401405 - narcissus - invalid label error on nest ternary

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Bob Clary [:bc] (inactive)]

js> a=b=c=d=e=1
1
js> q='a?b?c:d:e'
a?b?c:d:e
js> eval(q)
1
js> parse(q)
:1: Invalid label
js>

Comment 1

[Brendan Eich [:brendan]]

Attachment: fix

Comment 2

[Brendan Eich [:brendan]]

Attachment: forgot jsexec.js

Comment 4

[Brendan Eich [:brendan]]

Fixed:

js/narcissus/jsexec.js 1.27
js/narcissus/jsparse.js 1.25

/be

Comment 5

[crisp]

This fix is incorrect; although it does fix the erronous 'invalid label error' the result of 'a?b?c:d:e' is actually evaluated as 'a?b:c?d:e'

I fixed this by reusing 'CONDITIONAL' for nested ternaries, but giving this a higher precedence:

jsparse.js
654a655,656
>     // add special precedence for nested ternaries
>     CONDITIONAL: 3,
679c681
<     HOOK: 3,
---
>     HOOK: 3, CONDITIONAL: 3,
761a764,765
> 		if (x.hookLevel > 1)
> 		    n->type = CONDITIONAL;

jsexec.js
502a503
>       case CONDITIONAL:

I'm not sure what the current status of Narcissus is; is it actively used besides in ported projects?

Comment 6

[crisp]

err, n->type should read n.type ofcourse (doing too much PHP these days)

Comment 7

[crisp]

hmmz, my patch only moves the problem ahead; it now breaks stuff like:

a ? b ? c = 1 : c = 2 : c = 3

and we're back to bug https://bugzilla.mozilla.org/show_bug.cgi?id=330975

I wonder if this is solvable in this way or if it would be better to handle this as a seperate construct.

Comment 8

[crisp]

Since this issue is closed I opened a new one with a new proposed patch: https://bugzilla.mozilla.org/show_bug.cgi?id=492445