1. 21 Aug, 2016 3 commits
  2. 17 Aug, 2016 3 commits
  3. 14 Aug, 2016 1 commit
  4. 04 Aug, 2016 1 commit
  5. 03 Aug, 2016 1 commit
  6. 01 Aug, 2016 7 commits
  7. 31 Jul, 2016 2 commits
  8. 29 Jul, 2016 2 commits
  9. 26 Jul, 2016 1 commit
  10. 25 Jul, 2016 3 commits
  11. 23 Jul, 2016 1 commit
  12. 22 Jul, 2016 1 commit
  13. 21 Jul, 2016 2 commits
  14. 20 Jul, 2016 3 commits
    • Sven Brauch's avatar
      Hopefully fix screwed-up clang higlighting. · 5d076e29
      Sven Brauch authored
      If you see this happening again, please tell me.
      Make sure to also update kdevplatform.
      CCMAIL:kdevelop-devel@kde.org
      REVIEW:128480
      The issue was that the clang plugin wouldn't store the modification
      revision for files which were marked as "not modified" by KTextEditor.
      Thus, no translation of the highlighting to the new state could be
      performed for those documents. This is wrong, since a document can
      be "not modified", but can still have changed in the time between the
      start of the parse job start and end, etc. Just imagine you
      press save after each character typed.
      5d076e29
    • David Nolden's avatar
      Search declarations by ID in addition to their range · 50041d95
      David Nolden authored
      The declaration search based on "findContextAt" completely fails when
      there are multiple different versions of a header which were parsed
      differently depending on a macro. Also, findContextAt does a linear
      search, and may be slow when there is a large list of contexts.
      Whenever possible, try finding the declaration by its qualified
      identifier based on the duchain symbol table, which should be more
      efficient and scalable.
      
      Extract the qualified identifier from clang by
      following up the chain of semantic parents. After declarations
      were found based on the symbol table, they are filtered
      the same way as those found by findContextAt -- so there should
      be no regressions. When the search fails, then use the previous
      findContextAt method instead.
      
      Added a test which tests several difficult cases that failed
      without this change, and that succeed with it.
      
      This also fixes the testDeclarationsInsideMacroExpansion test,
      which was EXPECT_FAIL, but which succeeds now.
      50041d95
    • David Nolden's avatar
      If no pinned translation unit is available, ask the duchain · a36ae328
      David Nolden authored
      This fixes highlighting, navigation, etc. for complex cases
      where the primitive "buddy" mechanism doesn't help finding
      the translation unit. The buddy mechanism is not correct for
      this case, because it was built around header/source pairs,
      but fails in more complex cases. The duchain import chain
      directly leads to the original transition unit from where
      a header is included recursively.
      a36ae328
  15. 18 Jul, 2016 1 commit
  16. 17 Jul, 2016 2 commits
  17. 15 Jul, 2016 3 commits
  18. 14 Jul, 2016 3 commits