Re: RFR(xs): 8150379: [windows] Fix Leaks in perfMemory_windows.cpp

[David Holmes]

On 24/02/2016 12:37 AM, Thomas Stüfe wrote:

Hi David,

thanks for the review!

new webrev:
http://cr.openjdk.java.net/~stuefe/webrevs/8150379-fix-leaks-perfMemory_windows_cpp/webrev.01/webrev/

Looks good. I can sponsor this for you.

comments inline.

On Tue, Feb 23, 2016 at 2:24 AM, David Holmes <david.hol...@oracle.com
<mailto:david.hol...@oracle.com>> wrote:

    Hi Thomas,

    On 23/02/2016 12:43 AM, Thomas Stüfe wrote:

        Hi all,

        please take a look at this small fix for two leaks
        in perfMemory_windows.cpp.

        bug report: https://bugs.openjdk.java.net/browse/JDK-8150379

        webrev:
        
http://cr.openjdk.java.net/~stuefe/webrevs/8150379-fix-leaks-perfMemory_windows_cpp/webrev.00/webrev/


    Fixes look good, but there seems to be additional leakage here:

    1437   // get the name of the user associated with this process
    1438   char* user = get_user_name();
    1439
    1440   if (user == NULL) {
    1441     return NULL;
    1442   }
    1443
    1444   // construct the name of the user specific temporary directory
    1445   char* dirname = get_user_tmp_dir(user);
    1446
    1447   // check that the file system is secure - i.e. it supports ACLs.
    1448   if (!is_filesystem_secure(dirname)) {
    1449     return NULL;
    1450   }

    We need to free user and dirname before returning. These allocation
    methods are crying out for some RIIA assistance. :(


I fixed this place too and took a look at other functions returning C
heap array but did not find anything more. I also played around with an
RAII object, but the change got big quickly, so I decided to keep the
change small. I agree this is a mess. If we wanted to clean up this

Ok.

coding, I would do this as a separate issue. Also, instead of doing a
RAII wrapper, could we not just move some allocations to resource area?

I'm unclear on the allocation lifecycles in this code.

Thanks,
David

Thomas

    Thanks,
    David

        Thank you,

        Thomas