640852 - can we remove the atom-compartment-avoidance gunk from string marking?

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Luke Wagner [:luke]]

Attachment: rm

Bill and Gregor couldn't think of any reason it was bad to mark atoms in the atomCompartment during a per-compartment gc.  Removing the gunk not only removes a tidy bit of code, bit makes a simple marking microbench (the one in bug 616562 comment 0) 5% faster.

Comment 1

[Gregor Wagner [:gwagner]]

Comment on attachment 518609 [details] [diff] [review]
rm

Can we leave the part where we assert that all "chained" strings are in the same compartment or end in the atomsCompartment?

Comment 2

[Luke Wagner [:luke]]

Oh, I meant to comment on this:

My reasoning for nixing those is that the "same or atoms compartment" property is asserted by js_ConcatStrings which is the only way to get a rope, so all this really seems to be guarding against is some bizarre memory corruption.  I.e., I don't these asserts would catch compartment misuses since those would (and do) hit earlier (hopefully much earlier now that assertSameCompartment checks strings).

Comment 3

[Luke Wagner [:luke]]

http://hg.mozilla.org/tracemonkey/rev/4a73b3a12f30

Comment 4

[Chris Leary [:cdleary] (not checking bugmail)]

http://hg.mozilla.org/mozilla-central/rev/4a73b3a12f30