I have a matching app that uses Text::Compare (among a number of other matching algorithms/modules). When the same app is run on Linux vs. AIX, it baloons memory. I've narrowed this down to the langof() subroutine that Text::Compare calls.

If I set the $lang in Compare.pm to 'en' and run, it sits at a steady ~24-26MB on Linux, but if I let langof() be called (as the module *does* by default - no I'm not the maintainer), the app consumes all available memory and segfaults....

.... on Linux

.... but *NOT* on AIX. It runs fine on AIX. Always happily hums along at around 27MB for the same data set.

Now, assuming a value of 'en' isn't *that* damaging to my app (I just get more false positives - which is OK), but it'd be nice to know what's going on.

Anyone else come across this?

This is Perl 5.8.8 on both machines, both threaded, both with the same module set, and, in particular, both with the same versions of the above modules. (I have as nearly as possible an identical environment on both systems because the Perl code could potentially run on either).

Kind regards

Derek Jones