[boost] Re: Fast-test-cycle Proposal [was 1.31.0 Progress (or lack thereof)]

From: Bronek Kozicki (brok_at_[hidden])
Date: 2004-01-26 12:46:08

On Mon, 26 Jan 2004 11:02:37 -0500, Beman Dawes wrote:

> * Keep the reporting very simple (to speed cycle and upload times)
> * Run the quick-test very often - once an hour or more during heavy 
> activity. (Presumably on Win32, because GCC on Win32 is also indicative of 
> GCC results on other platforms.)
> * In the future, if quick-test is found to be useful, dedicate a machine to 
> run quick-test continuously.

I wonder if this could be mixed with two-stage commit, ie. all changes
to CVS are first automatically tested using "machinery" you just
described, and then (after one hour or sooner) commited to main CVS.
There might be more tests (ie. each library might have own quick-test)
with two assumptions:
* running each test must be cheap in terms of CPU usage, so running all
of them every hour is possible
* some changes will trigger extra tests, limited in scope to library
being modified (directly).

It's just an idea.

> Would this be useful and worthwhile?

Definitely. I think this is first complete (well, close to) solution to
problem "most developers have access to limited number of compilers".

B.