I never needed to (my new/delete implementations "cheat" some), but I guess > you could do something like the thing Linus figured out for some commonly > used kernel macros, like the user memory access function with area checking. > These use a form of exception handling to eliminate the checking overhead > from the main execution path. > > I'm not sure about what the compiler expects to find, though; probably > something similar to setjmp() and longjmp(). The gcc is expecting some DWARF unwind infos > A simple hack might be to just terminate the current thread if you run out of > memory, and then call a clean-up callback assigned to it. (Obviously, this > isn't very useful unless running out of memory really *is* fatal to the > thread - and if it is, you probably shouldn't be doing dynamic allocation in > the first place.) I though it's more a problem of syncing of threads why not allocate memory inside threads (C and C++) on the side of linux kernel ? Some days ago I attached an archive with new/delete, rtti and throwing exceptions. Unfortunally I was not able to catch it. The main part of the last is a rip of the gcc sources. It seems, there is no interesst since I got no response on it. I miss the deep knowlegde about this due to my profession: electronics and controlling. Regards Olaf -- [rtl] --- To unsubscribe: echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED] -- For more information on Real-Time Linux see: http://www.rtlinux.org/rtlinux/
