579937 - treehydra_startup assertion incorrect after fatvals

Status: RESOLVED FIXED

(Developer Infrastructure :: Source Code Analysis, defect)

Description

[Ehren Metcalfe]

Attachment: patch

I'm not quite sure what the point of this assertion in that I don't see any conversions from void* to jsval or anything that makes assumptions about the size of a jsval. Tests pass with tracemonkey after this change.

The JSVAL_(TRUE|FALSE) changes silence a couple of overflow in implicit conversion warnings.

Comment 1

[Ehren Metcalfe]

Attachment: patch v2

missed a JSVAL_TRUE -> JS_TRUE

Comment 2

[Luke Wagner [:luke]]

JSVAL_(TRUE,FALSE) are completely different values than JS_(TRUE,FALSE).  The former are tagged values and thus will always look "true".  I'm surprised these were assertions and not compile errors.  Are you compiling without DEBUG?  In DEBUG builds, I made jsval and jsid struct types so these errors would be caught at compile time.

Comment 3

[Ehren Metcalfe]

Comment on attachment 458383 [details] [diff] [review]
patch v2

I'm compiling tm with DEBUG. struct -> int should be an error... odd. Either way, after talking to Taras, there's a lot more to fix (building on 64 bit ftl).

Comment 4

[Ehren Metcalfe]

Attachment: fatvals patch

I'm breaking any spidermonkey rev. between the rooting changes and fatvals with this btw (about a month of check ins). Works on 32/64 bit with 4.5 or 4.3.

Comment 5

[(dormant account)]

Comment on attachment 462467 [details] [diff] [review]
fatvals patch

Cool, that's exactly what I wanted. Need some sort of an ifdef #error or a configure check for old spidermonkeys.

Comment 6

[Ehren Metcalfe]

Attachment: fatval patch

this will still work with older spidermonkeys (I fixed a couple of nits though)

Comment 7

[Ehren Metcalfe]

http://hg.mozilla.org/rewriting-and-analysis/dehydra/rev/fd66b553a16c