Does this constitute a bug ... ?

I wouldn't call it a bug, unless I was also prepared to say these are:

    print 0+"0xFF"; # 0
    print 0+"0755"; # 755
    print 0+"0b11"; # 0
[download]

But I don't think those are bugs, either.

It's all pretty much defined behavior. Any language that makes thin distinctions between strings and numbers (as Perl does) needs to have some defined way of what strings do when they're applied to numbers, and what numbers do when applied to strings. You might take issue with the way Perl does that, and you might take issue with the fact that Perl doesn't have a value-based type system.

But it's all well-defined and documented (more or less).

What I was getting at, is that in general perl promotes a number at the start of a string to the same number as if it was a literal. I was suggesting that since it doesn't in this case, you could consider it a bug.

I do consider it a bug, but I would not consider changing the current behaviour before perl6, because of backward compatibility. It would break too subtly in too many scripts.

As an example of "working" DWIM:

my $i = 1234e3;
my $j = "1234e3";

print join("\n",$i, $j, 0+$i, 0+$j);
[download]

"What should it profit a man, if he should win a flame war, yet lose his cool?"

should people just refrain from using _ as a NOP in numbers?

It's not that people should ... it's that people do refrain from using it that makes this nasty :-)

(And the people who do use it ... only use it every third digit, and so we aren't used to looking at these wierd numbers with random _'s in them.)

(Maybe this thread almost qualifies for the Obfu section...)