What steps will reproduce the problem?
1. compile the example from the Project Home, link with -lconcurrency
2. run the example
3.
What is the expected output? What do you see instead?
The context should have switched to the writer upon coro_call(cw, r);
What version of the product are you using? On what operating system?
svn r31, Linux x86_64, glibc 2.11.3
Please provide any additional information below.
I don't think the setjmp logic works at all. longjmp will always stay inside
coro_call() because there's also the setjmp that goes with it.
setjmp doesn't record the stack of the previous frame but the current one; and
that's coro_call().
Original issue reported on code.google.com by
sfreu...@gmail.comon 1 Nov 2013 at 10:46