- 02 2月, 2021 16 次提交
-
-
由 Stephen Toub 提交于
-
由 Ove Bastiansen 提交于
* adding span version of ReceiveMessageFrom fix #43933 * Started using new testsetup for the receivemessagefrom * removed code duplication in socketpal.windows * removed unused buffers parameter * adding non-zero offset to the ReceiveMessageFrom test * added documentation * offset fix for EAP tests * fixing pr comments
-
由 Peter Sollich 提交于
This fixes a problem where accidentally too much code was excluded with USE_REGIONS. This caused the compaction decision to be inconsistent between heaps, with some heaps doing compacting and some heaps doing mark&sweep. This is a problem, because the relocation phase relies on relocation information stored in the free space between used objects, and the free space is being overwritten by the sweep. It's an issue only for object pointers between heaps, but these are very common.
-
由 Andrew Au 提交于
-
由 Jeremy Koritzinsky 提交于
* Produce a .version file for Microsoft.NETCore.App in the runtime layout. * Update src/installer/pkg/sfx/Microsoft.NETCore.App/Microsoft.NETCore.App.Runtime.sfxproj * Update src/installer/pkg/sfx/Microsoft.NETCore.App/Microsoft.NETCore.App.Runtime.sfxproj
-
由 Aleksey Kliger (λgeek) 提交于
Use it to take the slow path in metadata lookups when there's at least one delta applied.
-
由 Andy Ayers 提交于
Change #47646 did not remove the code in `fgMakeBasicBlocks` that set profile data. That masked some flaws in `fgIncorporateProfileData`: * it should process all blocks; nothing has been imported yet * it must honor a constraint that some handler blocks can't be rare. This fixes the above, and removes the early profile setting.
-
由 Chuck Ries 提交于
-
由 David Wrighton 提交于
- Fix class type probes (the increment of the count wasn't correct) - Use the 64bit integer encoders for all the pgo data - New section for R2R PE Files containing instrumentation data - R2RDump functionality to display all the data embedded in the file - Enable BBOPT for optimized builds in a more unconditional fashion Future PGO work will include - Move Pgo type handle histogram processing into JIT (which will make type guessing work in crossgen2 as well as in the runtime) - Triggers for controlling Pgo data extraction - Size control for pgo instrumentation data With this checkin, the feature is functional from a crossgen2.exe point of view, but its not polished, and it cannot be used from the Crossgen2 sdk integration yet (as the sdk does not have the ability to pass an extra pair of arguments to the compiler.
-
由 Kevin Jones 提交于
The DisableCertificateDownloads property on the chain policy controls all network activity when building a chain on macOS, not just AIA fetching. If set to true, the (default) revocation policy would fail because the network would be treated as unavailable. On macOS, as a work around, permit the network activity if revocation checking is explicitly enabled.
-
由 Petr Onderka 提交于
-
由 John Salem 提交于
-
由 Stephen Toub 提交于
PLINQ's QueryTaskGroupState handles joining with queued tasks, and handles the exceptions that emerge from waiting on them. AggregateExceptions are special-cased, and are flattened to eliminate extra levels that may have been added explicitly or implicitly; the code then checks to see whether all of the exceptions contained in the aggregate represent cancellation. However, it fails to account for the possibility that there are no exceptions it contains, which could happen if the only exceptions thrown are themselves empty AggregateExceptions. If that happens, the code makes some invalid assumptions that lead to a failed assert and then throwing an ObjectDisposedException. The fix is just to check whether the flattened AggregateException is empty and to treat that not as cancellation but as a normal exception.
-
由 Stephen Toub 提交于
Uses the recently added support for SocketsHttpHandler.ConnectCallback to be used for synchronous operations to add support for HttpWebRequest.ReadWriteTimeout. That value is now used to set the Read/WriteTimeout on the underlying socket, with it then affecting all synchronous reads and writes on that connection. The ReadWriteTimeout is added to the cache key for whether the cached HttpClient instance can be used, as is whether the request is sync or async because that influences how connections are created from the HttpClient instance.
-
由 Adeel Mujahid 提交于
* Update license headers in new files * Run dotnet-format on changeset
-
由 Johan Lorensson 提交于
* Add support for EventPipe stackwalks and method formatting on Mono. * Review feedback. * Explicit gboolean -> bool conversion.
-
- 01 2月, 2021 6 次提交
-
-
由 Eirik Tsarpalis 提交于
* Add test validating against regression in #46469 * fix test bug
-
由 Kevin Jones 提交于
-
由 Marek Safar 提交于
It's strictly speaking not required but the type is handled by JIT specially and illink would need to recognize this one of case.
-
由 LateApexEarlySpeed 提交于
* Issue-20619 early return existing value. * issue-20619 Fix comment: Reuse check inside of SetItem; Also use customer-supplied value comparer for TryUpdate(). * issue-20619 Fix tab-space issue. * issue-20619 Fix comment: Revert: use customer-supplied value comparer for TryUpdate().
