a9f0f445ba
The end_sema k_sem was only initialized on one of the several paths that used it, leading to some crazy clobber-the-run-queue behavior that was dependent on linkage order (see the linked bug) when end_sema and the pipe object were made non-static.. Adding a k_sem_init() call fixes the corrupt issue, but really the right thing is to use the DEFINE macro, so do that instead. Note that that the initializer changes the linkage order too (by putting the semaphore in a separate segment), so... yeah, it's actually impossible to prove that this patch in isolation resolves the issue seen without manual validation. Issue: https://github.com/zephyrproject-rtos/zephyr/issues/4366 Signed-off-by: Andy Ross <andrew.j.ross@intel.com> |
||
|---|---|---|
| .. | ||
| main.c | ||
| test_pipe_contexts.c | ||
| test_pipe_fail.c | ||