381101 - Incorrect use of new setter syntax in decompilation

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jesse Ruderman]

js> (function(){ ({ x setter: function(){} + {} }) })
function () {
    ({set x() {} + {}});
}

Comment 1

[Brendan Eich [:brendan]]

Attachment: spot-fix extracted from patch for the genexp bug

I think we want this for js1.7src. Probably won't get approved for the 1.8.1 branch though.

/be

Comment 2

[Jesse Ruderman]

This uses the wrong form (get vs getter) for lots of simple testcases.  (Tested with genexp4b patch, which includes this patch.)

js> (function(){ ({ x setter: function(){} }) })     
function () {
    ({x setter: function () {}});
}

js> (function(){ ({ set x(){} }) })             
function () {
    ({x setter: function () {}});
}

js> (function() { ({x getter: ({abcdefghijklmn: function(id2) { return id2; }}) }) })     
function () {
    ({get x ijklmn:function (id2) {return id2;}}});
}

Comment 3

[Brendan Eich [:brendan]]

Attachment: spot-fix, v2

Comment 4

[Brendan Eich [:brendan]]

Attachment: fix, v3

What I'm checking in. It would suffice to test for *rval == '(' to detect a generator expression (seemingly generated by a JSOP_ANONFUNOBJ op), but I stuck with the strncmp(rval, js_function_str, 8) != 0 test instead, just to avoid too much coupling to genexp decompilation.

/be

Comment 5

[Brendan Eich [:brendan]]

Fixed on trunk:

revision 3.240
date: 2007/05/18 00:48:18;  author: brendan%mozilla.org;  state: Exp;  lines: +4 -3
Fix getter/setter decompilation to depend on generating op, not prefix string (381101, r=mrbkap).

/be

Comment 6

[Blake Kaplan (:mrbkap) (inactive)]

Could the generator expression change ss->opcodes to avoid the string mashing sadness?

Comment 7

[Bob Clary [:bc] (inactive)]

/cvsroot/mozilla/js/tests/js1_7/decompilation/regress-381101.js,v  <--  regress-381101.js
initial revision: 1.1

Comment 8

[Bob Clary [:bc] (inactive)]

verified fixed 1.9.0 linux/mac*/windows.