Rob Weir wrote:
On Thu, Aug 9, 2012 at 12:16 PM, Ariel Constenla-Haile
wrote:
Please don't mix frequent building on regular basis with what has been
happening here the last weeks. There have been two, three, RC build
proposal in the same week. The most notorious is the build announced on
Monday and the re build on the next day. IMHO this is mainly not fair
for volunteers doing QA work. ...
It should be possible to do this:
1) Take the first RC of the series and stick with it for a week or
more. Test it as completely as you want. So long as a test is not
blocked, you can report all bugs against that RC.
Of course this works, but I can understand Ariel's concerns since it's
hard, for those not following development closely, to identify builds
(RC1, RC2 and RC3 tend to be better names than r1369843 and such) and
understand which one comes first, and why a given version is available
for certain operating system and not for others (since we don't publish
all build simultaneously).
I agree that it would be difficult for a tester to take every RC and
interrupt their work to install and restart testing on it.
Tests like those in (Italian, but the English version is quite similar)
http://wiki.services.openoffice.org/wiki/QA/TestCases//it#Test_Generale_su_AOO_3.4.1
tend to take one hour or so for each tester, so it's unlikely that
someone sees their work interrupted.
But still, a more regular cycle (like publishing a Release Candidate per
week and having a kind of schedule for uploading builds and receiving
feedback) would allow smoother operations. This will become easier when
the project graduates: at that point we will only need one vote and 3
days, as opposed to 2 votes and 6 days like now.
Regards,
Andrea.
