Hi Christian,
Le 30 juin 08 à 13:08, Christian Lohmaier a écrit :
[...]
Yet another reason to select good milestones, and not provide all.
Again you're basing all your argumentation on a wrong starting
point. Your idea is that Maho's build are "releases" for the public.
No, that's not my idea, but just a fact: people (not me) believe
Maho's builds are "releases". Here is the problem.
This is just not the case.
Sure: we *all* agreed Maho builds are for QA ( we discussed the point
with Pavel on IRC yesterday too btw)
Read (in the thread) my answer to Philipp, where I wrote "We all
agree the only reason to provide such milestones is for QA. "
=> So there is no doubt, andmy argumentation is not what you wrote;
Maho's builds are milestone builds for interested users/developers/
QA-folks.
Agreed (again). but the fact is, people who are not aware of QA are
downloading them.
I think Jason mails is a good summary of what I think.
He did never announce them to "users", he announces availability of
the builds on developer lists.
So everyone who downloads the builds should know that those are
only snapshots, not "releases".
The reality is different. Here is the problem.
So, I ask to find a solution, to prevent such issues in the future.
What issue? That stupid users download non-released software?
"That stupid users downloading non-released software " as you wrote,
will propagate a bad image of the mac port.
And we don't need that at all.
Then
yes: Do something about it.
Indeed: that's why I proposed to not provide all the milestones.
I don't want to support my own builds. I'm only interested with code
understanding, and provide builds including new code for some
testers, to
have the faster feedback.
But how does that conflict with providing milestone builds?
They don't. I simply answered James statement, nothing else.
So to sum up:
Maho just builds every milestone and uploads every milestone (if
time permits).
(Sun would do the same if they could do it by a snap of their
fingers/if they had the ressources, so any argument telling "Sun skips
build as well" is plain wrong.
It is not question to skip milestones, it is questions to put them
elsewhere, when they are known to have problems.
It is true that some builds have problems. These faults are a
problem with QA. (of the cws that are about to be integrated, or
just because the milestones/the combination of the cws
is not tested as well as it should).
This is not Maho's fault as QA is not his job.
Just about the job, isn't it Maho who does manage QA ?
He is a provider of builds.
Presented like that I disagree. I'd like to see at least "
intelligent provider of builds " in your mind.
Means: able to decide to not provide a build. Nobody will blame him.
He spends his time and ressources on doing builds, not development,
not QA.
As QA lead, this is not the best example :-)
It is true that users use development builds (the milestones or
beta) and find bugs. See above: This is a problem with QA. Those
problems should not have
been introduced in the milestone in the first place. (but this is
only theory, you cannot test each and every feature in all
environments to catch all bugs).
Sure but when a bug leading to instant and random crashes is well
known, and needs time before the fix to be integrated, it sounds
reasonable to stop providing public milestones in meantime, at least
until the fix has been integrated.
That's why it is even more important that people do find those bugs.
That's what we did: I announced the crashes on IRC before WWDC. And
there a bit after .. and just now, some people seems to discover the
issue.
It also shows that communication of the Mac port could be better:
Why didn't the users downloading those build knew about those issues
already?
Cannot be bad.
This is not Maho's fault as he doesn't do any of it. He doesn't act
as a developer working on the code, he is not involved in cws process,
does not do QA, doesn't do marketing/advertise his builds to endusers.
Warning : troll inside ;-)
But he also wrote before: He doesn't follow development and doesn't
want to go through the mail-archives himself.
This is not normal, and I contest here : when one has the
responsability of doing public builds, the minimal requirement is to
keep an eye upstream.
You cannot tell a builder that a fix is a must-have, if the
developers didn't think it was worth fixing it in a cws as fast as
possible)
The point is a builder must have an eye on what happens upstream, and
be able to hear the warnings we send.
Using other words: something is in open loop mode here ...
Regards,
Eric
--
qɔᴉɹə
