[Libreoffice] Number of days for testing RC

[Andrea Pescetti]

On 10/12/2010 Petr Mladek wrote:
> OOo has only 5 days, see
> http://wiki.services.openoffice.org/wiki/Release_criteria#Stopper_issues
> How is it there with building and approving the national builds?

As shown in the page you link, it's "5 days including a weekend", so the
constraints proposed by Jean-Baptiste are still respected.

> My opinion is that 10 days might make sense for the first release
> candidate but it is too much for the next release candidates. Otherwise,
> we would do only 2 release candidates during one month and the release
> would be newerending story.

I agree that 2 release candidates a month are too few. But you needn't
reach 8 RCs every time! I mean, OpenOffice.org 3.3 RCs were produced
with a 3-7 days delay between each other and this was wrong and
confusing; a longer time between RCs allows to fix more bugs, provided
that testing continues. So, if a Release Candidates fixes several
stopper bugs, it can make sense to have fewer (and, in this case, you
will hopefully never reach 8 RCs).

> 95% of the functionality is language
> independent, so most of the testing can be shared and distributed.

Mostly, indeed. But still there are subtle bugs (fixed in OOo 3.3-RC
Italian: a wrongly translated style name that caused a name clash with
another style and caused bizarre behavior only when that specific style
was used and only in the Italian version) that suggest that, to be safe
from the start, a more thorough testing would be needed. On the other
hand, this would be the kind of bugs you would probably address in
LibreOffice with very frequent bug-fix releases, thus preferring
"on-the-field" QA by all LibreOffice 3.3 users to the preliminary QA by
volunteers.

Regards,
  Andrea.