23612 - JavaScript mishandles characters \u001C - \u001F

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect, P3)

Description

[Waldemar Horwat]

ECMAScript states that characters \u001C - \u001F in program source should be
treated as syntax errors.  Our engine treats them as whitespace.

js> eval("2\u0018+4")
3: SyntaxError: illegal character:
3: 2+4
3: .^

This is correct.

js> eval("2\u001F+4")
6

This should also generate a syntax error.

Comment 1

[Mike McCabe]

Forwarding to Roger.

Comment 2

[Waldemar Horwat]

This bug is most easily fixed by changing entries 0x1C through 0x1F of js_Y[] in
jsstr.c from 1 to 0.

Comment 3

[Brendan Eich [:brendan]]

Do regexps classify those characters the same as the lexical grammar does?  The
js_Y table (based on Java code) is used for both, IIRC.

/be

Comment 4

[Robert Ginda]

testcase ecma_3/Unicode/uc-001-n.js added

Comment 5

[Waldemar Horwat]

According to ECMA, regexps should not treat these characters as white space
either.  If regexps use the same table, then this fix would fix both problems.
We might want to add a test case for testing these characters against \s in a
regexp.

Comment 6

[Robert Ginda]

Added ecma_3/Unicode/uc-004.js testcase to cover regexp usage.

Comment 7

[rogerl (gone)]

Fix checked in.

Comment 8

[Robert Ginda]

Marking Verified.