1082107 - OdinMonkey: add maximum-length requirement to change-heap definition

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[Luke Wagner [:luke]]

Attachment: tweak-change-heap

This patch adds an additional maximum-heap disjunct for the reasons described in:
  http://discourse.specifiction.org/t/request-for-comments-switching-resizing-heaps-in-asm-js/641/17

Comment 1

[Luke Wagner [:luke]]

Sorry for the churn (particularly in tests).  I tried to be careful to make sure all the failing tests failed for the intended reason, not just b/c they were missing the third disjunct.

Comment 2

[Benjamin Bouvier [:bbouvier] (inactive)]

Comment on attachment 8504225 [details] [diff] [review]
tweak-change-heap

Review of attachment 8504225 [details] [diff] [review]:
-----------------------------------------------------------------

Easy to read once we're acquainted with AST pattern matching :)
Would it make sense to wait for comments on the specifiction thread before landing this?

::: js/src/jit-test/tests/asm.js/testResize.js
@@ +175,5 @@
>  
>  // Tests for validation of heap length
>  
> +var body = USE_ASM + IMPORT1 + 'function ch(b2) { if(len(b2) & 0xffffff || len(b2) <= 0x1ffffff || len(b2) > 0x80000000) return false; i8=new I8(b2); b=b2; return true } function f() { return 42 } return ch';
> +var m = asmCompile('glob', 'ffis', 'b', body);

Could you add a test involving a module compiled with a low max-limit, and an attempt to change the heap to a heap with an value higher than this limit? I think i didn't see any.

Comment 3

[Luke Wagner [:luke]]

(In reply to Benjamin Bouvier [:bbouvier] from comment #2)
> Easy to read once we're acquainted with AST pattern matching :)
> Would it make sense to wait for comments on the specifiction thread before
> landing this?

Yep, I was planning to do this.

> Could you add a test involving a module compiled with a low max-limit, and
> an attempt to change the heap to a heap with an value higher than this
> limit? I think i didn't see any.

Yes, good idea.

Comment 4

[Luke Wagner [:luke]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/8d4702c0db51

Comment 5

[Luke Wagner [:luke]]

Attachment: aurora.patch

Carrying forward r+; this is just the original patch with feedback addressed.

Approval Request Comment
[User impact if declined]: inconsistent validation of new asm.js feature between two versions
[Describe test coverage new/current, TBPL]: mozilla-inbound, has tests
[Risks and why]: low, this is just a validation tweak on top of bug 965880, which landed right before ff35 branched

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/8d4702c0db51

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/5feb6e3234e5