pkgbase: Add incremental package

[manu]

Summary

This adds a new target update-packages which will create the new packages compared to the last run.
This is how to use it:

At this point we cut a release

$ make buildworld ...
$ make buildkernel
$ make packages

1. There is now a PKG_VERSION directory with latest link pointing to it
2. Distribute the packages to server

$ something something that update the source tree
$ make buildworld ...
$ make buildkernel
$ make update-packages
You know have a PKG_VERSION directory in the REPODIR and latest link pointing to it.
In PKG_VERSION dir only the packages which differs from the latest run are named PKG_VERSION, otherwise the old packages are there.

The process is :
Build the new packages in the PKG_VERSION directory
Compare the internal data with the PKG_VERSION_FROM version. The comparison is done by checking the hash
of the files. By default PKG_VERSION_FROM is set to what the latest link points to.
If the old and new version matches, we rm the new package and cp the old one.

[0mp]

Makefile.inc1

1903

Why is it cd ${.CURDIR} && make ... instead of make -C ${.CURDIR}?

[manu]

Makefile.inc1

1903

Because it's done like that everywhere in this file.

[0mp]

Makefile.inc1

1903

Fair enough! Thanks!

[0mp]

Makefile.inc1

1886

This echo + subshell here are probably unnecessary.

[manu]

Simplify update no need for PKG_TIMESTAMP and kernel works.

[emaste]

Makefile.inc1

1880

or commit that part first

[manu]

Regen after r364103

[emaste]

If a script (post-install, post-deinstall ...) changes we don't detect it

I wonder if we should print a message to stderr indicating that incremental packages are experimental / have some missing functionality e.g.

[emaste]

Looks fine to me.

[manu]

In D25984#579165, @emaste wrote:

If a script (post-install, post-deinstall ...) changes we don't detect it

I wonder if we should print a message to stderr indicating that incremental packages are experimental / have some missing functionality e.g.

When we generate them you mean ?
If we add one at the end maybe people will read it but I'm unsure :)

[emaste]

When we generate them you mean ?

Yeah, that was my thought, although now that you ask perhaps a pkg message would be good - let users know there might be some issues.

[manu]

I'll update this review so we use pkg query '%X' from pkg 1.15 to get the internal checksum of the package and check if we need to update it.

[manu]

Use pkg query '%X' to have the internal package checksum, any changes to a package except the version will produce a different hash.

[manu]

Forgot to uncomment a line, we need the stage-package dep.