1. 24 Jan, 2019 1 commit
  2. 22 Jan, 2019 1 commit
  3. 21 Jan, 2019 3 commits
    • Thomas Schöps's avatar
      Contextbrowser: Ability to show combined problems and decl tooltip · 2071db63
      Thomas Schöps authored
      Without this patch, if both a problem and a declaration are available for a
      given cursor location, KDevelop would only show the problem tooltip. This makes
      the declaration tooltip inaccessible, which can be an annoyance. This patch makes
      it show a combined tooltip instead in this case, allowing to access both the
      problem(s) and the declaration.
      
      Test Plan: Tested manually.
      
      Reviewers: #kdevelop, mwolff
      
      Reviewed By: #kdevelop, mwolff
      
      Subscribers: mwolff, kdevelop-devel
      
      Tags: #kdevelop
      
      Differential Revision: https://phabricator.kde.org/D18229
      2071db63
    • Thomas Schöps's avatar
      Properly display argument names of template functions · 49d503f8
      Thomas Schöps authored
      When hovering over a C++ template function, the function argument names were
      shown incorrectly. For example, for the following function definition:
      
          template <int a>
          void foo(char b, char c) {}
      
      The function would be displayed like this in the tooltip:
      
          void foo(char a, char b)
      
      This is because argument names, different from argument types, are fetched from
      argumentContext->localDeclarations() for display. In case of template functions,
      both the template arguments and the function arguments are in this list, which
      was not accounted for. This diff changes the behavior to count the function
      arguments from the end of the local declarations instead of the start, which
      fixes the bug. (Note: I am not sure whether it is possible that the local
      declarations list also contains other entries. I did not observe this during
      some short testing. Also, not sure what the situation is for other languages
      than C++.)
      
      Test Plan: Some manual testing.
      
      Reviewers: #kdevelop, mwolff
      
      Reviewed By: #kdevelop, mwolff
      
      Subscribers: aacid, mwolff, kdevelop-devel
      
      Tags: #kdevelop
      
      Differential Revision: https://phabricator.kde.org/D18218
      49d503f8
    • Script Kiddy's avatar
      SVN_SILENT made messages (.desktop file) - always resolve ours · 37cc98f3
      Script Kiddy authored
      In case of conflict in i18n, keep the version of the branch "ours"
      To resolve a particular conflict, "git checkout --ours path/to/file.desktop"
      37cc98f3
  4. 20 Jan, 2019 6 commits
  5. 16 Jan, 2019 4 commits
  6. 15 Jan, 2019 11 commits
  7. 14 Jan, 2019 3 commits
  8. 13 Jan, 2019 2 commits
  9. 12 Jan, 2019 1 commit
  10. 11 Jan, 2019 4 commits
    • Friedrich W. H. Kossebau's avatar
    • Friedrich W. H. Kossebau's avatar
      appimage: skip unneeded cp of cmake, removed later again · ff326bb2
      Friedrich W. H. Kossebau authored
      Test Plan: AppImage still built and starting.
      
      Reviewers: #kdevelop, brauch, kfunk
      
      Reviewed By: #kdevelop, kfunk
      
      Subscribers: kdevelop-devel
      
      Tags: #kdevelop
      
      Differential Revision: https://phabricator.kde.org/D18175
      ff326bb2
    • Thomas Schöps's avatar
      Clang plugin: Handle CUDA files better · 71c65a65
      Thomas Schöps authored
      Summary:
      This fixes two issues in how KDevelop's clang plugin handles CUDA files (.cu source files and .cuh headers):
      1) Those file types were not treated as source files, therefore if such a file was modified, it was not passed among the list of modified source files to clang for re-parsing. The re-parsing thus incorrectly used the unmodified file on disk instead. This is addressed by the change to documentfinderhelpers.cpp and kdevclang.xml.
      2) The patch passes the general code parser settings also to CUDA files (change in parsesession.cpp). This is for example important to get flags such as -std=c++11 that are required for correct parsing. However, it is not correct: the build system may use separate options for CUDA files (e.g., CMake's CUDA_NVCC_FLAGS). But it might still be better than the previous behavior of not passing any options except -xcuda, since it might be somewhat likely that it is a reasonable guess.
      
      Additional comments:
      a) I am not sure whether the change in clanghelpers.cpp is required, but it seems coherent.
      b) I changed the CUDA mime types in kdevclang.xml to inherit from text/x-c++src/hdr instead of text/x-csrc/hdr since the CUDA files can contain C++ code.
      c) This patch is not sufficient to enable proper CUDA support. I'll post to the kdevelop-devel mailing list for that.
      
      Test Plan: Tested manually.
      
      Reviewers: #kdevelop, kfunk
      
      Reviewed By: #kdevelop, kfunk
      
      Subscribers: kfunk, apol, kdevelop-devel
      
      Tags: #kdevelop
      
      Differential Revision: https://phabricator.kde.org/D17909
      71c65a65
    • Script Kiddy's avatar
      SVN_SILENT made messages (.desktop file) - always resolve ours · e7bd7d39
      Script Kiddy authored
      In case of conflict in i18n, keep the version of the branch "ours"
      To resolve a particular conflict, "git checkout --ours path/to/file.desktop"
      e7bd7d39
  11. 10 Jan, 2019 4 commits
    • René J.V. Bertin's avatar
      clang: detect Clang builtin dirs at runtime on Unix · 41a0f0ce
      René J.V. Bertin authored
      ec5477f3 introduced runtime detection
      of the clang builtin header directory on MS Windows. This commit
      implements the same behaviour on Unix. Some refactoring and code
      cleanup too.
      
      Differential Revision: https://phabricator.kde.org/D17858
      41a0f0ce
    • Milian Wolff's avatar
      Merge branch '5.3' · d0fbbfe3
      Milian Wolff authored
      d0fbbfe3
    • Milian Wolff's avatar
      Show size and alignment information in tooltips for typedef or alias · be490331
      Milian Wolff authored
      Summary:
      This patch lets us view the size information when inspecting type
      aliases. E.g. for C++ code like the following, we'll now see the
      size (8) and alignment (also 8) when hovering Bar. Previously,
      that was only shown when hovering Foo:
      
          struct Foo { double m; };
          using Bar = Foo;
      
      Sadly, this patch by its own isn't yet enough for the more interesting
      cases, namely aliases of explicit template instantiations, such as
      
          using Ptr = std::shared_ptr<int>;
      
      Here, we still don't show any size information, since we don't track
      that information properly. In the builder, we only encounter std::shared_ptr
      as a template, which doesn't have any size information yet. We would
      need to build a declaration for the template instantiation of
      std::shared_ptr, at which point we could ask libclang for the concrete size...
      
      Still, this patch goes into the right direction already, I believe.
      
      Reviewers: #kdevelop
      
      Subscribers: kdevelop-devel
      
      Tags: #kdevelop
      
      Differential Revision: https://phabricator.kde.org/D18097
      be490331
    • Milian Wolff's avatar
      Actually cleanup the duchain from the background thread · 48e5892f
      Milian Wolff authored
      Turns out that my commit bf183ce7 from May 2017 moved the
      background cleanup task into the foreground thread... Sorry for that!
      
      When we connect to the timer event, we must use a context object
      that lives within the background thread. The thread itself *does not*
      live in itself!
      
      Fix this by passing the timer itself as context and
      add an assertion to verify we don't ever do this again.
      
      BUG: 388743
      48e5892f