cvs commit: src/share/mk bsd.cpu.mk

[M. Warner Losh]

In message: <44C8C0C8.4090906 at fs.ei.tum.de>
            "Simon 'corecode' Schubert" <corecode at fs.ei.tum.de> writes:
: M. Warner Losh wrote:
: > We need to find some way to accomplish having local changes in the
: > tree that doesn't screw new imports.
:
: Maybe I am missing something important here, but where is the
: problem with modifying files and taking them off the vendor branch?

Future integrations are harder.  As more and more files come off the
vendor branch, it becomes increasingly difficult to use CVS to manage
the delta from the point where it came off the vendor branch.  CVS'
vendor branches are extremely weak.

: Situation 1: the vendor sources don't include the feature that is
: implemented in the local changes.  -> The file needs to stay off the
: vendor branch.

That makes future merges difficult.

: Situation 2: the vendor sources integrated the modifications and
: these sources and the local (off-vendor-branch) sources are the
: same.  -> just put the file on the vendor branch again (basically
: cvs admin -b1.1.1)

That breaks checkout by date, because the change of default branch
isn't recorded in CVS.

: another interim solution is the way we are handling it for
: DragonFly: have patches somewhere else and apply them automatically
: during the build process (we have a bsd.patch.mk for that.  just set
: SRCS+= file1.c.patch file2.c.patch)

This is crap.  That's what a SCM is supposed to do for you.  Adding
patches outside of the SCM is like putting a comdom on after sex.  The
results might be more enjoyable for some aspects, but worse for
others.

Warner