On Wed, Mar 14, 2001 at 03:16:42PM -0800, Tommy Butler wrote:
> Just to stir the pot a bit, everyone knows about the tendency of the Perl
> debugger to occasionally give incorrect line numbering.
>
> Two questions:
> How long would that bug be allowed to last in a similar product from one of
> the evil empire companies?
> How would our criticism of that bug in that product differ from the casual
> way it's treated in Perl?
>
> -----------------
> Well, technically it isn't a bug, but an unclear implementation. You can
> get around it -- that's why it probably hasn't been fixed, because it's not
> really "broken". You have to take into account that re-tooling little
> things like this contributes to the size of the core. Perl is so great for
> speed, improving over tcl and python, even java in some cases. Many of us
> who maintain large enterprise software products in perl are more concerned
> with these issues than an error which is reported 5 lines from its apparent
> origin. We just educate ourselves about the issue and learn to understand
> what the compiler is telling us.
I think that's an attitude that most businesses do not share. Heck, I
don't share that attitude when it comes to a business environment.
Speed is *never* an issue when it comes to correctness. I don't care for
a fast, wrong answer.
If line numbers are reported wrong, the tool is broken. And you can
bent words in any way you want and declare it unclear implementation or
"not really broken", for someone evaluating its use in the coorperate
environment it's broken, and *his* opinion whether that's a bug or a
feature matters. If you care about Perl playing a prominent role in
many coorperation, you have to keep that in mind.
And I find it hard to believe that having the debugger report wrong line
numbers is a significant speedup it, especially considering the typical
difference in speed between a well written Perl application, and a well
written C or C++ application with the same functionality.
Abigail
