RE: 1.3.5

Stefan Küng wrote:

> I seriously doubt that an RC would have saved us here.

Maybe

> Not many people
> actually use/install an RC.

Not proven.

> And those who do most likely don't commit
> single files but commit from the parent folder.

Not proven.

> So that bug wouldn't
> have shown up.

Not proven.

Sorry, this is just a concatenation of assumptions, made up to justify that
you don't want to create RCs. If you don't want create RCs, for intermediate
releases, don't do it (see below).

> Also don't forget that we currently have a 'problem':
> all those who already tried an 1.4.x nightly build won't be able to
> use an 1.3.x version anymore without checking out all projects again,
> because once you use an 1.4.x version, the working copy format gets
> changed and can't be used anymore with previous clients.

That doesn't have anything to do with people testing release candidates or
nightlies in general, does it?
We want people who normally don't install nightlies (=additional testers) to
test our RCs. These won't have 1.4.x installed anyway.

>> Yes, but we want a lot of people to test our software. IMO the
>> nightlies are sort of well hidden from 'normal' uses, even though we
>> mention them at several places. Nightlies also have a very
>> developer-alpha-quality smell about them, which surely prevents a
>> lot people from installing them.
>
> The same applies to an RC. Maybe a little less, but I guess not much
> more people are willing to use/test an RC than are willing to test
> nightly builds.

I don't think that this applies to RCs, because that's a build that the
developers deem release-worthy, so it's more likely that non-devs will
install it.

How about a poll on SF:
Would you test RCs?
( ) never
( ) maybe
( ) of course

> Downloading an RC or nightly does *not* mean it also gets tested.

Hrmpf, How many people download software just for the sake of downloading
it? Once it's installed, it's under test.

> So
> stats about the downloads only can give you a very rough estimate on
> testing, not more.

Agreed, but 1000 downloads of a RC within two weeks without complaints
reported back to the dev list tell us more than an unknown number of
downloads of nightly builds.

> While I agree that this information would be
> useful,
...
> I really don't like the idea of pushing out RC's for
> intermediate releases. I'm the one who has to do it, and I'd rather
> spend my time with other things...

That's about the only argument you presented that I can (or have to) accept.

I checked our release history since 1.0 (~ two years) and I see at least
eleven obvious quick bugfix releases (withing ten days of a previous
release). During that period we made four major, 27 minor releases and had
three RCs up for testing.

That means that 11/31 = 35% of our official releases were so broken that we
had to fix them quickly and I am convinced that we could get below 10%
broken releases by publishing RCs.

We could restrict RCs to be without language packs. That would save some
build time and a lot of upload time.

If uploading, adjusting the web sites and announcing everything everywhere
is the part of the work that you don't like, I'll gladly do it. I can't
build TortoiseSVN, but I can take care of the administrative part.

Cheers
- Lübbe

--
       ___
  oo  // \\      "De Chelonian Mobile"
 (_,\/ \_/ \     TortoiseSVN
   \ \_/_\_/>    The coolest Interface to (Sub)Version Control
   /_/   \_\     http://tortoisesvn.tigris.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tortoisesvn.tigris.org
For additional commands, e-mail: dev-help@tortoisesvn.tigris.org