Re: Defaulting USE_PKG_ADMIN_DIGEST to yes

[Joerg Sonnenberger <joerg%bec.de@localhost>]

Am Sat, May 13, 2023 at 07:01:30AM -0400 schrieb Greg Troxel:
> Thomas Klausner <wiz%gatalith.at@localhost> writes:
> 
> > For switching away from CVS, joerg has prepared
> > USE_PKG_ADMIN_DIGEST=yes to compare file checksums using 'pkg_admin
> > digest' instead of using the RCS Ids for identifying when a package
> > needs to be rebuilt because the pkgsrc files have changed.
> >
> > I think we can turn this on independent of the actual CVS migration,
> > and I'd like to do this soon.
> 
> Not an objection, but I'd like to understand better.  I think this is
> about pbulk (and only pbulk), and that it stores this information in
> some database, but I'm having to guess to think that.  Can you explain
> the scope of the change?

See pkgdb/*/+BUILD_VERSION, the data is recorded to identify the state
of the pkgsrc tree at the time of the build. So far we used to record
the RCS ID, e.g. the last committed state. This fails for local
modifications of course and for any system that doesn't update the file
content on checkout or commit. With the USE_PKG_ADMIN_DIGEST flag set,
we record a hash of the files instead. This is good enough (and actually
better) for purposes like pbulk, but makes it a bit more annoying to
figure out what state the tree was in.

I've been doing bulk builds with it for a while, the only possible down
side is that it requires a newer pkg_install.

Joerg