103602 - Not getting a JavaScript strict warning for reassigning to a const

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Phil Schwartau]

This was discovered by jrgm. Here is a session in the JS shell: 

[]./js -w

js> const one=1;

js> const one=2;
4: TypeError: redeclaration of const one

js> one=2;
2

This shows that we get a warning if we try to redeclare a const,
but we do not get a warning if we try to reassign to a const.
It's true, the new value doesn't take: 

js> one
1

but there should be a warning, no?

Comment 1

[Phil Schwartau]

Attachment: HTML testcase by jrgm

Comment 2

[Alex Vincent [:WeirdAl]]

Oh, wow.  That's not good.

0.9.8 is coming up fast, any indications of work on this?

Comment 3

[Phil Schwartau]

cc'ing Brendan in case this is of interest -

Comment 4

[Brendan Eich [:brendan]]

That's per ECMA (don't blame me, I fought for what Netscape always did: throw an
error [could be a catchable exception since 1.3]).  Readonly properties, when
set by assignment, are not changed, but no error or exception is thrown.  The
value of the assignment expression is the value of the r.h.s.

If you want a *strict* warning, please change the summary of this bug to say so.

/be

Comment 5

[John Morrison]

Yeah, I only meant strict. I'm not particularly worried about whether/when this 
gets done. I just noticed that there was no warning for this, and it seems like 
something a strict mode should flag.

Comment 6

[Phil Schwartau]

Same behavior occurring with the -s option as with the -w option above: 

[]./js -s
js> const one=1;
js> one=2;
2
js> one
1

but no warning is given -

Comment 7

[Phil Schwartau]

Testcase added to JS testsuite to ensure what Brendan described in Comment #4:

          mozilla/js/tests/js1_5/Regress/regress-103602.js

Comment 8

[Brendan Eich [:brendan]]

shaver reminded me that exceptions didn't show up until js1.4.  This is a low
priority bug, are we sure it should be targeted at 0.9.8?

/be

Comment 9

[Kenton Hanson (gone)]

targeting future.

Comment 10

[Henrik Gemal]

sorry for the spam: just so it shows up in my javascript query...

Comment 11

[Brian Crowder]

Tested this out just now, I think it is fixed:

js> const one = 1
js> one = 2
typein:2: strict warning: one is read-only
2

Comment 12

[Bob Clary [:bc] (inactive)]

verified on 1.8, 1.9 windows.

Comment 13

[Brendan Eich [:brendan]]

Tradition would mark this bug WFM unless we know the bug that was patched to fix this symptom.  A little cvs annotate -r and cvs diff archeology discloses:

revision 3.230
date: 2005/12/01 21:05:12;  author: brendan%mozilla.org;  state: Exp;  lines: +24 -24
Followup to maximize error-reporting code sharing (315436, r=daumling).
----------------------------
revision 3.229
date: 2005/12/01 02:26:35;  author: brendan%mozilla.org;  state: Exp;  lines: +25 -7
Fix from Michael Daumling <[e-mail]> to make read-only assignment a strict warning (315436, r/sr=mrbkap/brendan).

This is when the fix went in.

/be

Comment 14

[Georg Maaß]

Will Moffat posted in mozilla.dev.tech.js-engine


I'm using Gran Paradiso rv:1.9a2 with javascript.options.strict = true
and JavaScript Shell 1.4

const one=1;
const one=2; --> TypeError on line 1: redeclaration of const one
one=2; --> 2  *No warning*
one --> 1

According to Bug 103602 - "Not getting a JavaScript strict warning for
reassigning to a const" this was verified fixed on 1.8, 1.9 windows.
Shouldn't Gran Paradiso contain this fix?

Any ideas why I don't see a warning?
regards,
--Will


I've tested this issue with nightly build from trunk [Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.9a3pre) Gecko/20070222 SeaMonkey/1.5a] where I get the warning as desired. So I can not verify a new regression. For me it still looks fixed.

Comment 15

[Brendan Eich [:brendan]]

Assignment is not declaration. The redeclaration warning is only when const is used to declare the constant identifier.

I don't think this regressed:

$ ./Darwin_DBG.OBJ/js -sw
js> const one=1
js> one=2
typein:2: strict warning: one is read-only
2

Are you (is Will, I mean) sure you have strict warnings turned on?

/be

Comment 16

[Will Moffat]

My bad, there is no regression. Two factors confused me:

The reassignment warnings go straight to the JS Console, they are not exceptions which can be caught and displayed by the JS Shell extension.

I was using an old build with bug 324142, so nothing appeared in the JS Console