526970 - Handle wide string literals in lazy_tree_string

Status: RESOLVED FIXED

(Developer Infrastructure :: Source Code Analysis, defect)

Description

[Philip Taylor]

lazy_tree_string calls convert_char_star which calls JS_NewStringCopyZ, for all strings. This fails when it's a wchar_t* string (it only copies the first byte before hitting a null byte and stopping), and makes it impossible to access the string contents from scripts. Strings with non-char precision ought to be converted properly into Unicode JS strings.

Comment 1

[Philip Taylor]

Attachment: rough patch

This is a nasty hack - I don't know the right way to implement the feature. But this at least worked for me. Also it fixes handling of non-wide strings with embedded null characters.

Comment 2

[(dormant account)]

Philip can you attach a testcase with instructions on how to run it?

Comment 3

[Philip Taylor]

Attachment: test case

A test case (which the patch passes), using the standard Treehydra test runner.

(I'd prefer it if the string was exposed as 't.string.str' or 't.string', rather than 't.string.str_full', but the latter seemed easiest to implement given that I don't quite understand how the code works.)

Comment 4

[(dormant account)]

nice hack! I think the right thing to do here is some magic. I think

t.string.str should do the unicode conversion that your .str_full hack does. Patch coming right up.

Comment 5

[(dormant account)]

Attachment: my version

Thank you for the excellent exploratory patch. It made it really easy to turn into something landable. Please confirm that this works for you and I'll land it.

Comment 6

[Philip Taylor]

Comment on attachment 410852 [details] [diff] [review]
my version

The only problem is I get a compiler error "treehydra.c:223: error: expected ‘)’ before ‘uint16_t’". If I #include <stdint.h> then it looks like it works fine.

Comment 7

[(dormant account)]

http://hg.mozilla.org/rewriting-and-analysis/dehydra/rev/0a5fc065b382