On Fri, Jan 3, 2014 at 8:11 AM, Manuel Klimek <[email protected]> wrote:
> On Fri, Jan 3, 2014 at 2:26 AM, Alexander Kornienko <[email protected]>wrote: > >> >> On 2 Jan 2014 20:06, "Manuel Klimek" <[email protected]> wrote: >> > >> > >> > lg. >> > >> > >> > ================ >> > Comment at: clang-tidy/ClangTidy.h:107-109 >> > @@ +106,5 @@ >> > +/// >> > +/// FIXME: Ideally we'd want to build a more generic way to use >> > +/// \c FrontendAction based checkers in clang-tidy, but that needs some >> > +/// preparation work first. >> > +class ClangTidyAction : public ASTFrontendAction { >> > ---------------- >> > Alexander Kornienko wrote: >> > > Manuel Klimek wrote: >> > > > Why? >> > > I'm not sure what exactly did you mean when writing this, and I was >> thinking already about asking you, if this comment is still applicable. >> > Yea, I don't understand my own fixme any more... Probably we should >> delete it :) >> > >> > ================ >> > Comment at: clang-tidy/ClangTidy.h:110 >> > @@ +109,3 @@ >> > + /// \brief Returns an action that runs the specified clang-tidy >> checks. >> > + virtual FrontendAction *create() LLVM_OVERRIDE; >> > + >> > ---------------- >> > Cool. One question is whether it's worth having this small function >> (with the trivial class) in here at all. Do we expect this to be called >> outside of clang-tidy itself? >> >> It's already called somewhere in tooling via the FrontendActionFactory >> interface. What would be a better alternative to this? >> > > Oh, this is now a FrontendActionFactory? I thought we'd wanted to get an > interface that is free of FrontendAction* stuff. > > To me, we have conceptually two parts in clang-tidy: > 1. a library interface that provides an ASTConsumer and PPCallbacks to be > used by other clang-based analysis pipelines > 2. something to run clang-tidy over a bunch of code > > Conceptually, the first one should be as far as I understand completely > independent of anything related to FrontendAction. The second part then > creates a FrontendAction that just runs clang-tidy's ASTConsumer and > PPCallbacks > Please take a look at the latest update. There's still no separation between ASTConsumer and PPCallbacks stuff (there's a FIXME in place), but it should now be closer to your proposed design.
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
