Documentation improvements in object_tracing.c

[srawlins]

I've updated docs in object_tracing.c a bit:

• trace_object_allocations_stop has improved wording.
• trace_object_allocations_clear now has a comment about the consequences on garbage-collecting.

I think that the example in the rdoc for trace_object_allocations should feature allocation_class_path and allocation_method_id but Koichi-san might want to do this in a specific way.

[zzak]

@srawlins Thanks for the patch!

I will check on the garbage collection note, and then commit your patch.

Feel free to submit a patch for allocation_class_path and allocation_method_id and we can discuss it there.

[zzak]

@srawlins You may also be interested in the following new methods to ext/objspace:

• ObjectSpace.trace_object_allocations_debug_start: 17e1c72
• ObjectSpace.reachable_objects_from_root: ede6826

[srawlins]

@zzak ObjectSpace.trace_object_allocations_debug_start looks interesting, but Koichi-san didn't leave any rdoc there. :/ The ChangeLog comment is cryptic.

ObjectSpace.reachable_objects_from_root does have some rdoc ("Return all reachable objects from root.") but it's not clear what is the difference between this and ObjectSpace.each_object, other than returning a Hash versus an Enumerator (maybe that is all).

I'd like to experiment with these method, thanks for the tip!

[zzak]

@srawlins I have confirmed with koichi-san that GC still runs during active trace_object_allocations, you can see yourself:

require 'objspace'

ObjectSpace.trace_object_allocations {
  loop { ' ' }
}

If you watch OS memory usage, it should indicate that GC is working.

I will commit your patch for trace_object_allocations_stop