539819 - The statement "0/0;" is not always compiled out

Status: RESOLVED DUPLICATE of bug 632020

(Core :: JavaScript Engine, defect)

Description

[Jesse Ruderman]

On the TM branch, jsfunfuzz discovered that whether "0/0;" is compiled out depends on whether the function as a whole is parenthesized.  This makes jsfunfuzz complain a lot, so I'll probably have to turn off its round-trip checking until this bug is fixed.

Maybe this is related to constant-folding and caused by bug 537673?

With parens:

js> (function() { (function b() {0 / 0;})(); })
(function () {(function b() {0 / 0;}());})

Without parens:

js> function () {(function b() {0 / 0;}());}
function () {(function b() {}());}

Comment 1

[Jeff Walden [:Waldo]]

Hmm, yeah, we should be hopping down a level or two, not skipping it entirely at that point.

Comment 2

[Jesse Ruderman]

Waldo, ping? We're missing a lot of jsfunfuzz coverage because of this bug.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

The js shell errors out now if the function isn't enclosed in parentheses. Is there another way to test this or is this no longer valid?

Not sure if this is really equivalent, but when I use print(), I get the following:
js> print((function () {(function b() {0 / 0;}());}))
function () {
    (function b() {}());
}
js> print(function () {(function b() {0 / 0;}());})
function () {
    (function b() {}());

Comment 4

[Jesse Ruderman]

I can still reproduce the bug, like this:

js> print(function() { (function b() {0 / 0;})(); })
function () {
    (function b() {0 / 0;}());
}

js> print((function () {(function b() {0 / 0;}());}))
function () {
    (function b() {}());
}

Comment 5

[Jesse Ruderman]

With the decompiler gone, this bug no longer bites jsfunfuzz.  But it's still present, as can be seen with:

dis("-r", function() { (function b() {0 / 0;})(); })
dis("-r", (function () {(function b() {0 / 0;}());}))

Comment 6

[Jan de Mooij [:jandem]]

I just noticed this bug in a bug list.

This also applies to things like a["x"] => a.x folding and so it's an interesting perf issue I want to track down.

Comment 7

[André Bargull [:anba]]

Probably the same issue as in bug 632020.

Comment 8

[Jan de Mooij [:jandem]]

(In reply to André Bargull [:anba] from comment #7)
> Probably the same issue as in bug 632020.

Yes, thanks :)