- 09 8月, 2022 19 次提交
-
-
由 Peter Sollich 提交于
This introduces a lookup table for regions where we can find the current generation, the planned generation and a couple flags efficiently, i.e. by simple indexing into table. The table has byte-sized elements where the low 2 bits represent the current generation and the high 2 bits represent the planned generation. The 4 bits in between are available for flags, 2 are used so far, one to show that a region was swept in the plan phase, and one to show that a region was demoted. The table is used in the mark phase, in the mark_through_cards_helper (which is also called in the relocate_phase), and in the write barriers (for now only the most frequently used ones, Array.Copy has its own way of setting cards that I haven't fixed). I have changed the write barrier to only set single bits for the case where a pointer to younger generation is stored into an object in an older generation. This costs an interlocked operation in the case the bit is not already set. In the test cases we looked at this is more than compensated by lower cost in card marking. For cases where the cost of the interlocked instruction is too high, an escape hatch has been provided the form of the COMPLUS_GCWriteBarrier setting which can be used to select another flavor of write barrier - either a more precise write barrier for regions, or an imprecise write barrier. Both set whole bytes and thus need no interlocked operations. The write barrier uses a global ephemeral range to decide quickly whether a stored pointer can be in an ephemeral generation at all. For now, this global ephemeral range is simply set to the address range of the whole heap, to avoid having to update the write barrier on the fly. A slightly more sophisticated approach that takes into account low gen 2 regions is present in the code, but disabled for now.
-
由 Viktor Hofer 提交于
* Avoid additional Pipes invocation and clean-up project * Update System.IO.Pipes.AccessControl.csproj
-
由 Michal Strehovský 提交于
* Replace MissingInteropDataException with NotSupportedException * Replace MissingRuntimeArtifactException with MissingMetadataException (it was already used as a MME in some spots anyway). I'll try to get rid of MME later. * Change the messages to refer to static analysis as the fix, not RD.XML.
-
由 Ilona Tomkowicz 提交于
* Fix of VHO for most test cases. * Cleanup. * New testcases with auto-props. * Adjusted ProtectionLvlTest to new changes.
-
由 Michal Strehovský 提交于
We no longer root all assemblies by default and we don't need to advertise the size switches to much. Also, we need to establish what switches we actually want to support (document on docs.microsoft.com) for real.
-
由 Hyungju Lee 提交于
-
由 Aditya Mandaleeka 提交于
-
由 David Wrighton 提交于
* Refactor input bubble checks to work in non-composite scenarios - Input bubble checks for making sure that the input bubble is valid are not currently processed for non-composite scenarios - Refactor them so that they are not n^2 in cost, and are handled for both composite and non-composite scenarios
-
由 Kevin Jones 提交于
-
由 Tamás Turnyánszki 提交于
Use ignoreCase=true when parsing enum values. Fix #71185
-
由 Ankit Jain 提交于
* xplat-setup.yml: Add support for setting `shouldContinueOnError` to .. `false`. Currently, xplat-setup.yml passes on `shouldContinueOnError` to the template with `true` if the parameter was `true`, else it sets the value based on whether this is for runtime-staging, on a PR. Which essentially means that runtime-staging defaults to `shouldContinueOnError: true` for PRs, or we can set it explcitly to `true` which would affect the rolling builds. But this does not allow making a runtime-staging job *fail* when *tests fail*. To support that this change adds a check for a special value `forceFalse`, which causes the param to explicitly set to `false`. Thus allowing `runtime-staging` builds to be marked as such. * [wasm] Add more paths to trigger running tests on CI * fix comment
-
由 Steve Pfister 提交于
* [iOS] Enable skipped System.Threading tests Fixes https://github.com/dotnet/runtime/issues/51400 * Redisable 60583 and remove added newline Co-authored-by: NSteve Pfister <steve.pfister@microsoft.com> Co-authored-by: Nmdh1418 <mitchhwang1418@gmail.com>
-
由 Thays Grazia 提交于
-
由 Pavel Savara 提交于
- add flat APIs to createDotnetRuntime callback parameter - assign IMPORTS instance (#73565)
-
由 Aleksey Kliger (λgeek) 提交于
