Re: obscure bugs

[Gregory Leblanc <gleblanc cu-portland edu>]

On 24 Jun 2001 11:52:42 +0100, Telsa Gwynne wrote:
> On Fri, Jun 22, 2001 at 08:22:40AM -0600 or thereabouts, John Fleck
wrote:
> > In trying to work through old bugs and get them categorized, I'm
> > running into many that fit a common pattern, and I'm wondering what
to
> > do with them.
> > 
> > They're typically old - GNOME 1.2 or before - and consist of a
> > description to the effect of "I don't know", or "it crashed,"
followed
> > by a backtrace - no description of what crashed or what was
happening
> > prior to the crash.
> > 
> > Is it reasonable to just mark these "INCOMPLETE" and be done with
> > them, or is there some value in saving them and putting them
> > somewhere, and if so, where?
> 
> I have wondered about this. I started off by making great duplicate
> chains of them. But then I realised that every time you do that, 
> all the people who own other bugs in the same duplicate chain get
> email too. So you can easily end up with thirty mails when you add
> one to a long chain; and the person who sent in the one they all
> got added to gets thirty emails, one for each addition. 

I'm all for making the duplicate chains.  If somebody decides to look at
fixing these bugs, we want them to be able to find all of the
information that we have, with a minimum of hunting.  If we don't mark
duplicates, then there's no chance that they'll be able to find all of
the information, nor even most of it.  

Oh, one note...  If you change a bug with 30 dupes, 30 emails are sent
out, but only 1 email per person.  Hopefully if you get a line of dupes
that long, somebody will consider it an important bug to fix, and it
won't exist for too long.  Once the bug with the long chain of
duplicates has been closed, new bugs can simply be marked "fixed", until
people upgrade.

[much verbage snipped]

> Or the same without the bug-buddy references. And I mark it as
incomplete.
> I do worry that I'll do that with the one message that has the
> critically-useful stacktrace, but there are several patterns, as you
> say. I see two a lot: the really old bug-buddy one with less than a
> dozen old lines all of "?? in ()". And another which is colossally
> long, all the functions are gtk_-something, and it appears to cycle
> through them and then stop. Those, I close as INCOMPLETE without
> compunction. 

I wonder how other people are going about bug hacking.  When I did
gnome-games a couple of months ago, I first went and moved everything I
could figure out from general into the right games categories.  Once
they were there, I picked 1 game, and looked at every bug, from oldest
to newest.  There were a lot of dupes, and I marked them as such.  I
marked a lot as "needinfo", and I've not heard back on more than 2 of
those so far.  A lot more got closed as 'incomplete' or 'fixed' (thanks
Zana).  Juan Paolo has gone through and actually fixed a bunch of the
bugs that were real bugs.  AFAIK, the gnome-games maintainer hasn't done
anything, except to make a new release of the package after jrb and zana
did their gnome-games hacking (a bit before the 1.4 release).  
	Greg