- 17 11月, 2015 1 次提交
-
-
由 Kevin Halverson 提交于
Fix hard-coded error message in DynamicView tests...
-
- 14 11月, 2015 15 次提交
-
-
由 Kevin Halverson 提交于
-
由 Kevin Halverson 提交于
Always "fail fast" in EE unit tests...
-
由 Jason Malinowski 提交于
Clean up complexity around OfficialBuild flag that is breaking official builds
-
由 Jason Malinowski 提交于
-
由 Jason Malinowski 提交于
The official build flag specifies which assembly version is baked into Roslyn. This was being used in various ways to either include or not include binding redirects, but we now always want binding redirects. Also, we were passing the flag along to the compiler to figure out which NewVersion to bake into the redirects, but setting nothing at all conveniently does exactly what we want.
-
由 Kevin Halverson 提交于
We always want to "fail fast" if there is an Exception in the EE unit tests (we never want to go down the "generate a non-fatal Watson" code path). Also, there were several places where we were generating non-fatal Watsons for cases that were not truly "exceptional" (the Exception was being handled reasonably and returned as an "error" result). We should only have a filter to generate a non-fatal Watson in places where it would also make sense to throw an unhandled Exception.
-
由 Andy Gocke 提交于
Mark *CompilationOption overloads as backcompat
-
由 Andrew Casey 提交于
Restore missing nuget implementations
-
由 Andrew Casey 提交于
...to Roslyn.VisualStudio.Setup.vsix
-
由 Manish Jayaswal 提交于
add imports for System.Console in response file - fixes #6768
-
由 Manish Jayaswal 提交于
-
由 Andy Gocke 提交于
-
由 Andy Gocke 提交于
NuGet ZIP is now used everywhere in CI
-
由 Jason Malinowski 提交于
Enable F5 deployment atop Visual Studio 2015 Update 1
-
由 drognanar 提交于
Handle case when not all projects exit break state
-
- 13 11月, 2015 10 次提交
-
-
由 Manish Vasani 提交于
Fix the batch pragma suppression fixer to handle the case where multiple diagnostics with different IDs are being suppressed on the same token Current logic in batch pragma fixer merges individual fixes in a document by applying each pragma fix sequentially and keeping track of new diagnostics spans for remaining diagnostics. The latter logic that computes new spans was using out of date spans, causing the subsequent fixes to be incorrectly applied. This change fixes that logic. Fixes #6455
-
由 Jason Malinowski 提交于
-
由 Tomáš Matoušek 提交于
Avoid hard test dependency on FX 3.5
-
由 Jared Parsons 提交于
Fix a couple of flaky suites
-
由 Tomáš Matoušek 提交于
Skip GAC test
-
由 Tomas Matousek 提交于
-
由 Jason Malinowski 提交于
-
由 Nick Guerrera 提交于
Adjust ErrorLogger to conform to current SARIF draft
-
由 Kevin Halverson 提交于
Rename VB Scripting Namespace...
-
由 Manish Vasani 提交于
Fix the batch pragma suppression fixer to handle the case where multiple diagnostics with different IDs are being suppressed on the same token Current logic in batch pragma fixer merges individual fixes in a document by applying each pragma fix sequentially and keeping track of new diagnostics spans for remaining diagnostics. The latter logic that computes new spans was using out of date spans, causing the subsequent fixes to be incorrectly applied. This change fixes that logic. Fixes #6455
-
- 12 11月, 2015 14 次提交
-
-
由 Neal Gafter 提交于
Documentation for csc and vbc command-line.
-
由 Jared Parsons 提交于
This test was failing because it's possible for a Task to execute on the thread where Wait is called (scheduler can run it inline). The particular assert here must run on a different thread as it's checking the ability to acquire a Mutex handle. Use an explicit thread now to ensure the check is correct.
-
由 Jared Parsons 提交于
Under stress scenarios even seeing if the timeout was at least a specific period of time fails. Going to a simpler check that we simply hit the value.
-
由 Jared Parsons 提交于
-
由 Jared Parsons 提交于
-
由 Jared Parsons 提交于
-
由 Jared Parsons 提交于
The compiler server suites were using a flawed method for tracking process used in the test. It tried to identify the processes used by the test by the installation path. At the end of any test it would then kill all processes originating from that path. That's flawed for a couple of reasons: 1. There is no guarantee that all processes with that path were started by the unit test 2. There is no guarantee external events won't prematurely kill the processes created by the unit tests. Both of these were happening in our build environment. To fix these tests I: - Remove all path based identification of processes - Changed all tests that used long running processes to instead use in-proc servers closes #6540
-
由 Jared Parsons 提交于
This suite was flaky because it relied on timeouts and cross process communication. The latter in particular was problematic because it depended on being able to control the lifetime of VBCSCompiler.exe. This is a process frequently killed in our scripts as it can lock DLLs and is always safe to kill. Fixed the flakiness by refactoring the APIs and eliminating the timeouts
-
由 John Hamby 提交于
Moving IOperation from future to master
-
由 Tomáš Matoušek 提交于
Enable vb test
-
由 Heejae Chang 提交于
put more information for analyzer exception
-
由 Tomas Matousek 提交于
-
由 Jason Malinowski 提交于
-
由 Jason Malinowski 提交于
We sometimes need an .exe and sometimes need a .dll, so let's just make it explicit. We can still get the assembly name either way.
-