svn commit: r415078 - in head: . Mk

[Alexey Dokuchaev]

On Fri, May 13, 2016 at 08:28:37PM +0200, Baptiste Daroussin wrote:
> On Fri, May 13, 2016 at 04:01:51PM +0000, Alexey Dokuchaev wrote:
> > On Thu, May 12, 2016 at 06:20:27PM +0000, Ed Maste wrote:
> > > New Revision: 415078
> > > URL: https://svnweb.freebsd.org/changeset/ports/415078
> > > 
> > > Log:
> > >   Record TIMESTAMP in make makesum
> > >   
> > >   This is intended to support development and prototyping for ports
> > >   reproducible build efforts which require some concept of a "last
> > >   updated" time. It is being committed now so that timestamp entries
> > >   will be populated "for free" as port distfiles updates happen.
> > 
> > Why does it have to be part of distinfo rather than obtained from keywords
> > or "svn info distinfo | grep ..."?  (The latter might seem like it could
> > slow things down a bit, but 1) it's unclear if it would make a significant
> > difference without actual tests, and 2) polluting distinfo looks worse, to
> > me at least.)
> 
> First the svn keyword date will be updated too often (each cosmetic change
> in a makefile) resulting in making reproducible builds less efficient.
> 
> Second the ports tree is available in plenty of forms where the keywords
> are not expanded: git, hg etc. breaking as such the reproducible builds[.]

(*sigh* I knew that I shouldn't have brought keywords into the argument.)

Anyway, why this information can't be fetched from the metadata?  Why does
it have to be tied to distinfo?  What about patch files, sed(1) calls in
the Makefile, etc.?  Is there a place we can read more about this whole
reproducible builds thing?

./danfe