Peter N. Lundblad wrote:
> +<p>Before a release or release candidate is offocially made public, it is
> +made available in a temporary location for committers to test and sign.
> +The point is to have the tarballs tested on more systems than that of the
> +person who rolled the release. When there are three signs from full
> +committers for each of the <tt>.tar.bz2</tt, <tt>.tar.gz</tt> and
> +<tt>.zip</tt> files, the release (candidate) can go public.</p>
Perhaps indicate somewhere in this section what is considered "signable"
testing. Ideally, this would be a 6-way test of all RA methods against
all backends (file/svn/http and bdb/fsfs), coupled with a successful run
of the bindings tests. A signature is acceptable as soon as the
commiter verifies that one particular RA/backend couple passes tests.
However, before the files go public, it is desirable that all
combinations be tested at least once, or at a pinch (if a meteor
destroys all but a few subversion commiters and we want to release) that
each component be tested once. So, according to that last rule, if one
commiter signs (file|svn|http)/bdb + bindings, all that is needed to go
public is for someone to test any RA method against FSFS.
Perhaps that last bit is a little complicated. Perhaps restrict the
text in that section of hacking.html to "6-way tests if possible, or
else as much as you can" and leave the "can we go public?" logic to the
release manager's section. However, I find it good that people who are
signing know what it will take to get a release out the door. It helps
getting signatures for the bits that haven't been tested yet.
This is why Karl (among others) is much appreciated at the signature
stage, as he plays by the full rules and does full 6-way testing, thus
letting us slackers be lazy.
- Dave.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Nov 22 23:00:38 2005