[opensource-dev] Review Request: Put the viewer version into marker files, and report errors only when the version matches

[Ricky]

Well, all I have is anecdotal evidence from my own family of four, everyone
an SL user: we all keep multiple versions of the SL viewer, and even a few
TPVs, on our machines.  If one viewer crashes it's more often that the
person it crashes on will relog in another version or viewer - just to
avoid the perceived, if possibly inaccurate, cause of the crash. We may
switch viewer versions on the second or third crash, depending on the
individual, but the effect is the same.

Like I said, it's anecdotal. Also I can hardly take my family as typical of
SL users, but it's the only population of SL users I can observe.  To me it
seems that if it's both possible and feasible, the viewer should report and
attempt to clean up after a crash - even if it is not its own.  This would
eliminate any deviation from appearing from this source.

Either that, or the marker files should be filename keyed to a version, in
which case the next time that particular version launches it will be able
to do its own report and cleanup.  However this has the negative that the
viewer version may not ever be launched again, leaving the keyed marker
files cluttering the filesystem.  Not a good picture on this logical path
either - though it could be kept sane by having any markers older than X be
automatically removed - if they are too old it matters little.  This path
of logic does keep the deviation very minimal, though not eliminated: only
those viewer versions that crash and are never re-launched get their report
missed.

Ricky
Cron Stardust

On Thu, Nov 1, 2012 at 10:34 AM, Argent <secret.argent at gmail.com> wrote:

> Wouldn't this only be an issue, normally, for people switching from one
> the viewer to another, where there was an leftover crash filer from the
> older viewer? Doesn't seem like leaving these events out would cause
> significant deviation.
>
>
> On Thu, Nov 1, 2012 at 11:09 AM, Ricky <kf6kjg at gmail.com> wrote:
>
>> Wouldn't it be better if the crash could be reported anyway - just
>> marking the correct version?  With this in place at least crashes won't be
>> misreported, but they also will be not reported to the servers at all,
>> causing statistical deviation - what I believe is the core idea to be
>> fixed.  More comments in the JIRA,
>>
>> Ricky
>> Cron Stardust
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/opensource-dev/attachments/20121101/3dc1f6f8/attachment.htm