On 5/23/2010 00:50, wy...@volny.cz wrote:
Wolfram Sang<wolf...@the-dreams.de> wrote:
I noticed that Wylda uses "--private keyword: bisected"
when appropriate. IMHO this could be useful as a real keyword, e.g.
if you search for bugs you'd like to try tackling. Has this been
considered already?
There is a keyword 'regression', that should be enough.
Hi, it depends, if regressions are considered as bugs of top priorites.
I would wish, if they could be considered, because killing them early
means, that quality of wine is not getting worse during the ongoing
development.
Imagine, that someone used to have a working game in 1.1.42 and his distro
offers him upgrade to 1.1.44. He goes ahead with upgrade and apps stops
working. He fills in a bug - UNCONFIRMED& REGRESSION keyword. Then
someone other have exactly the same problem, so confirms that and we
have the bug with NEW& REGRESSION. So something serious, which should
attract dev's attention. But when developer begins to work on that,
he spends some time with getting the demo, verifies it's behaviour and
then finds out, that the problem is wrong Ubuntu package. So his time
with this bug was wasted.
I believe developer's attention doesn't depend on bug state
(confirmed/uncofirmed) at all.
The flow is to try to fix regressions caused by your own patches if it's
not too complicated and there's
nothing more serious to do at this point. If it's too complicated to fix
fast and a lot of apps (potentially) affected
a change is reverted.
Compare this with regressions which are bisected (aka "bug served on
silver tray"), so saves a lot of time. And of course such a keyword
would help in dev's triage what to fix first.
Actually a "regression" keyword supposes to mean exactly the same. It
just happens that
it's added every time someone decided to add it. IMO it should be added
only when regression
test results are available but this isn't going to happen of course for
obvious reasons (some reporters don't
bother to respond in months). If no test was performed a developer will
see a report anyway,
searching for a module of interest.
Of course this has a sence and helps only if _truly_confirmed_ regressions
are taken seriously, i.e. fixed as first. But Wine is still not in this
state.
