351988 - e4x adds spurious quotes when decompiling an XMLPI object initializer

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Robert Sayre]

js> var x = function() { var y = <?foo bar?>; }

function () {
    var y = <?foo "bar"?>;
}

Comment 1

[Brendan Eich [:brendan]]

js> var x = function() { var y = <?foo bar?>; }
js> x
function () {
    var y = <?foo "bar"?>;
}
js> dis(x)
flags: LAMBDA
main:
00000:  string "bar"
00003:  xmlpi "foo"
00006:  setvar 0
00009:  pop
00010:  stop

Source notes:
  0:     6 [   6] decl     offset 0

Trivial fix suggests itself: use JSOP_QNAMEPART instead of JSOP_STRING for the first bytecode.  Robert, you gonna take this one?

/be

Comment 2

[Robert Sayre]

Attachment: use JSOP_QNAMEPART

Comment 3

[Brendan Eich [:brendan]]

Comment on attachment 237571 [details] [diff] [review]
use JSOP_QNAMEPART

Cool, thanks.

Trivially safe and good for 1.8.1 if you want to nominate, and to help the decompiler fuzzer avoid a special case.

/be

Comment 4

[Robert Sayre]

Checking in jsemit.c;
/cvsroot/mozilla/js/src/jsemit.c,v  <--  jsemit.c
new revision: 3.205; previous revision: 3.204
done

Comment 5

[Mike Schroepfer]

Comment on attachment 237571 [details] [diff] [review]
use JSOP_QNAMEPART

a=schrep for JS decompiler bugs

Comment 6

[Robert Sayre]

Checking in jsemit.c;
/cvsroot/mozilla/js/src/jsemit.c,v  <--  jsemit.c
new revision: 3.128.2.41; previous revision: 3.128.2.40
done

Comment 7

[Bob Clary [:bc] (inactive)]

Checking in regress-351988.js;
/cvsroot/mozilla/js/tests/e4x/Regress/regress-351988.js,v  <--  regress-351988.js
initial revision: 1.1
done

Comment 8

[Bob Clary [:bc] (inactive)]

verified fixed 1.8 1.9 2006091022 windows/mac*/linux