Refactoring mb_str(r)(i)pos out of bound checks/errors #5109
Conversation
|
There are some conditions where this will miss out of bounds offsets, for example if haystack is shorter than needle, which the mbfl code short-circuits. |
|
I've now updated mbfl_strpos() to always detect offset error conditions. Can you please rebase? |
Girgias
force-pushed
the
mbstring-pos-func-oob
branch
from
459ec62 to
5251ae5
Compare
Girgias
force-pushed
the
mbstring-pos-func-oob
branch
from
5251ae5 to
f9cbad0
Compare
| @@ -2084,13 +2084,13 @@ static void handle_strpos_error(size_t error) { | |||
| case MBFL_ERROR_NOT_FOUND: | |||
| break; | |||
| case MBFL_ERROR_ENCODING: | |||
| php_error_docref(NULL, E_WARNING, "Unknown encoding or conversion error"); | |||
| zend_value_error("Unknown encoding or conversion error"); | |||
There was a problem hiding this comment.
Let's leave this one as a warning. This error doesn't actually occur (as far as I know), but if it did, it would be due to invalid encoding in the input string, which shouldn't throw. You can change the error message to just "Conversion error" to make that clear. (Unknown encodings are caught separately.)
|
Side note, mb_strrchr still rejects empty needles. |
I'll have a look at it, must have glossed over during the (lengthy) other PR :D |
This is in part splitting #4971 where these changes were combined into.