Suggestion for a new warning

[p5pRT]

Migrated from rt.perl.org#1808 (status was 'resolved')

Searchable as RT1808$

[p5pRT]

From mike@tecc.co.uk

My suggestion​: I'd like to be able to turn on a warning that lets me know if my program contains function calls to functions that I haven't defined. So for example​:

  my $a1 = foo1();   if ($a1 == 3) {   $a1 = foo2();   }   print "a1 is $a1\n";

  sub foo1 {   return 2;   }

should report "foo2() not defined" when given the appropriate options. (This is a lot like the sort of reporting you get from the linker when you use Boring Old-Fashioned Languages.)

I appreciate that there is an overhead here\, so you'd not want this to be the default behaviour. But I think it's trivial to implement (right? Just walk the parse-tree before starting the run). It would help to ensure that programs that work today don't stop working tomorrow when someone feeds them input that wanders off down a different code-path.

Thanks for listening\,

  Mike Taylor   TECC Ltd.   mike@​tecc.co.uk   Tel. +44 181 880 4040

(Oh\, and thanks for Perl -- loving it!)

[p5pRT]

From [Unknown Contact. See original ticket]

At 03​:21 PM 11/19/99 +0000\, Mike Taylor wrote​:

Apologies for sending this to what I suspect is an inappropriate address\, but would you mind forwarding it for me please? (And what's the \suggestion\-box@​perl\.org address?)

My suggestion​: I'd like to be able to turn on a warning that lets me know if my program contains function calls to functions that I haven't defined.

Except this will kill​:

1) AUTOLOADed things 2) A goodly chunk of the perl core modules that have platform-specific function calls in code protected by "if ($^O eq 'whatever')" 3) Code that requires things 4) Code written by folks getting altogether too clever with symbol tables

Nothing insurmountable\, though fixing #2 would take some work. (Presumably people affected by 1\, 3\, or 4 just Won't Do That)

  Dan

----------------------------------------"it's like this"------------------- Dan Sugalski even samurai dan@​sidhe.org have teddy bears and even   teddy bears get drunk

[p5pRT]

From [Unknown Contact. See original ticket]

Date​: Fri\, 19 Nov 1999 10​:35​:30 -0500 From​: Dan Sugalski \dan@​sidhe\.org

My suggestion​: I'd like to be able to turn on a warning that lets me know if my program contains function calls to functions that I haven't defined.

Except this will kill​:

1) AUTOLOADed things

Uuuh ...

2) A goodly chunk of the perl core modules that have platform-specific function calls in code protected by "if ($^O eq 'whatever')"

Erm ...

3) Code that requires things

Duh!

4) Code written by folks getting altogether too clever with symbol tables

Ahem.

But otherwise\, it's great idea\, right? :-)

Nothing insurmountable\, though fixing #2 would take some work. (Presumably people affected by 1\, 3\, or 4 just Won't Do That)

Hmm\, I can't see it. Never mind\, and thanks for the ultra-prompt reply! (BTW.\, was perlbug a reasonable address to send this to?)

  Mike Taylor   TECC Ltd.   mike@​tecc.co.uk   Tel. +44 181 880 4040

[p5pRT]

From [Unknown Contact. See original ticket]

But otherwise\, it's great idea\, right? :-)

Yes\, actually.

Hmm\, I can't see it. Never mind\, and thanks for the ultra-prompt reply! (BTW.\, was perlbug a reasonable address to send this to?)

No\, you have a good point. It's exceedingly annoying to have a program die halfway through due to a seldom traversed codepath that includes a call to the lenght() function. Some pragma could be arranged to help this.

  no autoload; or   use warnings 'undefuncalls';

or some such.

That doesn't solve the issue for method calls\, and it is compile-time only warnings. I'm not sure whether this should be foisted off into B​::Lint as it is now. Did you know about that? This​:

  perl -MO=Lint\,-context\,-undefined-subs myperlprogram

might make you happy. Then again\, that's hardly a user-friendly front end. What ever happened to the idea of a plint program that did the icky work for you\, akin to perlcc?

--tom

[p5pRT]

From [Unknown Contact. See original ticket]

At 03​:56 PM 11/19/99 +0000\, Mike Taylor wrote​:

But otherwise\, it's great idea\, right? :-)

Yep. Heck\, I'd love to have it at times too. None of the problems are insurmountable\, just a touch tricky.

  Dan

----------------------------------------"it's like this"------------------- Dan Sugalski even samurai dan@​sidhe.org have teddy bears and even   teddy bears get drunk

[p5pRT]

From [Unknown Contact. See original ticket]

Date​: Fri\, 19 Nov 1999 09​:01​:41 -0700 From​: Tom Christiansen \tchrist@​jhereg\.perl\.com

But otherwise\, it's great idea\, right? :-)

Yes\, actually.

Oh\, thanks :-)

I'm not sure whether this should be foisted off into B​::Lint as it is now. Did you know about that? This​:

perl \-MO=Lint\,\-context\,\-undefined\-subs myperlprogram

might make you happy.

Ah\, no I didn't know about that -- many thanks\, I'll check it out. But does it make me happy? Not necessarily\, since the scripts where I most want to be able to do this sort of static checking are CGI scripts running on IIS/NT \ which I fear I will have trouble persuading to do what I want. Then again\, as this is a purely compile-time issue\, I guess I can still run the lint phase on Unix ... Hmm\, I'll try it out.

Then again\, that's hardly a user-friendly front end. What ever happened to the idea of a plint program that did the icky work for you\, akin to perlcc?

No idea\, I've only been writing Perl since July!

  Mike Taylor   TECC Ltd.   mike@​tecc.co.uk   Tel. +44 181 880 4040

[p5pRT]

From [Unknown Contact. See original ticket]

Dan Sugalski writes​:

At 03​:56 PM 11/19/99 +0000\, Mike Taylor wrote​:

But otherwise\, it's great idea\, right? :-)

Yep. Heck\, I'd love to have it at times too. None of the problems are insurmountable\, just a touch tricky.

Since it would be lexical\, it will not be even as bad as that. You try it once\, then you switch it off. ;-)

Ilya