Closed Xazax-hun closed 5 years ago
Notes are parsed from the clang-tidy output and a bug event is created from it in the generated plist. I think we should still use the plist output from the clang static analyzer even if notes are supported in the next release. A plist output would be better from clang-tidy too.
I am not sure that an event is a good way to store a note. The notes are not path sensitive pieces of the reports. And sometimes I get unspecified empty name results from tidy. Maybe there is a bug lurking in the output parser.
This requires work at both the client and server side and the schema will change.
Note that, the notes are not emitted to the plist yet: https://github.com/llvm-mirror/clang/blob/master/lib/StaticAnalyzer/Core/PlistDiagnostics.cpp#L284
Could you provide some cland-tidy output which we fail to parse correctly?
Oh, the issue I experienced might be unrelated to the title. Here is a sqlite database of a run that has some check results without name: https://dl.dropboxusercontent.com/u/4131903/codechecker.sqlite.tar.lz
I've created a new issue for the check results without the name, that problem should be discussed further there.
I think this issue has already been solved in #1818. What do you think @gyorb?
Yes, it is done we can close this.
Clang tidy sometimes output notes. These notes are connected to a warning and provides the user with additional information e.g.: where a variable was declared. This is not part of path sensitive information. It would be great to store/display those notes.
Also note that, the clang static analyzer will also support and emit notes in the next release: https://reviews.llvm.org/D24278