[prev in list] [next in list] [prev in thread] [next in thread] 

List:       mozilla-builds
Subject:    Re: assertion question
From:       Akkana Peck <akkana () netscape ! com>
Date:       1999-05-05 17:19:33
[Download RAW message or body]

Adam Lock wrote:
> When running in debug mode, Mozilla assertions simply crash apprunner
> outright without printing a useful dialog box with a message telling me
> what has gone wrong and letting me retry/ignore/cancel to investigate or
> ignore it in a sane way.

This should be regarded as a bug; it was probably written that way by a
programmer on some other platform (I can't imagine any Unix programmer
*liking* the behavior of crashing without printing a message,
especially when Unix is the only platform where it crashes).

We should fix this.  Maybe I will (when I get my build working again --
right now it's exiting on startup with "_X11TransSocketUNIXConnect:
Can't connect: errno = 111").

Daniel Veditz wrote:
> Why would you leave it? If NS_ASSERTION called DbgAssert() instead you
> still get the option of dropping into the debugger, but you don't *have*
> to.
> 
> It would be one thing if asserts were always used only for truly awful
> "things that are going to crash anyway", but I've hit a heck of a lot of
> 'em that are simply placeholders for "I really should do error handling
> here someday".
[ ... ]
> On Unix it just crashes. Really useless.

Agreed.  Let's fix it.  In the old (4.*) code, when we got tired of
asserts crashing, we changed them print a message and a control-G, so
if you weren't watching stdout you'd still get an alert that something
had happened.  That seemed to work well; how 'bout if we do it that way
again? 

	...Akkana

[prev in list] [next in list] [prev in thread] [next in thread]