John,

> File Descriptor Leak:
> http://cr.openjdk.java.net/~chegar/8000203/webrev.00/
> (Jira bug ID number: 8000203)

(*it's not to your changes but as far as you touch this code *)

607  else is not needed here

609  realloc should not touch original pointer in case of fail,
     so this code leads to hidden memmory leak

     its better to do:

       newLoRoutes = realloc(
       if (newLoRoutes == NULL){

         // What you do here depends to whether you
         // plan to keep incomplete table or not.

          free(loRoutes);
          break;
       }


-Dmitry


On 2012-09-28 18:11, John Zavgren wrote:
> Greetings:
> 
> I just posted the webrev images for two networking code bugs:
> 
> File Descriptor Leak:
> http://cr.openjdk.java.net/~chegar/8000203/webrev.00/
> (Jira bug ID number: 8000203)
> 
> Uninitialized memory:
> http://cr.openjdk.java.net/~chegar/8000206/webrev.00/
> (Jira bug ID: 8000206)
> This change doesn't actually fix a bug... the original code initialized 
> "optlen" before it was referenced, however, parfait (static code analysis) 
> believes "optlen" MAY be used before initialization. I added the assignment 
> statement to "spoof" parfait, and it no longer flags a bug. I assume it's 
> better to put minor harmless tweaks in our code than to add state information 
> to parfait, that would cause it to ignore certain "situations". That option 
> seems complicated and dangerous.
> 
> Thanks!
> John Zavgren
> john.zavg...@oracle.com
> 


-- 
Dmitry Samersoff
Java Hotspot development team, SPB04
* There will come soft rains ...

Reply via email to