- 16 9月, 2020 15 次提交
-
-
由 Rikki Gibson 提交于
-
由 Rikki Gibson 提交于
-
由 Gen Lu 提交于
Exclude empty item from completion list
-
由 Sam Harwell 提交于
-
由 msftbot[bot] 提交于
Disable flaky optprof test
-
由 Jared Parsons 提交于
This test is hanging on test runs. Tracking issue #47720
-
由 CyrusNajmabadi 提交于
Show events completion inside NameOf
-
由 Gen Lu 提交于
Keep track of the trigger location for async completion
-
由 Jonathon Marolf 提交于
always nuget restore in graph mode
-
由 David Wengier 提交于
* Classify function pointers correctly * Use a stable sort for classifications
-
由 Gen Lu 提交于
-
由 Rikki Gibson 提交于
-
由 Fred Silberberg 提交于
-
由 Allison Chou 提交于
Add change signature support for target-typed new
-
由 CyrusNajmabadi 提交于
Fix lambda completion inside `new()` arguments
-
- 15 9月, 2020 21 次提交
-
-
由 Julien Couvreur 提交于
-
由 Youssef Victor 提交于
-
由 msftbot[bot] 提交于
only specify minimum heights and widths
-
由 msftbot[bot] 提交于
Add test to validate non-repro issue.
-
由 Jonathon Marolf 提交于
-
由 Fredric Silberberg 提交于
-
由 Fredric Silberberg 提交于
This is a mitigation for the threading issue discussed in https://github.com/dotnet/roslyn/issues/35818. When building VS, there is a flaky threading issue where an interpolation with multiple parts can end up with different nodes on different threads, causing an optimization in `SetParentOperation(ImmutableArray)` to be invalid. This makes the optimization much narrower, ensuring that it is actually valid for this case. The specific scenario here is an interplotated string with multiple parts. The steps are: 1. 2 threads attempt to get the children of an interpolated string operation at the same time. 2. Thread 1 gets the child for the first part of the string and caches it. 3. Thread 2 gets the child for the first part of the string and gets that same cached element. 4. Both threads attempt to get child 2. Neither hits the cache, so they get different nodes. 5. Both threads construct the array of interpolated string parts. 6. Thread 1 sets the parent of the interpolated string parts to be itself. 7. Thread 2 attempts to set the parent of the interpolated string parts to be itself. This returns early, because part 1 already has a parent, leaving part 2 with an unset parent. 8. Later, thread 2 asks for the parent of part 2. It is unset, violating the invariants, and we infinite loop. As an additional mitigation step, we also ensure that the process of constructing an interpolated string's inserts does not hit the cache. Since we never ask for an operation within a single expression, this will prevent 2 threads from seeing the same first element and differing subsequent elements.
-
由 CyrusNajmabadi 提交于
Fix convert-to-switch handling of comments at the end of blocks
-
由 Julien Couvreur 提交于
-
由 Gen Lu 提交于
-
由 Alireza Habibi 提交于
-
由 CyrusNajmabadi 提交于
Switch to an async nuget api to avoid blocking the UI thread.
-
由 Youssef Victor 提交于
-
由 Cyrus Najmabadi 提交于
-
由 Allison Chou 提交于
-
由 Rikki Gibson 提交于
-
由 Neal Gafter 提交于
Related to #47471
-
由 Neal Gafter 提交于
* Decision dag: union the set of values when there are two predecessor states. If a state has no set of values for a temp on one of the edges, that means it is unrestricted. This fixes a bug in decision dag construction. I was not able to construct a test that is affected, but the change makes a visible difference in an internal dump of the decision dag for the newly added test `MultiplePathsToState_01`, which we walked through in a compiler team meeting on 2020-09-09. * Use a builder as suggested in review.
-
由 Allison Chou 提交于
Enable unused integration test parameter
-
由 Tomáš Matoušek 提交于
* Fix steaming ISB service hang * Fix comment
-
由 Alireza Habibi 提交于
-
- 14 9月, 2020 4 次提交
-
-
由 Cyrus Najmabadi 提交于
-
由 Cyrus Najmabadi 提交于
-
由 CyrusNajmabadi 提交于
Fix incorrect mapping of nodes/tokens in VBgrammar generator
-
由 David Wengier 提交于
Co-authored-by: NCyrusNajmabadi <cyrus.najmabadi@gmail.com>
-