> On Mon, Aug 27, 2001 at 10:53:22PM +0200, Arthur Bergman wrote:
> > I would suggest that smoke scripts all do make regen_headers before smoking not to
>report trivial header file mixups. Or because you got a snapshot inbetween a patch
>and a make regen_header submit to the repository.
>
> Aren't the targets in the Makefile supposed to ensure that headers get
> regenerated in the case where they are out of date w.r.t. the scripts that
> generate them?
That might be the case, but we still need to do a make regen_headers and submit these
into the source repository. And taken from the emails it doesn't seem to do that for
people.
> [I seem to remember adding an extra un-needed dependency to one to con gnu
> make into running one of the regen scripts before another, so that it
> the date stamps ended up in "order" so that make didn't end up re-running
> one of the regen scripts next make and consequentially completely rebuilding
> everything.
>
> ie if a regen headers was necessary
>
> make
> make test
>
> would re-make some headers on make test, and then make all the object files
> again. This con cured the remake.
>
> Sorry for this digression]
>
> If the current dependencies can't always detect the need to make
> regen_headers, can we fix them rather than working round it in the smoke
> test?
Indeed that would be a good solution. However I can see potential problems with the
dates getting mixedup on downloads.
I still would like to smoke script to a make regen_headers; make; make test; since
catching the fact that the headers haven't been checked into p4 isn't realy what
smokers are for.
> Nicholas Clark
>
Arthur
