On Tue, Jul 14, 2026 at 4:40 PM Nathan Sidwell <[email protected]> wrote: > > oops > > On 14/07/2026 10:24, Nathan Sidwell wrote: > > On 14/07/2026 03:03, Richard Biener wrote: > >> On Mon, Jul 13, 2026 at 4:53 PM Nathan Sidwell <[email protected]> wrote: > > > >>> This annoying. > >>> > >>> I contemplated just changing AWK to only accept the names at the > >>> beginning of a > >>> line (with whitespace), but then considered just running the C > >>> preprocessor > >>> first -- that way one gets all the CPP behaviour and no surprises. > >> > >> Hmm, I'm not sure it's good to open up the can of worms of using the host > >> preprocessor (is there a CPP_FOR_BUILD?). Can't you fix the awk script > >> to honor comments? Can't we perform what the awk script does by > >> clever use of C++ constexpr of some sorts? > > > > grep didn't find a CPP_FOR_BUILD, but I do agree it's a bit of a Pandora's > > box. > > Here's the earlier approach I mentioned. It's sufficient to skip > > line-comments, > > which is IMHO the natural way to temporarily stub out a pass. > > > > not sure why awk wanted '[[:blank:]]` (rather than `[:blank:]`, but one awk > > complained and the other didn't > > > > I'm just looking for a simple solution here, not reengineer things with > > exciting > > C++ > > > > ok?
The gen-pass-instances.awk change looks OK. I find the Makefile change just obfuscates things, I can no longer see what happens, so NACK from my side for that part. Richard. > > nathan > > > > > > > -- > Nathan Sidwell
