On Tue, Sep 03, 2019 at 09:36:23PM +1000, Jonathan Gray wrote:
> jsing has a pending diff which does that and adds a target
Cool; I'll just wait for his diff either ways, then.
> > - drop python handling in pre-build; done automatically in pre-configure
>
> This used to be required but if it works with all targets without this
> without a local 'python' symlink when building aarch64 targets with
> rk3399 sure.
port-modules(5) tells you that the shebang is replaced automatically in
pre-configure, even if your port has no real configure stage. But I've
also manually verified that the shebang is fixed after removing pre-build,
cleaning and running configure.
> > - use python 3 at build
>
> I think there may be some spots where python2 is used as it checks for
> both, but python3 is the direction all the linux people are pushing.
I'll double check the builds with Python 2 removed from the build host.
> > - introduce ${MK} helper variable: better readability, less repitition
>
> Not so sold on this as the currents lines are designed to be
> close to bsd.port.mk.
Can you elaborate? MK really is just an alias in the end to chop
excessive repetition, it still follows the aspects of bsd.port.mk(5),
that is it does the same as before, honoring all the variables.
> > - turn ${WRKSRC}/build/${BOARD} into ${WRKBUILD}/${BOARD}
> > - create build directories at once prior to build
> > - pass BL31 as make flag not environment variable to enable unified ${MK}
> >
> > Bits are still written to WRKSRC as its size increases by a few
> > megabytes after build, but I think the split is still worth it and makes
> > it easier to look into these remaining matters.
>
> Why do you want to change this?
So `make clean=build' wipes WRKBUILD only without rextracting and
patching again; keeps the build environment clean and separated.
> As this changes HOMEPAGE shouldn't there be a REVISION bump?
Indeed, thanks.
