Front page | perl.perl5.porters |
Postings from August 2010
> On Fri Aug 04 18:10:13 2000, jfriedl@yahoo-inc.com wrote:
>> With versions since 5.004, this program (with line numbers shown)
>>
>> 1| $_ = "foo";
>> 2| s{
>> 3|
>> 4| f
>> 5| o
>> 6| o
>> 7|
>> 8| }{
>> 9|
>> 10| warn "now";
>> 11| warn "now";
>> 12| warn "now";
>> 13|
>> 14| "FOO";
>> 15| }gex;
>> 16|
>> 17| warn "now";
>>
>> prints:
>>
>> now at /tmp/t3 line 4.
>> now at /tmp/t3 line 5.
>> now at /tmp/t3 line 6.
>> now at /tmp/t3 line 17.
>>
>> Perl 5.003 prints what I would consider to be correct:
>>
>> now at /tmp/t3 line 10.
>> now at /tmp/t3 line 11.
>> now at /tmp/t3 line 12.
>> now at /tmp/t3 line 17.
> This was an apparently intentional change in perl-5.003_26 (4fdae800)
> from Chip Salzenbrg titled "Refine 'runaway string' heuristic". 'Fixing'
> it is as simple as commenting out the line "PL_multi_start =
> first_start;" in toke.c, although "first_start" can be removed
> completely then. No tests fail as a result, although in a way I wish
> there were.
>
> I would prefer "warn" have the exact line, although there may be more
> cases it changes.
>
> Opinions?
>
I just noticed this.
The purpose of the patch was to fix the "runaway string" warning for
substitutions. Try removing the final "}" in the test code. I'll bet
5.003 says that there's a runaway substituion starting at line 8, rather
than line 2. I had no idea that PL_multi_start would affect the line
number diagnostics in s///e.
I'm inclined to remove my old patch to fix the above problem. A more
complete fix would be duplicating PL_multi_start to PL_multi_startwarn
(e.g.) and only reset one of them.
Thread Previous
|
Thread Next