818621 - Perl 5.16 complains with "Variable length lookbehind not implemented in regex" when the Example extension is enabled

Status: RESOLVED FIXED

(Bugzilla :: Extensions, defect)

Description

[Matt Selsky [:selsky]]

Bug 745197 added a variable length lookbehind regex in the Example extension.

This causes a test failure.

t/001compile.t ....... 29/230 Variable length lookbehind not implemented in regex m/(?<=<td id="error_msg" class="throw_error">).*(?=</td>)/ at ./extensions/Example/Extension.pm line 412.
Compilation failed in require at Bugzilla/Extension.pm line 70.

Comment 1

[Koosha KM]

I don't think you're right. I cannot figure out where the variable-length regex is.

Comment 2

[Matt Selsky [:selsky]]

$ /usr/local/bin/perl -e 's/(?<=<td id="error_msg" class="throw_error">).*(?=<\/td>)/$new_error_msg/si'
Variable length lookbehind not implemented in regex m/(?<=<td id="error_msg" class="throw_error">).*(?=</td>)/ at -e line 1.

Restricting the character set to ASCII seems to help.

$ /usr/local/bin/perl -e 's/(?<=<td id="error_msg" class="throw_error">).*(?=<\/td>)/$new_error_msg/siaa'

I'm running perl 5.16.1

It looks like behavior changed in perl 5.14 when additional Unicode support was added.

Comment 3

[Frédéric Buclin]

(In reply to Matt Selsky [:selsky] from comment #2)
> It looks like behavior changed in perl 5.14 when additional Unicode support
> was added.

It didn't change in 5.14, or at least 5.14 doesn't complain. AFAIK, only 5.16 is affected. Now you are no longer allowed to write "ss" in the regexp above ("ss" appears in the word "class"), else it's also seen as the german ß character. This is a completely stupid behavior from 5.16! We cannot use /aa either because it's only understood by Bugzilla 5.14 and newer. We could use \K instead, but it has been implemented in 5.10 only.

The workaround is trivial though. Patch coming.

Comment 4

[Frédéric Buclin]

Attachment: patch, v1

Comment 5

[David Lawrence [:dkl]]

Comment on attachment 689007 [details] [diff] [review]
patch, v1

Review of attachment 689007 [details] [diff] [review]:
-----------------------------------------------------------------

Otherwise looks good and works as expected. r=dkl

::: extensions/Example/Extension.pm
@@ +409,5 @@
>      $new_error_msg = html_quote($new_error_msg);
>      # There are better tools to parse an HTML page, but it's just an example.
> +    # Since Perl 5.16, we can no longer write "class" inside look-behind
> +    # assertions, because "ss" is also seen as the german ß character, which
> +    # makes Perl 5.16 to complain. The right fix is to use the /aa modifier,

nit: s/to complain/complain/

Comment 6

[Frédéric Buclin]

Committing to: bzr+ssh://[e-mail]/bugzilla/trunk/
modified extensions/Example/Extension.pm
Committed revision 8521.

Committing to: bzr+ssh://[e-mail]/bugzilla/4.4/
modified extensions/Example/Extension.pm
Committed revision 8486.

Committing to: bzr+ssh://[e-mail]/bugzilla/4.2/
modified extensions/Example/Extension.pm
Committed revision 8180.