Note: There are a few cases of duplicates in user autocompletion which are being worked on.

TI: Assertion failure: !fe->isNotType(JSVAL_TYPE_DOUBLE), at methodjit/FrameState.cpp:715

RESOLVED FIXED

Status

()

Core
JavaScript Engine
--
critical
RESOLVED FIXED
6 years ago
5 years ago

People

(Reporter: decoder, Unassigned)

Tracking

(Blocks: 2 bugs, {assertion, testcase})

Trunk
x86_64
Linux
assertion, testcase
Points:
---
Dependency tree / graph
Bug Flags:
in-testsuite +

Firefox Tracking Flags

(Not tracked)

Details

(Reporter)

Description

6 years ago
The following testcase asserts on TI revision 32e8c937a409 (run with -m -n -a),
tested on 64 bit:

test();
function test() {
    function f(a, b, c) {
        return c;
    }
    if (gczeal == 'function') actual = f(1.5, 1.25, 1.125)
    else expect;
    (expect, actual, summary);
    var actual = '';
}
Regalloc bug from the SSA conversion.  When branching and computing the register allocation at the target bytecode, we assign FP registers at the target to any variables which are currently in FP registers and meet some other criteria.  However, with SSA we can have a variable which is a known double at the branch but is not known to be anything at the target, and would still assign an FP register at the target.  The fix filters out such variables from the register allocation, by looking at types of the target's phi nodes.

http://hg.mozilla.org/projects/jaegermonkey/rev/d834388eed47
Status: NEW → RESOLVED
Last Resolved: 6 years ago
Resolution: --- → FIXED
(Reporter)

Updated

6 years ago
Blocks: 676763
(Reporter)

Comment 2

5 years ago
A testcase for this bug was automatically identified at js/src/jit-test/tests/jaeger/bug655990.js.
Flags: in-testsuite+
You need to log in before you can comment on or make changes to this bug.