On 09/03/2011 16:26, Alan Bateman wrote:
Chris Hegarty wrote:
Alan, Michael,
In Java_java_net_Inet(4|6)AddressImpl_getLocalHostName
JVM_GetHostName, a.k.a gethostname, is returning ENAMETOOLONG. Then
falling back to use "localhost", the assumption was that failure would
only happen when something went wrong, maybe networking is not setup.
It appears that on many Linux variations the given buffer length
should include space for the null terminator. Our given buffers
already account for this, just need to pass down the correct length to
gethostname.
http://cr.openjdk.java.net/~chegar/7024560/webrev.00/webrev/
-Chris.
Are you sure that it is failing with ENAMETOOLONG? I just checked the
I tested on a few different Linux machines and instrumented the error
handling code, it is definitely failing with ENAMETOOLONG.
man page and is reads "In case the null-terminated hostname does not
fit, no error is returned, but the hostname is truncated. It is
unspecified whether the truncated hostname will be null-terminated",
which suggests to me that we should be careful to null terminate on the
success path.
Yes, the man page does leave the door open for non null-terminated
strings, but we've never seen this in the past so I didn't think it
would be a problem in reality.
-Chris.
-Alan.
