[Data-Dumper] fix string comparisons with $] to use numeric compariso…

[book]

…n instead

The fix follows Zefram's suggestion from
https://www.nntp.perl.org/group/perl.perl5.porters/2012/05/msg186846.html

On older perls, however, $] had a numeric value that was built up using
floating-point arithmetic, such as 5+0.006+0.000002. This would not
necessarily match the conversion of the complete value from string form
[perl #72210]. You can work around that by explicitly stringifying
$] (which produces a correct string) and having that numify (to a
correctly-converted floating point value) for comparison. I cultivate
the habit of always stringifying $] to work around this, regardless of
the threshold where the bug was fixed. So I'd write

use if "$]" >= 5.014, warnings => "non_unicode";

• This set of changes does not require a perldelta entry.

[book]

This is ported from the psc/ppc0025 branch. The fix is useful even if we don't change the version scheme for Perl.

[jkeenan]

Data-Dumper is one of those blead-first, dual-life CPAN distributions which people actually do try to install against older perls. Do we have a way of testing the modifications in this p.r. against older perls (before merging this into blead)?

[haarg]

The GitHub actions check these against older perl versions, and those checks are passing.

[book]

It seem the cygwin-related failures can be ignored.

Excerpt from IRC #p5p on 2025-01-07:

12:49 <mauke> some kind of cygwin bug
12:51 <mauke> I was trying to test the changes in older perls,
              but somehow they fail to load the XS parts of the module,
              and several tests are broken without XS
12:51 <mauke> some have been for years (decades?)
12:51 <mauke> shows you how often people try to run the test suite
              against pure perl Data::Dumper :-)