Re^5: erroneous warning involving locale and input encoding: perl bug?

[dave_the_m]

The fault can be reduced to the following:

use experimental 'smartmatch';
use POSIX 'locale_h';
use locale ':ctype';
setlocale(LC_CTYPE, 'en_US');

$_ = "x";
utf8::upgrade($_);
/x(y|z)?/;

which gives an assert failure on bleadperl. The locale-variant of the TRIE code in the regex engine appears to be treating the 'no more chars' special value of nextchr (-10) as a real large utf8 character:

                        && UTF8_IS_ABOVE_LATIN1(nextchr)

By all means perlbug it

Dave.

[raygun]

Thanks, Dave, for the detective work! I confirm that your example produces the warning on v5.22.2 as well. (I'm not quite sure from your description whether the failure method on bleadperl is the same as that on v5.22.2. An assert is a fatal error (in C, anyway), whereas the message I'm seeing does not halt execution; if I add another statement after the end of your code sample, it does get executed.)

I'm happy to submit a bug report, but since you know much more about what's going on under the hood, you're probably a better contact person. If you don't have the time or inclination, however, I'll do it.

[raygun]

Bug reported.