Re: Patch for Review: JvGetStringUTFChars

[Bryce McKinlay <bryce at mckinlay dot net dot nz>]

On Wednesday, Aug 13, 2003, at 19:52 Pacific/Auckland, Mohan Embar 
wrote:

> After careful consideration, I've decided not to go with Bryce's
> suggestion of checking buffer size:
>
> http://gcc.gnu.org/ml/java-patches/2003-q3/msg00308.html
>
> ...because:
>
> - this would involve an additional call to JvGetStringUTFLength
>   for bounds checking

It is going to leave application code vunerable buffer overrun bugs in 
user code if they do not call JvGetStringUTFLength, so you are shifting 
the responsibility of doing this onto the user. If you are making the 
user do the JvGetStringUTFLength call then this new call doesn't make 
life much easier for anybody. I thought the point of it was to reduce 
the amount of code people have to write to convert strings?

> - JvGetStringUTFRegion does not do this bounds checking either

JvGetStringUTFRegion cannot be safely used without a corresponding call 
to JvGetStringUTFLength. If its being used somewhere without one then 
that is a serious bug.

> - the caller can call JvGetStringUTFLength beforehand if s/he wants to

They MUST do so. I'm sure your aware of what happens when code that 
fails to check bounds ends web servers and such. I'm sure someone will 
want to run a web server on GCJ some day.

> - the _Jv_TempUTFString class and JV_TEMP_UTF_STRING macro
>   (subsequent patch) will also shield us from this

Seems like these are two solutions to the same problem. Why do we need 
both?

Regards

Bryce