review (XS) for 6765546: Wrong sscanf used to parse CompilerOracle command >= 32 characters could lead to crash

[Igor Veresov]

Thanks, Tom!

igor

On 12/14/10 4:45 PM, Tom Rodriguez wrote:
> Sure.  I added that in and tested to make sure array signatures were accepted now.  I've updated the webrev too.  Thanks.
>
> tom
>
> On Dec 14, 2010, at 3:54 PM, Igor Veresov wrote:
>
>> Looks good!
>>
>> May I also suggest piggybacking that on this change:
>>
>> @@ -489,7 +489,7 @@ void CompilerOracle::parse_from_line(cha
>>      line += bytes_read;
>>      // there might be a signature following the method.
>>      // signatures always begin with ( so match that by hand
>> -    if (1 == sscanf(line, "%*[ \t](%254[);/" RANGEBASE "]%n", sig + 1,&bytes_read)) {
>> +    if (1 == sscanf(line, "%*[ \t](%254[[);/" RANGEBASE "]%n", sig + 1,&bytes_read)) {
>>
>> This basically makes it accept signatures with arrays.
>>
>> Thanks,
>> igor
>>
>> On 12/14/10 3:30 PM, Tom Rodriguez wrote:
>>> http://cr.openjdk.java.net/~never/6765546
>>>
>>> 6765546: Wrong sscanf used to parse CompilerOracle command>= 32 characters could lead to crash
>>> Reviewed-by:
>>>
>>> The buffer for a sscanf isn't long enough to include the null
>>> termination and we're missing a check for unknown commands.  Tested
>>> with various command lines.
>>
>