Re: Some example code for a new crash handler



On Fri, 2002-01-11 at 22:14, Daniel Hauck wrote:
> You know, your idea is probably a very good one...for now.  But look where
> it leads.  We (I say "we" as if I were more than a user) need to focus on
> crash protection, not handling measures, in the long run.
> 
> I am not sure where Gnome's instability comes from except to quote Scotty
> from one of those Star Trek movies -- "The more you overtake the plumbing,
> the easier it is to stop up the drain."  Is Gnome too complex for its own
> good?
> 
> ....
> 
> Fix the problem, don't cover it.  That's my message.  I hope it could
> propogate as a new programming philosophy, but I don't carry the weight
> needed to get such a message to be listened to in a broad scale.  The
> plumbing has been overtaken too much and now problems are hard enough to
> track down that people are taking measures to simply clean up the mess AFTER
> the failure instead.
> 
> I can't deny the need for this measure because of its over-all impact on
> functionality, but remember what brought us to this point in the first
> place.
> 
> 


Well I agree that segfaults should be fixed, and not covered.
However, every program *will* crash once in a while, that is inevitable.
All programs have bugs, no matter how hard one try to keep it bug free.
Bugs should be fixed and not covered, but is that a good enough reason
to tell the user: "The program has crashed. It must die now and you
can't even try to save your documents."?
I think it's much better to warn the user that the program is about to
crash, and tell him/her to save any open documents, and then close the
program.

About Havoc's comments on garbage in the saved file:
if the program is really that corrupt, then even their own
emergency-save handler won't help much.
And not all programs implement emergency-saving.
If we put this in gnome-libs, then all apps will get it transparently.





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]