Re: [VOTE] Adjust voting required for alpha/beta releases.

Ben Reser wrote on Mon, Apr 07, 2014 at 14:57:44 -0600:
> On 4/7/14, 2:08 PM, Justin Erenkrantz wrote:
> > To be pedantic, your last phrase could be mis-interpreted - I'm pretty
> > sure that you mean that someone can indeed test with multiple
> > platforms; not that "each person only has one vote".
>
> Yes that's what I mean. You can't vote twice for Windows. But you could vote
> for Windows and Unix.
>
> > Perhaps this would be easier if we just codified this into a script?
> > didthevotepass.com.
>
> That's difficult because there's no standard way people vote. I've often
> wished we had something like that. I often use the release.py check-sigs as a
> check, but you can't use the signature count for anything useful. For one
> thing the signature doesn't tell us what platform was tested (people often sign
> the zip files when they only voted for unix). Another problem is you may have
> multiple signatures from the same person on the same file (either as a mistake
> as has happened in the past or due to signing with multiple keys).
>
> So I think we'd have to agree to use a given template for release voting.

Sounds like check-sigs does what you want except for there things: it
doesn't uniquify multiple signatures by a single signer; its output
isn't oriented at that particular task; and it doesn't partition the
signers into unix/windows.

The first is easy to solve (eg, since PGP keys are bound to email
addresses). The second is trivial. The third I would suggest to solve
by an approximation: simply assume that all signatures from Bert or Ivan
are for windows, that all signatures from Ben and Stefan Sperling are
for unix, etc. Sure, there may be exceptions (Bert may sign for Unix
and Ben for Windows), but if the script did the approximate counts, the
RM would need only to remember the exception.

See attached. It works like this:

    % wget https://www.apache.org/dist/subversion/subversion-1.8.8.{tar.gz,tar.bz2,zip}.asc
    % ./sigcounter.sh 1.8.8
    unknown: 1 (stefan2)
    unix: 3 (breser, brane, philip)
    windows: 2 (jcorvel, rhuijben)
    
    Remember: while these counts are accurate, the division into windows/unix may not be.
    %

It should probably have been written as a release.py subcommand, but I
finished writing it before I realized that :(

HTH

Daniel