalling call. Because there can be left over data from previous
calls in the buffer the string length calculation will be incorrect
and we will end up leaking in extra data into the cased string.
- The StringBuilder marshalling as a whole won't work if we have
embdedded nulls in a string (we'll end up dropping characters on the
transition from native back to managed) but that's something that we
need to be able to handle.
Ideally there would be a way to construct a mutable string, case into
its buffer and the freeze the string, but we don't have a way to do that
today, so we need to stick with this allocation.
Eric Erhardt [Tue, 8 Sep 2015 18:34:11 +0000 (13:34 -0500)]
Responding to PR feedback.
Eric Erhardt [Wed, 2 Sep 2015 23:18:20 +0000 (18:18 -0500)]
Implement Japanese Era information.
Eric Erhardt [Tue, 1 Sep 2015 16:49:42 +0000 (11:49 -0500)]
Implement CalendarData.GetCalendars.
Eric Erhardt [Tue, 25 Aug 2015 21:55:30 +0000 (16:55 -0500)]
Implement CalendarData on Linux.
This is the first round of CalendarData implementation on Linux using ICU
for the information. It contains Month Names, Day Names, DateTime format
patterns, etc.
Steve Harter [Wed, 2 Sep 2015 17:42:36 +0000 (12:42 -0500)]
address code review feedback
Steve Harter [Wed, 2 Sep 2015 17:31:03 +0000 (12:31 -0500)]
address code review feedback
Steve Harter [Thu, 27 Aug 2015 17:31:25 +0000 (12:31 -0500)]
add number formatting