822283 - Poison TokenStream::strictMode() when not in the parser

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[:Benjamin Peterson]

After parsing is done, if someone calls TokenStream::strictMode(), the StrictModeGetter will still access the parser's shared context to look for the strict flag. This is wrong when BytecodeEmitter is reporting a strict error a nested strict function (see bug 821470 and bug 821103). TokenStream::strictMode() should be poisoned after parsing, so errors can be detected.

Comment 1

[:Benjamin Peterson]

Attachment: make strictness apis private

Here's something. It makes APIs that access strictness not directly related to tokenizing private. I removed reportStrictWarning, since it is no longer used.

Comment 2

[Nicholas Nethercote [inactive]]

Comment on attachment 693182 [details] [diff] [review]
make strictness apis private

Review of attachment 693182 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good, thanks.

::: js/src/frontend/TokenStream.h
@@ +545,5 @@
>                                            va_list args);
>  
>    private:
> +    // This are private because they should only be called by the tokenizer
> +    // while tokenizing.

Maybe add something like "E.g. they shouldn't be called from the BytecodeEmitter."?

Comment 3

[:Benjamin Peterson]

https://hg.mozilla.org/integration/mozilla-inbound/rev/391f4cf534e1

Comment 4

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/391f4cf534e1