rev-upgrade and a database of binary information

[Dan Ports]

On Tue, Sep 18, 2012 at 12:58:50AM +0200, Clemens Lang wrote:
> Yes, we could copy the information in the binary header into a database
> and run the testing there, marking packages as broken if a package that
> has been rebuilt no longer satisfies a dependency.

Oh... actually, I'd kinda figured that was how rev-upgrade already
worked. :-)

Now, having actually looked at the code, I see that it in fact scans
all active binaries each time it runs. I suppose that explains why it
takes so long to run (about 3 minutes on my system, with about 7500
binaries). 

Aside from being useful on the buildbot (where there are lots of
archives that aren't active), having this database might be useful for
regular installations. It would certainly speed things up if we could
just identify the files that linked against a library that's been
upgraded or deactivated.

I suppose that looking for broken ports this way would fail to detect
some breakages that rev-upgrade currently finds, but might be
worthwhile for a faster check after upgrade.

> Sounds doable, do you want to give it a shot? ;)

If I could find some more free time...

Dan

-- 
Dan R. K. Ports                UW CSE                http://drkp.net/