[sldev] Bug triage on Friday

[Rob Lanphier]

Hi Alissa,

Comments inline....


On 4/13/07 8:40 AM, Alissa Sabre wrote:
>> I've been talking with folks at Linden Lab about how to deal with bugs
>> filed in the public JIRA.  A group decided a good next step would be to
>> have a bug triage, where we start working through the issues, and
>> decided to co-opt my next in-world office hour to do it.  Friday, 3pm
>> PDT at Grasmere at my cubicle.
>>     
>
> I will not be there.  The following is a list of my concerns on the
> issue.
>
> 1. in-world bug reports and JIRA
>
> Are in-world bug reports listed on JIRA?  I'm afraid not.  Are those
> submitted through support phone calls?

In-world reports aren't.  We're considering changing it so that they
are, but until we have a good process for dealing with the issues that
we're getting now, it's probably not smart for us to unleash the
firehose of in-world reports.

>   How about bugs found by Linden
> employee?

Not yet.  There's an exporter that someone (probably me) needs to write
yet, which would push items out to the public JIRA in the same way that
they currently get pushed to this page:
http://secondlife.com/support/known-issues.php

>   I believe the bug triage should cover *all* bugs.  If we
> use JIRA that lists bugs from Open source developpers only ,the
> resulting ordering will not work very well.
>   

This meeting isn't a holistic look at all of the open bugs that have
been filed against Second Life.  There are just too many.  This is an
attempt to better cope with an existing pile of bugs that we have.

> 2. Repro counts
>
> I don't understand why the number of repro cases is important.  A bug
> is a bug, even if it hits only one condition.  The frequency of
> hitting the bug or number of people affected by it may count, but the
> number of cases is different from that.
>
> I believe one valid repro is enough for a bug report to be evaluated.
>   

We have a /lot/ of bugs to get through.  We have to set some criteria,
even if it's not the best criteria.  Otherwise, we can waste a lot of
time fixing bugs that only affect one person while letting bugs that
affect a lot of people go unnoticed.

If a patch is attached, that makes a big difference.  It still not
automatic, because it requires evaluation and testing, but a
clearly-written patch which fixes an obvious defect would probably not
need to clear the three repro hurdle.

>> please add whatever debug information you can, and vote
>> for the issues that you think are most important.
>>     
>
> Hmm.  I'm not sure what you mean by debug information.
>
> Do yo need something more than repro?  Do you want memory dump or
> something?
>   

Depends on the bug.  Sometimes a clear repro is enough.  Other times,
the system information in Help->About is really important, especially if
it's likely that the hardware or video drivers are playing a role. 
Stack traces are always handy in the event of a crash.

(I probably could have just said "see Jesse's response")

Rob

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20070413/573e9db6/signature.pgp