Replace the static error string in ZIP_Put_In_Cache0 with on stack memory
Alan.Bateman at oracle.com
Wed Apr 11 13:09:05 PDT 2012
On 11/04/2012 15:58, Sean Chou wrote:
> Hi hotspot guys,
> Would any one like to take a look at this? I'm trying to fix a
> potential race in ZIP_Open, it is found classLoader.cpp uses this
> function. So a webrev for hotspot is made as well, but I need a
> sponsor from hotspot as suggested by Alan Bateman.
> The start of this thread is
> The webrevs:
> and http://cr.openjdk.java.net/~zhouyx/7159982/webrev.00/
> <http://cr.openjdk.java.net/%7Ezhouyx/7159982/webrev.00/> .
> And, I found JDK_Version::is_gte_jdk18x_version suggested by Alan
> does not exist in hotspot yet, shall I add it to the patch too ?
I had a brief chat with Coleen Phillimore about this. One suggestion
that would avoid HotSpot changes is to change ZIP_Open so that it
doesn't return an error message or returns something fixed. It's not
used in the VM anyway (as per Coleen's reply). We need the error message
for the java.util.zip APIs but that doesn't go through ZIP_Open so it
means it can be freed and *pmsg = NULL returned (or a pointer to one
fixed generic message).
More information about the hotspot-dev