Re: minutes 2006-11-14

[Stan Cox <scox@redhat.com>]

3381 teardown race (cmoller)
-should teardown be changed to be more robust
-what is the scope of the problem?
 -kills are asynchronous, so there is a race
 -not utrace dependent
 -buggy behavior is in buggy kernel
  -no need for teardown to handle buggy kernel
  -teardown needs to handle asynchronization
-frysk fires off kill/sigkill
 -don't use nohang, use wait without nohang
 -waitpid works for attached or unattached process
-frysk teardown does kill/detach/waitpid
 -detach means you are waiting on a process that isn't yours
-FC6 has SIGKILL bug because SIGKILL isn't specially handled
-hang bug may be a bug

3500 (Mark)
-possibly not a kernel issue but a timing issue

3492 (Nurdin)
-in progress
-stamping out bugs

Command Line Utilities
-fcore writes all segments and gcore doesn't, otherwise identical.  
 -FP regs are done.
 -next up is static analysis
 -how to handle a corefile in frysk
  -handle as part of current Host, just dead OR 
  -create a new Host
 -software layers will get registers and memory from the corefile
  instead of the task
  -32 on 64 with bytebuffer is a glitch
  -what about 2 level byte buffer for 32 on 64?
 -libelf bindings testings
-would be nice to use properties to get configuration info into frysk

source window
-Mike working on bugs
 -line stepping test
-Rick on jdom test cases
 -finding the right pieces can use $DATADIR
-dejagnu support added for "expect" testing

breakpoints
-Tim looking at Mike's stepping support to get events from core for hpd
-should this go into frysk.rt?
-look at gdb sig* testcases for edge cases

Open source java discussion

Changing the way observers are added
-currently observers are added explicitly
-is it better to add explicitly or implicitly?