On 10/19/20 9:46 AM, Pavel Begunkov wrote:
> close(dup(io_uring)) should not neccessary cancel all requests with
> files, because files are not yet going away. Test that it doesn't hang
> after close() and exits, that's enough.
Applied with the below incremental.
diff --git a/test/lfs-openat.c b/test/lfs-openat.c
index 3fa0b99db1b3..b14238a9a41d 100644
--- a/test/lfs-openat.c
+++ b/test/lfs-openat.c
@@ -133,10 +133,9 @@ static int test_drained_files(int dfd, const char *fn, bool linked, bool prepend
{
struct io_uring ring;
struct io_uring_sqe *sqe;
- struct io_uring_cqe *cqe;
char buffer[128];
struct iovec iov = {.iov_base = buffer, .iov_len = sizeof(buffer), };
- int i, ret, fd, fds[2], to_cancel = 0;
+ int ret, fd, fds[2], to_cancel = 0;
ret = io_uring_queue_init(10, &ring, 0);
if (ret < 0)
--
Jens Axboe
