find_method op should get sub pmc for pmclass method

[ronaldws]

This bug report recommends, based on an IRC discussion, that the find_method operation be able to retrieve a sub pmc for a method declared in a pmclass declaration.

I asked on IRC about getting a sub pmc for a pmc method declared in a pmclass declaration without looking for the method name in a namespace. It was suggested that I try the find_method op on a pmc instance and, when that didn't work, some of the discussion participants, who felt that it should, recommended a trac ticket.

I have also attached a patch that would add a relevant test to t/pmc/pmc.t.

[jkeenan]

Can we get an update on the status of this patch?

Thank you very much.
kid51

[ronaldws]

A revised, updated, patch of tests related to this ticket

[ronaldws]

When I apply the updated patch and run the tests, the patch shows that the fundamental issue raised in the ticket has been addressed at this time.

Looking at the original tests, it seems that, during the IRC discussion that started this trac ticket to begin with, a request was made to prevent pmc method subs from being retrievable through the class namespace. I vaguely recall that the issue was namespace pollution but am not sure and it may also have been about inheritance or other possible pmc instance transformations. I also don't really know whether it is still considered a concern.

Reviewing the IRC discussion for the link below, I don't see the reason for the restriction, but am a little reluctant to just drop it without another pair of eyes taking a look.

 http://irclog.perlgeek.de/parrot/2009-03-10#i_972463

[ronaldws]

The issue mentioned in my prior comment about pmc method subs being retrievable through the class namespace seems to be covered by trac ticket 389  http://trac.parrot.org/parrot/ticket/389. I am inclined to move the test from pmc.2.patch labeled _ns_test out, possibly to ticket 389, apply the remainder of the patch to pmc.t and request the closure of this ticket. Commentary, particularly by anyone familiar with TT 389 appreciated.

Thanks, Ron

[ronaldws]

New revised test file with one more test for builtin

[ronaldws]

I found an existing test for this functionality in t/pmc/object-meths.t coded in pasm. I could not figure out how long it had been there but after looking at the svn log found no reason to believe it had been added or modified recently. My initial test case used a String PMC object which was created from the builtin String PMC class which, it now appears, may have been the source of the problem with find_method in the first place, I'm not sure. In any case the latest attached patch translates the existing pasm test to pir and adds a test for the builtin pmc String class that opened the ticket.

I think it is pretty much time to close this ticket. I would be in favor of applying the attached object-meths.patch but I have been away from parrot development for a while and would like someone more closely involved to review my suggestions. Ticket 389, mentioned in my previous comment, is now closed.

Thanks, Ron

[whiteknight]

Applied with some modifications in 6c3990d. Resolved.