* [wasm-ep] Use DOTNET_DiagnosticPorts to configure Diagnostic Server Parse it the same way that the C code does: ``` <uri>[,<connect|listen>][,<suspend|nosuspend>] ``` - uri should be a websocket uri - listen is not supported as it doesn't make sense with a WebSocket - connect is the default if omitted - suspend is the default if omitted --- Additionally, move `mono_wasm_diagnostics_init` to later in the startup flow. This gives Blazor a chance to set DOTNET_DiagnosticPorts from their `onRuntimeInitialized` callback. Fixes https://github.com/dotnet/runtime/issues/73011 * Initialize diagnostic server in different places for Blazor and non-Blazor It has to be after environment variables are set, but before mono_wasm_load_runtime is called. There is no good place that's common to both startup paths. Try it on both. Use a flag to make diagnostics initialization run at most once * update browser-eventpipe sample to use DOTNET_DiagnosticPorts * remove unused imports
-
由 David Wrighton 提交于
Fix issue where calculation of valuetypeshapecharacteristic on Arm64 does not match native runtime (#73406) - We didn't check to make sure that the type of the Vector<T> was a primitive numeric as the runtime does - Re-enable test Fixes #60036
-
由 Buyaa Namnan 提交于
* Compare type instead of module, check MetadataUpdater.IsSupported in HashCode * Perf tweaks for Equals and GetHashCode - No need to use GetHashCodeOfPtr when we are wrapping the hashcode using HashCode.Combine that takes care of the randomization - Use underlying type handle for type hashcodes. It is faster than the default object hashcode. - Delete special casing for MetadataUpdater.IsSupported from RuntimeMethodInfo.GetHashCode. It is actually a de-optimization for RuntimeMethodInfo with the two changes above, and about neutral for the rest. - Avoid virtual call for ReflectedType comparison - Restore comment for RuntimeMethodInfo Co-authored-by: NJan Kotas <jkotas@microsoft.com>
-
由 Juan Hoyos 提交于
-
- 08 8月, 2022 16 次提交
-
-
由 Zoltan Varga 提交于
* [mono][jit] Transition the x86 backend to use SSE for fp arithmetic. * Add SSE2 and FCMOV to the cpu requirements for mono on x86. * Also force the usage of r4fp on x86, the same as on arm. * Most of the code is copied from amd64-codegen.h and mini-amd64.c. * Reenable some tests. * Fix build failures. * Remove r4fp conditionals. * Add missing RCONV_TO_I opcode. * Fix OP_MOVE_F_TO_I4 and OP_MOVE_I4_TO_F. * Remove fpstack support code. * Fix warnings. * Add back MONO_ARCH_FLOAT32_SUPPORTED on x86. * Fix dreg type for r4_conv_to_i1 etc. opcodes.
-
由 Vitek Karas 提交于
Syncs latest linker changes to NativeAOT * Improvements in the compiler generated code space * Fixes a source build problem * Sync the tests which are already enabled in NativeAOT
-
由 Radek Zikmund 提交于
* Improve idle timeout tests * Minor change * Code review feedback
-
由 Viktor Hofer 提交于
Fixes https://github.com/dotnet/runtime/issues/73463. Tested locally just now.
-
由 Stephen Toub 提交于
* Fix nullable annotations on IEquatable/IComparable constraints * Fix LastIndexOfAny null dereference bug If the first two values are non-null and don't match the input and the third value is null, it gets dereferenced and results in a null reference exception.
-
由 Krzysztof Wicher 提交于
* Fix #41061: Add null checks in various XML methods * Remove test checking if we throw InvalidOperationException for null argument
-
由 Eirik Tsarpalis 提交于
-
由 Radek Zikmund 提交于
-
由 Adeel Mujahid 提交于
* Delete some old Jenkins scripts * Delete arm64_post_build.py
-
由 Vlad Brezae 提交于
If we memset a vt containing references, then, in order for preemptive GC and concurrent GC to work, writing to reference slots must happen atomically, otherwise GC might encounter a garbage ref. Also speeds up memset by about 15% for smaller valuetypes hitting this path on 64bit. More for extra large valuetypes.
-
由 Egor Bogatov 提交于
Co-authored-by: NStephen Toub <stoub@microsoft.com>
-
由 Michal Strehovský 提交于
On ARM64 Linux. See if #73216 helped with the hang.
-
由 Aaron Robinson 提交于
* Undefined behavior with null
-
由 Egor Bogatov 提交于
Co-authored-by: NJakob Botsch Nielsen <Jakob.botsch.nielsen@gmail.com>
-
由 Josie Bigler 提交于
* Return null DeclaringType for GlobalFunction * Fix Assert.Null failed for Mono in System.Reflectio.Emit Fixes #72858
-
-
- 07 8月, 2022 4 次提交
-
-
由 Andrew Au 提交于
-
由 Dan Espinosa 提交于
* Process incoming frames before sending frames for Http2Connection. Moves the ProcessIncomingFramesAsync method before sending the preface. This fixes a deadlock when using a NamedPipe as a transport where incoming frames need to be read before being able to write. Also rename _expectingSettingsAck to _receivedSettingsAck to clearly specify that the settings frame has being received. Fix #73097 * Addressed PR Feedback. Added test that will prevent regresions when using Http2 with NamedPipe as a transport. Addressed PR Feedback Fix #73307 * Update ConnectCallback_UseNamedPipe_Success test Fix #73307 * Moved ConnectCallback_UseNamedPipe_Success test to SocketsHttpHandlerTest_ConnectCallback_Http2 class Fix #73307 Co-authored-by: NDaniel Espinosa <luespino@microsoft.com>
-
由 Filip Navara 提交于
KRB5/GSSAPI on Arm32 is unable to handle dates far in the future
-
由 Sergei Pavlov 提交于
-
- 06 8月, 2022 1 次提交
-
-
由 Zoltan Varga 提交于
-