Do not sort output from .xml.in

There should be no need to introduce a new dependency for this feature. Also, we should probably keep the sorting in, but do it on the source file location, just like xgettext does anyway.

Mmmh, if only my Perl skills were better :-/ Could this bug get higher priority?

Claude! You are joking right?

This bug has been relegated to forgotten wherever it is reported, even when it comes with a patch. The most response it ever got was for a coder to painfully, and badly, explain why it wasn't a bug.

The irony for me is that this bug so severely impacts localisation quality it makes me wonder about two things 1) Are programmers at all qualified to evaluate the importance of localisation bugs, 2) How much rubbish localisation must be out there if translators are translating with zero context information.

Yes, we definitely need such cludges, because... programmers cannot move their lazy butts and (gasp!) actually document what a string means! Writing comments is hard...

Giving it a higher priority would not help much since we have very limited time to work on these bugs. I'd be very delighted to get a complete patch which doesn't introduce a new dependency, and ensures all the tests are updated, and at least some exercise the sorting of messages. Ideally, it'd sort by *real* source reference, but I would not insist on that to solve the bug (because that'd be much harder with the intltool framework which actually lets xgettext extract messages from generated .h files).

Tie::IxHash does look like a useful way to handle this internally. That module's licensing is "same terms as Perl itself", which is an option of Artistic or GPL, and intltool is GPL also. Therefore, it seems permissible to include the module code directly in intltool itself. The Tie::IxHash module is just a single .pm file (~400 lines of code), so could just bundle that file with an appropriate installer directive (or maybe only do it conditional on the system not't have the module already available?). Or else just copy the code into directly into intltool-extract.in.

Looking at the specific uses of %messages, it would be pretty easy to hack together a poor-man's Tie::IxHash right in this one script, since all you need to do is store/retrieve by key and generate a sorted list of keys (not need to delete, take slices, etc.). I'll try to take a look in a few days.

Indeed, it should be quite simple to provide just the needed functionality. Thanks for looking into it, and I look forward to a patch :)

Seeing how no progress has been made on this, I'll be working on this today.

Fixed in r721 of lp:intltool, please test.