On Wed, 22 Jan 2003, Luke Mewburn wrote:
> I agree that `-R RELEASEDIR' hasn't got the most optimal behaviour.
> However, it may be too late to change this (too many people might rely
> upon the behaviour of -R).  This is why I added `-V var=value', to allow
> arbitrary variable setting (such as RELEASEDIR) without invoking the
> action that -R (et al) does.

Ah!  Time to change my wrapper scripts to use -V.

I don't think it's too late for an incompatible change to the behaviour
of build.sh.  I suspect that very few people use it in any way other
than by hand.

>   | My second biggest gripe (which might be a subset of the above) is about
>   | the fact that the "-i INSTALLWORLDDIR" option implies making a release,
>   | instead of just installing an already-built release.  OK, you can force
>   | it to be sensible by saying "-b -i INSTALLWORLDDIR" instead of just "-i
>   | INSTALLWORLDDIR", but it's easy to forget.
> 
> The documentation `-i idir' does not imply "make a release" to me. 

My use of the phrase "making a release" was probably wrong.  I was
referring to the behaviour that's equivalent to "make build", which is
the default if you use build.sh's "-i" option without any other options.

> All `-i idir' does is installs from DESTDIR to idir after the
> buildtarget (which defaults to "build") completes.

That's the problem.  It would be less confusing if the buildtarget
defaulted to "do nothing" (or the do_buildsystem variable in build.sh
defaulted to false) and if there were a separate command line option to
set do_buildsystem=true and buildtarget=build.  Unfortunately, both "-b"
and "-B" options are already taken.

One way forward would be to add some action args (like "bmake", "tools",
"build", "release", "install") so we could do stuff like

	build.sh [-options] build install

to build and install (like "-i idir" today), or

	build.sh [-options] install

to just install (like "-b -i idir" today).

--apb (Alan Barrett)