- 08 8月, 2015 7 次提交
-
-
由 Jared Parsons 提交于
We no longer use SCC bindings in our project yet the definitions remained around and caused a dialog prompt on solution open. Removing these as they are no longer needed
-
由 Wonseok Chae 提交于
Initial work to add telemetry to Interactive Windows - add telemetry for create/open/close Interactive Windows.
-
由 Charles Stoner 提交于
Refactor MetadataFileReferenceResolver classes
-
由 Wonseok Chae 提交于
-
由 Balaji Soundrarajan 提交于
Switch Section honors NewLineForBracesInControlBlock option
-
由 Charles Stoner 提交于
-
由 Tomáš Matoušek 提交于
Report a better exception when compiler test resource is not found
-
- 07 8月, 2015 23 次提交
-
-
由 Jared Parsons 提交于
The cross platform legs are causing roughly ~10% of all our Jenkins jobs to fail (3% Mac failure, 7% Linux failure). The main source of the instability is the NuGet and build step. As a short term work around I'm adding retry logic to those steps to get the false failure rate down.
-
由 Jared Parsons 提交于
Update the metadata reader to 1.1.0-alpha-00009
-
由 Manish Vasani 提交于
Fix some analyzer load diagnostics IDs and a duplicate roslyn diagnostic ID Fixes #3805 and #2990
-
由 Manish Vasani 提交于
Ensure that syntax tree actions are also re-executed by the solution crawler when compilation options (or analyzer rule severity) are changed. Fixes #3796
-
由 Andrew Casey 提交于
Fix mapping from line numbers to prompt spans
-
由 Manish Vasani 提交于
Fixes #3805 and #2990
-
由 VSadov 提交于
Fixed the order of optimizing argument reordering temps and packaging…
-
由 Andrew Casey 提交于
Bonus: Tighten projection span list asserts.
-
由 Charles Stoner 提交于
Conflicts: src/Workspaces/Core/Desktop/Workspace/MSBuild/MSBuildWorkspace.cs
-
由 Manish Vasani 提交于
Ensure that syntax tree actions are also re-executed by the solution crawler when compilation options (or analyzer rule severities) are changed. Fixes #3796
-
由 Charles Stoner 提交于
-
由 Jared Parsons 提交于
Modifications to bootstrap Roslyn on CoreClr without exceptions
-
由 Jared Parsons 提交于
Open source build improvements
-
由 Manish Vasani 提交于
Report diagnostics for exceptions within the analyzer driver.
-
由 Tomas Matousek 提交于
-
由 Manish Vasani 提交于
We execute the analyzer driver's initialization and core analysis tasks on a background thread. Any exceptions from the driver itself (not the analyzer callbacks, we already report diagnostics for those) were getting swallowed and analyzer execution also skipped silently. See #2980 (comment) for an example. We now report an exception diagnostic for analyzer driver crash (with the complete exception trace), so that the user knows what happened and we can diagnose the bug when the issue is reported. Fixes #3005
-
由 Andrew Casey 提交于
We used to maintain a separate data structure (```PromptLineMapping```) that kept track of the correspondence between line numbers in the surface buffer and the indices of the corresonding prompt subject buffers. Unfortunately, this data structure was being updated incorrectly, resulting in multiple prompt entries for individual line numbers and incorrect lookups because binary search was operating on an unsorted list. A simpler approach is to associate a line number with each ```ReplSpan``` (a standin for a subject buffer). We already have code for keeping this list up to date and the entries are guaranteed to be ordered by line number so we can still binary search for specific prompts. A number of new asserts validate all modifications of the ```ReplSpan``` list to ensure that it behaves as expected.
-
由 Richard L Ford 提交于
This change solves a number of problems that were resulting in problems compiling Roslyn with Roslyn on the CoreClr. I believe these problems were being masked by catching exceptions. The motivation for the changes is to be able to run Roslyn building Roslyn on Clr using the LLILC jit. LLILC cannot currently produce code to catch exceptions so it is important to avoid needless throwing of exceptions. 1. CoreClr was raising FileNotFound exceptions when trying to load the following assemblies: ``` System.Threading.Overlapped System.Threading.ThreadPool ``` I added references to these in src/Compilers/CSharp/CscCore/project.json. 2. System.Runtime in the version previously referenced in project.son had a reference to Internal.Uri. This is no longer the current module (it has been replaced by System.Private.Uri). So I changed the version number used in project.json from 22816 to 23109. 3. The path in these ruleset files: * src/Compilers/Core/CodeAnalysisRules.ruleset * src/Compilers/CSharp/CSharpCodeAnalysisRules.ruleset * src/Compilers/VisualBasic/BasicCodeAnalysisRules.ruleset was "..\..\Tools\Microsoft.CodeAnalysis.Toolset.Open\Rulesets\Roslyn.ruleset" which is not the current location of Roslyn.ruleset. It now is in "..\..\..\build\Rulesets\Roslyn.ruleset". The incorrect path was causing an exception to be thrown. 4. On the CoreClr csc uses a NoOpAnalyzerAssemblyLoader which just returns null when asked to get the assembly. This was causing a null reference exception. I added a null check to just return in this case. 5. The VbcCore project.json was updated to match CscCore.
-
由 Manish Vasani 提交于
Don't skip executing analyzer compilation actions in the IDE when ClosedFileDiagnostics flag is turned off OR the analyzer reports only hidden diagnostics. Compilation actions and compilation end actions can report diagnostics on open files, so they must always be executed. Fixes #4068
-
由 Heejae Chang 提交于
removed more dead code related to old ITaskItem.
-
由 Tomáš Matoušek 提交于
[Portable PDB] Implement GetMethodFromDocumentPositionm, GetMethodsIn…
-
由 Balaji Soundrarajan 提交于
Fix #3256 Switch section now honors the NewLineForBracesInControlBlock option
-
由 Tomas Matousek 提交于
[Portable PDB] Implement GetMethodFromDocumentPositionm, GetMethodsInDocument and GetSourceExtentInDocument APIs
-
- 06 8月, 2015 10 次提交
-
-
-
-
-
由 vsadov 提交于
Temp optimization relies on matching of the arguments and and corresponding temps if there are any. If params are morphed into ArrayCreationExpr before the temp optimization, some temps may not be matched to their arguments. A simple solution is just to do these operations in a different order. Fixes #4197
-
由 Jason Malinowski 提交于
ProcessOutput's members were IEnumerable which could have been very slow for repeated enumeration. Instead, use IList to guarantee stronger guarantees.
-
由 Jason Malinowski 提交于
If you had many, many errors (say, above 10,000), we'd spend many minutes of CPU time concatenating strings and garbage collecting the result. We should not do that.
-
由 Jason Malinowski 提交于
This only takes effect if you're building in a versionless command prompt.
-
由 Jason Malinowski 提交于
Rather than failing the build, if Roslyn is installed globally just don't try to deploy. Fixes #4189 (https://github.com/dotnet/roslyn/issues/4189)
-
由 Jason Malinowski 提交于
This flag was only used to add unecessary differences of behavior from local developer builds. By removing the flag everything acts the same.
-
由 Jason Malinowski 提交于
Since we're now on a stable VS version base, all tests should run everywhere now, so let's just do that. Fixes #860 (https://github.com/dotnet/roslyn/issues/860)
-