Front page | perl.perl5.porters |
Postings from November 2013
Ricardo Signes wrote:
> Right now, the situation is already lousy:
>
> ~$ perl -Mwarnings -e 'sub z { return 0 } do z()'
> Use of "do" to call subroutines is deprecated at -e line 1.
>
> ~$ perl -Mwarnings -e 'sub x { return 0 } do x()'
>
> My assumption has always been that the 'x' operator is interfering.
I have just had a look at the code. 'do' followed by a keyword is
always do-file.
x is special-cased to allow x() function calls, unlike other infix ops.
You will find that BEGIN, END, etc. act the same way.
$ perl5.18.1 -we 'do BEGIN()'
Undefined subroutine &main::BEGIN called at -e line 1.
The same with do CORE(), until I changed it without realising in com-
mit a96df64385. (I am proving your point for you. :-)
>
> Anyway, I think we're better off simply eliminating the special case of `do
> SUB(LIST)`, meaning it would fall back to EXPR. On the other hand, I think
> this is where the question of backcompat will come in.
>
> If there's code with `do foo($x)` out in the wild, and it becomes equivalent to
> `do(do { foo($X) })`, something stupid is going to happen somewhere. I imagine
> we may be doing the world a favor by inserting a period of fatality.
Which is what the patch does, though mostly by accident I think. I
could follow it up with a clearer 'do SUB has been removed' message
after the patch is applied.
Should we make an exception for do CORE() and restore the do-file
functionality I inadvertently changed?
Thread Previous