[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Monotone-devel] Thoughts about 'testresult'...

From: Richard Levitte - VMS Whacker
Subject: Re: [Monotone-devel] Thoughts about 'testresult'...
Date: Thu, 21 Apr 2005 17:10:10 +0200 (CEST)

In message <address@hidden> on 21 Apr 2005 07:49:11 -0700, Emile Snyder 
<address@hidden> said:

esnyder> On Thu, 2005-04-21 at 00:55, Richard Levitte - VMS Whacker wrote:
esnyder> ...snip...
esnyder> > I'm proposing to change that behavior by making things more
esnyder> > explicit.  Instead of getting implicitely stuck on
esnyder> > revisions with true testresult certs, there should be a way
esnyder> > to explicitely ask to get only those.  As fat as I can
esnyder> > tell, the simplest way would be with another selector.
esnyder> > Since we're talking about results, I propose 'r', with the
esnyder> > following format:
esnyder> > 
esnyder> >   r:keyid
esnyder> So, what commands would use this, and how would the final
esnyder> behavior differ?

'monotone update' and any other command that can use selectors.
Basically, it would search for the latest occurance of a true
testresult cert signed by keyid.  It's a little like searching for a
tag, just a different cert...

Basically, the real difference is to make certain things more visible
and explicit, especially when the default checks in std_hooks.lua
makes for unexpected behavior depending on luck :-).

esnyder> Does update now never update you to a rev with a true
esnyder> testcert *unless* you give it this selector?  That seems
esnyder> counter-intuitive in the other direction.

Why?  Basically, whoever wants to go for revisions where I confirm
that 'make check' works for me would simply do this:

        monotone update r:address@hidden

Nothing stops you from getting that revision anyway if that happens to
be the (single) head, AND it wouldn't get us into the surprising
effect we had a few days ago.

An advantage is also that you get to ask for specific tests (specific
keyids) instead of getting to trust anyone's test, more or less

esnyder> > keyid would simply be the key with which the testresult
esnyder> > cert has been signed, since that key is supposed to
esnyder> > represent the test that has been successfully performed.
esnyder> > 
esnyder> > Comments?  Is this a completely wacky idea, or would people
esnyder> > like this (I would, most obviously)?
esnyder> r: makes me thing revision, is t: taken?

t is for tag.  Otherwise, I'd agree with you.


Please consider sponsoring my work on free software.
See for details.

Richard Levitte                         address@hidden

"When I became a man I put away childish things, including
 the fear of childishness and the desire to be very grown up."
                                                -- C.S. Lewis

reply via email to

[Prev in Thread] Current Thread [Next in Thread]