Revisions code in 2.2 [was "Re: time travel ;)"]

From: Mark Gilbert <mg_abimail_at_yahoo.com>
Date: Tue Jun 08 2004 - 03:10:21 CEST

Dev list: I'm ccing the user list on this one.
User list: This is from a thread that touched on 2.2's status, this reply is about the revisions
and document history features (and misfeatures) in 2.1.x (2.1.3 has been tagged and uploaded and
will likely be released very soon).

--- msevior@physics.unimelb.edu.au wrote:
>
> 2.2. We might have to disable all the revision marks code which
> is still extremely buggy and very, very dangerous.
>

Yeah, we said that before 2.0 (-:

> We'll see what our fantastic QA team comes up with.

Finding (and hopefully fixing) the crashes (like 'insert anything in revisions mode') isn't so
hard, the dogfooding to seek out more subtle but equally feature-destroying runnable bugs (like
'midline backspace after second line post-undo misplaces caret/char, leaves ghost chars' or the
even more disturbing 'yada yada yada with revisions causes dataloss on save') will be the hard
part, requiring lots more time localizing and verbose bug filing. Also, much like the random
crashes vs. easily reproducable crashes situation, this causes a great deal more distress for the
users. Why? First and foremost, just like the qaers who'll have trouble localizing, the users
don't know what's going on or how to avoid it. May even think they're doing something wrong, but
usually will just ditch Abi on the side of the road. Second, it severely hampers productivity,
slowing you down while working in abi or forcing you to backtrack and do stuff over that got
messed up. Third, it enables the dataloss situation where user, rather than being unable to do
anything, gets a whole lot of work done and only then loses it all, along with the time he/she
spent doing it. Very upsetting. Fourth...bah, you get the idea.

Lots of "dogfooding" (real-use or quasi-real-use testing) needed on revisions, **even working
around the crashers to test the other functionality or lack thereof**.

Users who might have been saying "well, I'd love to get working revisions, but I don't know how to
debug a crash" can help greatly just by localizing testcases for runtime misbehavior and filing
profusely.

We need tests of all variety (as usual), including using rtl text, using non-typical-english fonts
(microscope the rendering) and encodings, using revisions in different 'special' or 'other than
plain text' cases (the ones that don't crash, which hopefully is increasing), using doc history
(and all the previous tests with doc history), using revs/history to make and/or edit long and
complex docs, using revs/history over multiple sessions (which is what they were meant for), revs
with spell checking, saving and exporting of revs/history-using docs...

I'm sure I could come up with more, hopefully you all (users that don't mind spending time with an
unstable beta version and would like to help, as well as developers) can too. Even if you can't
(or don't want to) take the time to localize test cases and file reports, emailing one of the
lists or me or the qa list or stopping by on irc, to let us know of a buggy revisions situation
that hasn't been filed yet (http://bugzilla.abisource.com/query.cgi), at least gets that on the
radar, so that hopefully a(nother) developer or qa-er can pursue it.

Best regards
-MG

        
                
__________________________________
Do you Yahoo!?
Friends. Fun. Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/
Received on Tue Jun 8 03:00:08 2004

This archive was generated by hypermail 2.1.8 : Tue Jun 08 2004 - 03:00:08 CEST