Re: [libreoffice-users] Re: fixing as a service ...

Hi there,

On Fri, 2014-10-03 at 08:42 -0400, Tanstaafl wrote:
> Setting up an entire automated build/test environment is simply way
> beyond my capability. The best I can do is download installable builds
> and simply test them by using them.

        Sounds good :-) that's why we have daily (and sometimes hourly)
snapshots of master - our latest development version available: to make
it easy for people to chase the latest features & help developers with
QA as they implement features.

> Meaning - a developer should be responsible for the code they write,
> including fixing bugs when they are found *without* resorting to
> "'patches welcome' or 'pay someone to fix it'"

        So the "when they are found" bit is of course the key =) No-one intends
to introduce new bugs (well, ok some people argue that a new feature is
a bug but this is fairly rare). They happen as an accidental side-effect
of feature development, and/or fixing other bugs - which can often be
entirely un-related and somewhere completely different.

        However the cost of a fix - in developer time goes dramatically up the
longer that it takes for the bug report to come in. I know some QA /
development guys who work in a quite tight loop together as a new
feature lands and is polished: that is -by-far- the most effective way
to provide feedback. A bug a year later far more expensive to fix.

        Failing that, we provide great tools to try to close the gap between:
"it broke" and "who is responsible / interested ?" - which is 'bibisect'
- which allows you to go back and run historic versions to find out at
what point (and almost 'by whom') the issue was introduced. Most
developers when you have bibisected, and CC'd them to point at the patch
feel responsible and jump in to fix the bug.

        Of course - some reports (and I assume it's not you doing this) at this
point (if you're lucky - often this happens when they find a bug at
all ;-) get a sense of outrage & entitlement and start shouting at the
embarrassed developer, demanding their work is reverted, demanding
processes to stop XYZ committing ABC until they (personally) are happy
etc. ;-) this piece of the puzzle tends to have a predictably
counter-productive outcome =) It is worth working hard to not
(accidentally) look like that interaction.

> If the vast majority of the developers don't agree with this principle,
> and in fact believe that they should be able to just commit code for
> something, then go on their merry way and/or respond with the "'patches
> welcome' or 'pay someone to fix it'" responses

        Given a generic bug reported loong after the development took place:
(ultimately) all bugs are caused by some developer either by action or
omission - I think that's a reasonable approach. The length of time it
takes to file it is -usually- a sign of its relative importance vs. the
other 6000+ open issues =)

        Of course, if some reporter wants to help pin-point the regression to a
specific commit, and does a chunk of work to help interest a developer
in fixing it then that might work well too - that reduces the cost of a
fix to (hopefully) a simpler spare-time task.

        From 10k feet though looking with an ecosystem perspective it is clear
that we have to stack the economics here to make things sustainable.
Expecting a (perhaps paid) developer to provide effectively indefinite
free support, to all users of every feature they ever implemented across
all other changes from others that may impact it is really not
realistic. If it was, we'd still have Sun around to fix the umpteen open
regressions vs. OpenOffice 2.0 ;-)

        I'm sure that's a familiar model to someone who sells services. If you
do a for-a-fee install a deployment of say 60x Windows version N
machines - as a one-off without an ongoing support contract. You are not
surprised to hear the customer phoning to endlessly complain about XYZ
change not working, and ... expecting that your time is free - and you
can 'just' help them with a given problem for free etc. ;-) [ perhaps
you don't ?] but the situation is reasonably analogous. Normally you
solve these problems by having a short period: "Report any issues inside
a month" - after which, they have to pay for support. Not a perfect
analaogy, but ... in this case no-one is paying at all it seems ;-)

        Hope that helps,

        ATB,

                Michael.

-- 
 michael.meeks@collabora.com  <><, Pseudo Engineer, itinerant idiot


-- 
To unsubscribe e-mail to: users+unsubscribe@global.libreoffice.org
Problems? http://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: http://wiki.documentfoundation.org/Netiquette
List archive: http://listarchives.libreoffice.org/global/users/
All messages sent to this list will be publicly archived and cannot be deleted