Age | Commit message (Collapse) | Author | Files | Lines |
|
Arm64 build fixes
|
|
Fix "call to 'sqrt' is ambiguous" error when building on SunOS
|
|
Building on SunOS will fail with the message of "call to 'sqrt' is
ambiguous" because it can't determine the type of the constant value
passed to it. This change adds a cast to a double because every usage
uses this data type
|
|
Fix a bug in morphing GT_COMMA with op1 being the new GetType() intri…
|
|
Remove dead code from JIT's earlyprop.cpp
|
|
Fix LNK4044 again
|
|
Disable LNK4221 again
|
|
Fixed issue where files in /opt/local/include are not found and LTTng is required on SmartOS
|
|
Enable ilasm round trip test for xPlatforms
|
|
The GetType intrinsic is side effect free and the morphing of GT_COMMA is
optimizing out side effect free op1a. The native implementation of GetType
- NativeObject::GetClass - can throw NullReferenceException and such
optimizing the call out is illegal in such case. Also added GTF_EXCEPT
to the intrinsic node to note the other optimizers that the call can
throw an exception.
The coreclr and corefx tests passed. Also minimal, correct asmdifs. The
daily tests passed as well.
|
|
|
|
This was fixed in #2310 and the fix was accidentally reverted in #2244 and #2425
|
|
Linker warning 4221 was disabled in #2320 and was accidentally reenabled in #2244
|
|
|
|
1. Enable CLRTest.Execute.Bash.targets similar to CLRTest.Execute.Batch.targets.
2. _IlasmSwitches uses '-' instead of '/'
3. Disable optimization on the prebuilt asmparse.c due to https://github.com/dotnet/coreclr/issues/2305
|
|
Modified to not require LTTng on SunOS
|
|
Priority 1 tests for OSX & Linux x64 Release builds should now run daily (fixed)
|
|
Enable Incremental build
|
|
|
|
Actually lock the project.lock.json
|
|
|
|
Temporarily disable these until we can sort out what goes wrong on some of the test machines.
|
|
Enable ildasm/ilasm round trip testing
|
|
Hopefully this clears up some test issues on ubuntu with the serialization benchmarks.
|
|
This enables ilasm round-trip testing for Windows.
In order to test il{d}asm on xPlatforms, I injected ilasm round trip before executing the test.
This means this step is a part of the runtest not the buildtest -- unlike runtest, currently buildtest is only on Windows.
1. Test batch file is extended to run ildasm followed by ilasm.
2. BuildTest has now an optional flag "ilasmroundtrip" to enable the round-trip tests. Note under this flag, we run both the original binary and the round-trip binary respectively.
3. Test timeout is increased since the round-trip is part of runtime.
4. Did some clean-up on CLRTest.Execute.Batch.targets
5. Bail out the round-trip tests that are built against mscorlib instead of .Net Core.
6. Added /raweh to ildasm due to https://github.com/dotnet/coreclr/issues/2481
7. ilasmroundtrip.targets (~10 tests) are added to enable them by default in a regular test run.
Tested with enabling the entire tests (>5K tests) with 'ilasmroundtrip' on, which should be on in a daily or weekly stress mode.
|
|
Merge changes from TFS
|
|
[tfs-changeset: 1560604]
|
|
Priority 1 tests should run nightly for OSX and Linux
|
|
|
|
|
|
Automating Event Logging Infrastructure
|
|
Simple serialization benchmarks
|
|
Using ilasm for test population
|
|
As with the simple serialization benchmarks, these tests are intended to stress aspects jit code quality and are not meant to compare various deserialization approaches.
|
|
Benchmarking a few serialization methods. Not really trying to compare them, just setting things up to see how well the JIT does at each.
|
|
Common Globals for TestWrapper
|
|
Fix assertion in ilasm
|
|
Individual test in test wrapper initializes and gets the same globals from environment variables.
This increases not only wrapper size but also runtime overhead when running multiple tests.
This refactors this common part out as _Globals so that each test just accesses it.
|
|
required for Event Logging
will be generated as part of build
|
|
This enables the use of ilasm from CoreCLR build to populate il tests instead of using a stock ilasm on Windows desktop.
Note currently we populate tests (even for xPlatforms) on Windows only.
But when we migrate this test population on xPlatforms, this can be simply used as well.
All permission related attributes are dropped which CoreCLR ignores anyhow. Or ilasm will fail to parse such attributes.
|
|
This is a simple fix for incorrect assertion.
Assertion location is changed and dwErrorIndex is zero-initialized.
ilasm on coreclr does not handle permission set/security attribute which coreclr does not support.
So, such attempt will end up with emitting error on coreclr instead of failing to assert.
|
|
|
|
Merge changes from TFS
|
|
[tfs-changeset: 1559946]
|
|
Merge changes from TFS
|
|
[tfs-changeset: 1559929]
|
|
[tfs-changeset: 1559927]
|
|
Unify casing of tests\src\JIT\jit64\eh\FinallyExec
|
|
|
|
GC OS interface refactoring
|