Item12446: Dependency check is inconsistent with Perl version comparisons

Foswiki::Dependency appears to treat decimal version to the right of the decimal as a simple number. This showed up in a question on IRC: http://irclogs.foswiki.org/bin/irclogger_log/foswiki?date=2013-03-22,Fri&sel=21#l17

hi, i am trying to install FullCalendarPlugin. But it shows missing dependencey Date::Calc version >= 6.13 required. I have already installed version 6.3 which is the latest in cpan

Confirmed with testing that 6.3 should compare as newer than 6.13.

$ perl -Mversion -e'print version->parse("1.13") < version->parse("1.3");'
1

1.13 < 1.3 evaluates as true. But Foswiki::Dependency evaluates that as false.

We can't use version for comparisons, as Foswiki::Dependency deals with versions that would fail Perl's comparson: dates, alpha strings, $Rev: 1234$ SVN revisions, etc. But the unit tests could test version against Dependency for supported version strings.

-- GeorgeClark - 25 Mar 2013

There's a function in Foswiki::Configure::Dependency::compare_cpan_versions - surely you can simply swap that out for use version?

3 is most definitely less than 13, despite what use version appears to say :-/

-- CrawfordCurrie - 25 Mar 2013

Since the $VERSION is external to core - coming from extensions, CPAN packages, etc. I think we need to be more robust than just using version. Things like $VERSION = "$Rev: 9845$" will cause version to throw an error. So we need to either carefully validate the string, and then pass it off to version, or just fix our own comparison.

I wish they called the module PerlVersion instead of just version. Sentences are so much harder to parse.

And last I checked, 1.3 is definitely larger than 1.13.

-- GeorgeClark - 25 Mar 2013

Hi Crawford,

I've changed Dependency.pm to use CPAN:version if the version strings pass a regex. There are definitely a few differences from our older Dependency comparisons. Please give this a look. Thanks.

-- GeorgeClark - 19 Apr 2013

1.30 is larger than 1.13; and 1.03 is smaller. But there is no convention on how to interpret a decimal. IMHO it's stupid to use a decimal; it has absolutely no advantages over using a simple integer, so why bother?

Your changes make sense to me; though this is a minefield

-- CrawfordCurrie - 19 Apr 2013