55114 - for... in loops not behaving as expected on Java strings

Status: VERIFIED FIXED

(Core Graveyard :: Java: Live Connect, defect, P2)

Description

[Phil Schwartau]

Getting failures on these LiveConnect test cases:

                js/tests/lc2/Arrays/array-003.js
                js/tests/lc3/forin/array-001.js

The problem seems to rest on how the characters of Java string objects 
are being enumerated -

Comment 1

[rogerl (gone)]

Here's a patch, need review/approval etc...

diff -u -r1.14 jsj_JavaArray.c
--- jsj_JavaArray.c     2000/08/26 02:30:19     1.14
+++ jsj_JavaArray.c     2000/10/03 23:18:43
@@ -252,6 +252,7 @@

     old_reporter = JS_SetErrorReporter(cx, NULL);
     result = access_java_array_element(cx, jEnv, obj, id, NULL, JS_FALSE);
+    if (result) *objp = obj;
     JS_SetErrorReporter(cx, old_reporter);
     jsj_ExitJava(jsj_env);
     return result;

Comment 2

[Phil Schwartau]

Nominating rtm -

Comment 3

[Phil Schwartau]

Setting P2 priority -

Comment 4

[Patrick C. Beard]

This is crucial for correct functioning of LiveConnect Java array access. The fix 
is very low risk.

Comment 5

[rogerl (gone)]

looking for r=,a= comments etc

Comment 6

[Michael La Guardia]

Marking "need info".  Please get your review and super review and renominate.

Comment 7

[Brendan Eich [:brendan]]

a=[e-mail].  Was this a regression in the code?

/be

Comment 8

[rogerl (gone)]

yes, I think when the changes in the for..in stuff was going in, or maybe 
earlier - we had a blank period on liveconnect testing.

Comment 9

[Mike McCabe]

r=mccabe.

I stepped through it and objp not being set causes us to mishandle the
enumeration.  Not to mention that it's obviously missing.

I think fur-style would have the the then-clause on a new line :)

Comment 10

[rogerl (gone)]

Fix checked in to trunk, appropriately en-furred.

Comment 11

[Patrick C. Beard]

Can we get an rtm++?

Comment 12

[Brendan Eich [:brendan]]

Crash bug, no?  If so, add crash keyword.

/be

Comment 13

[rogerl (gone)]

It's not reported as a crash, just test failures. But that must be just lucky 
since I guess there's likely to be garbage in *objp otherwise.

Comment 14

[Jim Roskind]

Small... stops crash.... rtm double plus.
Please land on N 6 branch asap.
Thanks,
Jim

Comment 15

[rogerl (gone)]

Fix checked in to branch

Comment 16

[Phil Schwartau]

Verified fixed using debug and optimized LiveConnect shells built on WinNT:

2000-11-01 (trunk)
2000-11-05 (MN6 branch)