[boost] Re: Re: Re: Re: Re: improved assertions?

From: Hendrik Schober (boost_at_[hidden])
Date: 2002-06-20 12:02:44

"David Abrahams" <david.abrahams_at_[hidden]> wrote:

  [my newsreader again mis-threaded your posting, which 
  is why I only saw it now]

> [...]
> >   I don't think it is more useful for the customer,
> >   since are core dump usually also means that they
> >   lost their work.
> 
> assert() == abort() on failure
> or at least, that's what I thought you meant.

  I didn't. 
  For me, a piece of code (an event handler) fails. 
  So there must be a bug in it. What can I say? With 
  the resources we have here, all our code has bugs. 
  Not a single user would pay for bug free software. 
  I simply have to live with the fact, that assertions 
  fail in shipped code. At customer's. Not on a daily 
  base, but they do fail once in a while. 
  My task was to deal with this, since developers 
  weren't asserting, because they shouldn't to abort 
  in the case of an error. So I did set up an 'ASSERT' 
  macro, which will throw a special exception. 
  As I said, that might stop the currently handled 
  event (e.g. print document etc.), throw a really 
  scary message onto the GUI, but it will allow the 
  user to proceed, if he/she wants to try to save 
  the last two hour's work. 

> I didn't realize you were trying to recover from a failed assertion.
> Dangerous, but sometimes it's the only practical approach.

  I'm not only a programmer, I am a user, too. And I 
  hate those messages that the OS not only slapped the 
  app I'm using onto it's fingers, but also killed it 
  right away. And I do appreciate to be told that 
  something really, really bad happened and that I 
  _should_ close the app and try again. And usually 
  I'll follow that advice. (Working on Windows for 
  years I got into the habit to constantly safe my work.) 

> [...]
> I was discussing the importance of having a stack tracing mechanism that
> wasn't dependent on exceptions. [...]

  I see. Sorry for the misunderstanding. 

> [...]
> >   Well, of course, once you have a portable/ported way
> >   to get a call stack from any point of the program,
> >   using this in the exceptions ctor would make a lot of
> >   sense.
> 
> Or, hopefully, from other places not involving exceptions...

  Of course. 

> -Dave

  Schobi