[Zope-CMF] Re: [CMF-checkins] CVS: CMF/CMFCore - FSMetadata.py:1.5

[Tres Seaver]

On Mon, 2003-06-23 at 14:18, Florent Guillaume wrote:
> On Mon, 2003-06-23 at 19:37, Tres Seaver wrote:
> > On Mon, 2003-06-23 at 13:08, Andy McKay wrote:
> > > > I'd try to recreate it if I knew the specifics of the errors you
> > > > encountered. As it is, I'll mark on my todo list to try and revisit
> > > > this. I hate bare excepts.
> > > 
> > > There's a lot of them in Zope code to hate :) I'll try to recreate and 
> > > track down the *real* problem now I have slightly more time.
> > 
> > Bare excepts in Zope are particularly nasty, because they swallow
> > ConflictErrors;  in any currently-released Zope, such ill-mannered
> > behavior can lead to transactions which commit using inconsistent data.
> > 
> > I would vote for disabling 'except:' altogether in TTW code, and would
> > make 'ConflictError' and all its bases inaccessible from TTW code as
> > well (there is *no way* TTW code can know enough to do the Right Thing
> > in the face of a conflict error;  most filesystem code can't either, but
> > we "trust" those authors :( ).
> 
> Except for TTW code that wishes to do:
>    try:
>       somestuff()
>    except ConflictError:
>       raise
>    except:
>       raise ValueError("check your values bob")
> 
> But that's still quite bad.

Right, because *every* bare except in TTW code should do that.  I just
finished a jihad in co-developed code where we went through and added
such tourniquets, which sucked a lot.  I would rather have TTW code be
*unable* to catch a bare 'except:' at all.

Tres.
-- 
===============================================================
Tres Seaver                                tseaver@zope.com
Zope Corporation      "Zope Dealers"       http://www.zope.com