1. 04 3月, 2015 6 次提交
    • M
      Re-enable passing test TestFixAllInDocument_QualifyWithThis. Matt Warren fixed... · 22112a91
      Manish Vasani 提交于
      Re-enable passing test TestFixAllInDocument_QualifyWithThis. Matt Warren fixed the underlying issue in SyntaxTree.GetChanges(SyntaxTree).
      22112a91
    • P
      Made auto-properties readonly where possible. · 9dea9584
      Paul Harrington 提交于
      9dea9584
    • S
      Use cached true, false, and empty tasks in a handful of places · 5ed2a672
      Stephen Toub 提交于
      In a handful of places, Task.FromResult(true), Task.FromResult(false), and Task.FromResult(default(T)) were being used, even though equivalent cached tasks are available in SpecializedTasks. This commit just changes those ~20 call sites to use already cached tasks.
      5ed2a672
    • B
      correct text navigation in VB (and C#) strings · c4a4da28
      Brett Forsgren 提交于
      c4a4da28
    • B
      Fixed some edge cases around leading trivia and added more tests. · c1c9a614
      Balaji Krishnan 提交于
      Also added a smart, where, if we are in an expression bodied member context
      and if the right side of arrow clause contains a block, we generate local
      as if we're in a block context. This leads to an experience where we can
      keep the expression body syntax and also have introduce a local (since we
      have a block) without having to rewrite it to an equivalent full bodied
      member whose body is always a block.
      c1c9a614
    • D
      Better handle dismissed rename tracking sessions · 499a46f5
      David Poeschl 提交于
      Fixes #540 "Stale rename entries in the lightbulb menu"
      
      Prior to this change, dismissed rename tracking sessions continued to
      provide codefixes which would silently fail when invoked, without
      completing the rename operation.
      
      With this change, we now do the following when rename tracking is
      manually dismissed:
      
      - Trigger the diagnostic service to reanalyze the document, thus
      removing the rename tracking diagnostic.
      - Stop providing a codefix when requested
      - Show a dialog if a cached codefix is invoked explaining that the
      rename was not performed.
      
      Here are the there user scenarios:
      
      1. The user dismisses rename tracking
      
          - The lightbulb will continue to show in the margin, regardless of
      whether any codefix/refactoring is actually still available.
      
      2. The user expands the lightbulb without it ever having been expanded
      before.
      
          - In this case, we are queried for fixes and will not return any, so
      either the lightbulb will disappear or it will contain other relevant
      codefixes/refactorings at that position, excluding rename.
      
      3. The user expands the lightbulb after it was previously expanded.
      
          - Our fix was already cached, so we can only give a reasonable message
      when it is invoked.
      
      Once #885 is fixed, we can instead implement IDiagnosticUpdateSource and
      avoid forcing diagnostic reanalysis.
      499a46f5
  2. 03 3月, 2015 4 次提交
  3. 28 2月, 2015 4 次提交
  4. 27 2月, 2015 9 次提交
  5. 26 2月, 2015 2 次提交
  6. 25 2月, 2015 4 次提交
    • M
      Fix a test failure. · 7b8a3347
      Manish Vasani 提交于
      7b8a3347
    • M
      Address couple of issues related to analyzers: · 5cd0be7c
      Manish Vasani 提交于
      1) Don't crash command line builds when reporting analyzer diagnostics. Reporting analyzer diagnostics may involve invoking user code for LocalizableString message, which may throw while finding resources or for other unknown reasons. We catch exceptions from these callbacks and report a diagnostic with the inner exception message.
      I verified that this addressed the crash seen in #271 and #813 and we started seeing this new info diagnostic.
      
      2) Workaround for #271: The underlying cause of the crash is that our assembly resolver/CLR is unable to fallback to the main assembly for resources and ends up looking for the satellite resource assembly. We had seen this crash quite a while back, but it went away and this seems to have come up again. Workaround, as applied back then, is to mark the analyzer assembly with NeutralResourcesLanguageAttribute.
      I verified that adding this attribute to our core analyzer assemblies caused the resource to be correctly found and we don't even see the exception diagnostic from (1) above, but instead the actual diagnostic reported by the analyzer.
      5cd0be7c
    • V
      Add explicit StringComparison.Ordinal where it was intended · 52399299
      Vladimir Reshetnikov 提交于
      Fix #409:
      Add explicit StringComparison.Ordinal or StringComparison.OrdinalIgnoreCase where it was intended,
      or replace single-char string literals with chars literals.
      Also, provide InvariantCulture in TryParse where it was intended.
      52399299
    • B
      Make OutliningSpan Creation not assume MultiLine Comment always end with'*/' · 9a950a8f
      Balaji Soundrarajan 提交于
      Fixes #791: The code '\*' is parsed as MultiLine Comment even with the
      presence of the corresponding '*\' in some cases. Hence, OutLining Span
      creator should not assume that '*\' will always be present for Multiline
      Comment and handle the creation appropriately
      9a950a8f
  7. 24 2月, 2015 8 次提交
  8. 23 2月, 2015 3 次提交