On Mon, Aug 5, 2024 at 5:39 PM Théo Lebrun <theo.lebrun@xxxxxxxxxxx> wrote:
>
> Hello Masahiro,
>
> On Sat Aug 3, 2024 at 10:42 AM CEST, Masahiro Yamada wrote:
> > On Thu, Jul 4, 2024 at 6:28 PM Théo Lebrun <theo.lebrun@xxxxxxxxxxx> wrote:
> > > +def filter_entries(datastore, filters):
> > > + for entry in datastore:
> > > + if filters == []:
> > > + yield entry
> > > + continue
> >
> > Maybe, this can be checked on the caller side.
> > (Note, I did not test this at all)
> >
> > if args.file_filter:
> > datastore = filter_entries(datastore, args.file_filter)
>
> Agreed.
>
> > > +
> > > + assert entry['file'].startswith(entry['directory'])
> > > + # filepath is relative to the directory, to avoid matching on the absolute path
> >
> > Does this assertion work with the separate output directory
> > (O= option)?
> >
> > Just try this command:
> >
> > $ make LLVM=1 O=/tmp/foo clang-tidy
>
> Indeed this does not work. It requires some changes to Makefiles to see
> that it does not work, as this assertion is only used when a filter is
> passed (which current cmd_clang_tools cannot do).
>
> What would you recommend to fix that? If we take the first entry:
>
> {
> "command": "clang ...",
> "directory": "/tmp/foo",
> "file": "/ABSOLUTE/PATH/linux-clang-tidy/arch/mips/crypto/crc32-mips.c"
> },
>
> I don't see an easy way to know that "/ABSOLUTE/PATH/linux-clang-tidy/"
> should be removed. We could probably compute it from the command value
> but that isn't a great way forward.
>
> There are some entries that look like this. Those are easy to deal with.
>
> {
> "command": "clang ...",
> "directory": "/tmp/foo",
> "file": "/tmp/foo/drivers/tty/vt/defkeymap.c"
> }
The top of the source tree is derived from __file__
srctree = str(pathlib.Path(__file__).parents[2].resolve())
--
Best Regards
Masahiro Yamada
