32674 - JS_SetVersion hides error; also, jsscan.c keyword bug

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect, P3)

Description

[Brendan Eich [:brendan]]

The patch in this bug fixes a performance and infallible-API-with-fallible-impl 
bug in JS_SetVersion.  It also fixes what looks to me like a bug in the recent 
Unicode ECMA Edition 3 changes: if you ever spelled an identifier with Unicode 
escape sequence(s), you'd reset the unescaped atom's kwindex to -1, preventing 
even the unescaped spelling from being scanned as a reserved identifier.

Roger, mccabe, can you code buddy?  I'll check in if given the OK.

/be

Comment 1

[rogerl (gone)]

Urg I hadn't thought about that, sorry. The patch from the mail looks fine, how 
are you going to fix my kwindex bug? 
I just checked Rhino and it doesn't have the same problem because it does a 
complete separate lookup for the keyword.

Comment 2

[Brendan Eich [:brendan]]

The patch fixes the kwindex problem by treating any identifier that 
hadUnicodeEscape as non-reserved.

/be

Comment 3

[Brendan Eich [:brendan]]

Attachment: proposed fix

Comment 4

[Brendan Eich [:brendan]]

On second thought, I'm on sabbatical.  Roger, can you get this into the trunk, 
and then we can figure out what branch to get it into for JS1.5?  I don't see 
why the nscp_beta1_BRANCH needs it, but JS1.5 needs a branch.  Maybe we should 
have a JS1_5_BRANCH.  Whatever the resolution there, this change should go into 
the trunk ASAP.  Can you get it in, r=[e-mail]?  Thanks,

/be

Comment 5

[rogerl (gone)]

Applied patch, ran tests. Checked in.