-
由 Simon Nattress 提交于
* Remove duplicated unwind and GC info table entries. * Overall, reduces the framework size with CG2 by 2.4MB (89,835,064 -> 87,439,416 bytes)
-
由 monojenkins 提交于
This error was introduced in my last PR about icordebug implementation. Fixing it as @lambdageek told me that it was reported on coverity scan. Co-authored-by: Nthaystg <thaystg@users.noreply.github.com>
-
- 31 1月, 2021 3 次提交
-
-
由 Tom Deseyn 提交于
-
由 Andy Ayers 提交于
Instead of trying to set block weight data when initially building the flowgraph, do it as a phase just before importation. The importer looks at block weights, so these must be set before importation. Note profile incorporation relies only on the IL level view of the method, and does not depend on the IR.
-
由 monojenkins 提交于
Co-authored-by: Nvargaz <vargaz@users.noreply.github.com>
-
- 30 1月, 2021 15 次提交
-
-
-
https://github.com/mono/linker由 dotnet-maestro[bot] 提交于
Microsoft.NET.ILLink.Tasks From Version 6.0.0-alpha.1.21076.1 -> To Version 6.0.0-alpha.1.21079.1 Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
由 Stephen Toub 提交于
* Use [NotNullWhen(true)] in more places Did a quick search/audit for methods that returned bool and took nullable object as the first input, and added [NotNullWhen(true)] where it was obviously correct. * Address PR feedback
-
由 Jeff Handley 提交于
-
由 Anton Lapounov 提交于
-
由 Omair Majid 提交于
The destructor calls delete[] to free up the memory allocated with new[] in the constructor.
-
由 Omair Majid 提交于
The code tries to write the status from `rcClose`, but accidentally writes out the value of `rc` instead.
-
由 Omair Majid 提交于
memcpy is defined to only work when the source and destination memory regions do not overlap. In here, they do. So let's use memmove instead.
-
由 Omair Majid 提交于
Using NewHolder results in the data being `delete` d after it's used. The underlying data is an array, so it should be `delete[]` ed instead.
-
由 Omair Majid 提交于
These free() calls are done in the middle of the method body; it's possible that a duplicate free() can be called on these at the cleanup code that runs before the method returns. So set these pointers to null after free()ing them to avoid free() already free()d memory.
-
由 Ankit Jain 提交于
* [wasm] Add $( WasmBuildAppAfterThisTarget) to control which target .. triggers WasmAppBuild. Defaults to `Publish`. * [wasm] emcc, mono-aot-cross doesn't like relative output paths eg. `-o foo/bar.js` .. so convert that to absolute paths. * [wasm] Check for WasmMainJSPath close to it's use * [wasm] Add new $(WasmGenerateAppBundle) to allow skipping app bundle .. generation. This is useful when you want to do all the other steps for wasm, but handle the packaging/bundle generation yourself. Eg. Blazor * [wasm] Ensure that generated dotnet.{js,wasm} get copied to the app dir This is usually done as part of the bundle generation (NativeAssets). But in cases when that step is being skipped, we still want to copy over generated dotnet.{js,wasm} (eg. in case of AOT) to the app dir. This is the case that blazor uses. * [wasm] Allow skipping mono-cil-strip step In some cases, eg. blazor (till we fix it), `mono-cil-strip` can break the app, because of bugs, or linker issues etc. Also, do this only in the AOT case. * [wasm] Add support for building wasm projects outside the tree .. using local runtime builds. * [wasm] fix up the support for local build dir * [wasm] add comments for new properties * [wasm] LocalBuild.*: add more comments, validation * [wasm] sample: Use BeforeTargets for Prepare target, as a psuedo test for that
-
由 Jeremy Koritzinsky 提交于
Since we're in an IL stub, we don't need to support breakpoints in the stub itself. The debugger will try to resolve the target method and move the breakpoint there. In actuality, that's the part that's failing. Somewhere in the ReversePInvoke helper call chain, we're touching the secret arg register. As a result, when the debugger tries to move the breakpoint, it crashes. By doing the same work ourselves, the debugger doesn't need to resolve the stub secret arg since it never sees a breakpoint in the stub itself. Use UMEntryThunk::GetManagedTarget like what the old system used.
-
由 Stephen Toub 提交于
-
由 Tomáš Rylek 提交于
During my experimentation with callchain optimizations I noticed that the profile map only contains meaningful information when we're generating PDB or PerfMap at the same time - apparently I had the PDB generation turned on during debugging so I overlooked this. Thanks Tomas
-
由 Stephen Toub 提交于
This is showing up a bit on ASP.NET startup paths, and we can be more efficient in how we handle allocations and work.
-