Comparing changes
Open a pull request
- 20 commits
- 61 files changed
- 20 contributors
Commits on Feb 19, 2025
-
-
Fix new EH hang on DebugBreak (#112640)
* Fix new EH hang on DebugBreak The new exception handling doesn't work well with the DebugBreak in some cases. E.g. when it is invoked from FATAL_GC_ERROR. The new EH attempts to handle the STATUS_BREAKPOINT stemming from the DebugBreak, allocate a managed exception object and hangs since it cannot do that when the GC is running. The cause is a missing check for the breakpoint exception in the ProcessCLRExceptionNew that is present in the old ProcessCLRException. To fix it, I've copied that code to the ProcessCLRExceptionNew. Close #112599 * Never process breakpoints via the new EH
-
-
JIT: Don't use
Compiler::compFloatingPointUsedto check if FP kills… -
Fix LINQ handling of iterator.Take(...).Last(...) (#112680)
When the Take amount is larger than the number of elements in the source `Iterator<T>`, Last ends up throwing an exception and LastOrDefault ends up returning the default value, rather than returning the last value in the taken region. As part of fixing this, I sured up the tests to try to cover more such sequences of operations. In doing so, the tests got a lot slower, so I tracked down and fixed places where we were doing a lot of unnecessary work.
-
[H/3] Fix test closing connection too fast (#112691)
* Fix test closing connection too fast * Same fix for similar test
-
JIT: array allocation fixes (#112676)
Fix some issues with array stack allocation that are currently hard to observe: * use correct method table for shared array types * pad array layout out to a multiple of TARGET_POINTER_SIZE It's currently not easy to observe the type of a stack allocated array as accessing the type causes escape. But this becomes possible when supporting span capture or once we have a version of the covariant store check that can handle stack allocated arrays. Non-padded layouts may mean elements just off the end of an odd-length short or byte array won't be zeroed, which some unsafe uses may expect. Added some test cases.
-
JIT: Always track the context for late devirt (#112396)
* Track InlinersContext in LateDevirtualizationInfo * Rework fix * Update inline context for statement as well * Use IL offset from call * Always set DebugInfo * Track InlineContext * Cleanup * Always record late devirt info * Always record late devirt info * Copy gtLateDevirtualizationInfo * Refactor to take InlineContext explicitly * Set inlinersContext to nullptr * Late devirt info always available * Oops * Remove the reduandant flag
-
-
-
Reduce allocations in XmlWriterApiTests.TCFullEndElement (#112688)
* Reduce allocations in XmlWriterApiTests * Apply code review improvements
-
JIT: Remove
LclVarDsc::GetArgRegandLclVarDsc::GetOtherArgRegus……es (#112652) Switch these to use new ABI info. The only uses left are in logging and in the old ABI classification happening during `lvaInitUserArgs`. Once we remove uses of all the information stored there we can get rid of the fields entirely.
-
arm64: Change EQ/NE node to SETCC if the operand supports the zero fl…
…ag (#112235) * arm64: Change EQ/NE node to SETCC if the operand supports the zero flag * Add one line tests for adds, ands, bics & subs * Add tests for negs
-
-
More clean up to debugger attribute tests (#112450)
I missed a few places with ``` [SetupLinkerKeepDebugMembers ("true")] ``` back when I landed #89011 -
-
Fix path to
dotnetinAndroidSampleAppproject (#112590)Non-Windows path was missing a `..`. This switches to using `DotNetTool`, which is set by the Arcade SDK (RepoLayout.props).
This comparison is taking too long to generate.
Unfortunately it looks like we can’t render this comparison for you right now. It might be too big, or there might be something weird with your repository.
You can try running this command locally to see the comparison on your machine:
git diff cc2019bc0b...457e3eb93e