On 15/Jul/2009 09:38, Oliver Deakin wrote: > I committed the classlib part of this patch last week and am ready to > commit the working_jdktools part now. There are two issues I have come > across: > > 1) In trace output, the seconds part of the time stamp is garbled. I'm > not sure exactly what is causing this right now, but I get a feeling > there might be a problem in our str_ftime port library function. > 2) On Linux, I see a crash during shutdown when debugging. It appears > that as the JDWP agent is shutting down it starts freeing up some of the > resources it's been using, including the sockets, before other parts of > the agent are finished using them. We then make a > hysock_set_nonblocking() call with what appears to be a null/bad socket, > and this crashes having passed that socket to ioctl() [1]. It looks to > be a timing issue - I think the shutdown phase of the agent needs closer > inspection to make sure that resources are not freed while they are > still being used. > > Although these issues still exist, I would like to commit the code to > the Java 6 branch so I can continue working on fixes for them directly > in Harmony, rather than my local workspace, and also to allow others to > try out the new agent. Once committed, Ill raise a JIRA for both of > these issues to keep track of them. > > Does anyone have any comments/objections to my committing the code in > it's current state?
Go for it. Assuming it does not break any of the existing committer tests then you'll have time to fix it up before the next milestone. Regards, Tim
