Subject: Re: [boost] [Boost-testing] Release branch purge policy request
From: Jim Bell (Jim_at_[hidden])
Date: 2015-03-14 16:00:44
On 2015-03-13 12:17 PM, Rene Rivera wrote:
> On Wed, Mar 11, 2015 at 11:29 AM, Jim Bell <Jim_at_[hidden]
> <mailto:Jim_at_[hidden]>> wrote:
> Just wanted to put this out there again since it seems to have
> flown under the radar...
> I suggest that, for the release branch, we lengthen the 15-day
> purge to
> 90 days or more.
> My immediate answer is no. At minimum because..
> Those looking to that matrix are determining where the last boost
> release works or breaks for their platform. A different audience than
> I don't see how extending the purge would address that. The master
> branch testing is for the *next* release. Hence it doesn't say
> anything about previous release (not even the last release).
While I agree that the master branch is for the next release, I'd
dispute strongly that it doesn't say "anything" about previous
releases. It says just about everything for those libraries which
haven't undergone revision, and still says a lot--or at least
something--for the rest. My personal experience has been finding the bug
I suspected in a library by looking at the matrix, sometimes with the
library a couple revs out of date.
Just guessing, but I'd say it's a pretty good bet statistically that a
yellow mark on the master regression will still be yellow after 90 days
When a boost user looks at an out-of-date matrix, what does (s)he see
for his/her platform? Mostly green (hopefully!), some yellow that's been
fixed (maybe!), some that hasn't. That's more valuable than seeing
nothing at all.
(Ok, ok, 90 days is pretty long, but still...)
Cross-posting with the main list since (I say) the audience is wider.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk