[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: nptl 0.5



Hong Zhang wrote:
- pthread_kill_other_threads_np (used by Sun JRE)

I don't see any reason to break code that could otherwise work.


Sun JRE assumes a lot of internal knownledge about linuxthreads.
The code will not work in anyway under NPTL, period.

Well, actually most of the dependency on LinuxThreads *has* been removed during the past two years. We no longer assume fixed stack layout, process forker/reaper code has been completely rewritten. Things like using non-public interface or fast get TLS are all gone. We do have several workarounds still in place for various LinuxThreads problems, but they were written in such a way that they don't rely on internal knowledge about the threading library and we believe they are benign under NPTL (but less efficient). (I can't say "all" now because I haven't got the VM working yet under NPTL, but I do believe the problem I'm seeing is due to my setup not VM.)

That being said, I don't think pthread_kill_other_threads_np
is needed if abort and core dump can work reliably in multi-threaded
app. I don't see a reason why we need the function if abort()
doesn't hang VM.

regards,
-hui

FYI: I am currently responsible for the HotSpot Linux port.





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]