On Sun, Jul 16, 2023 at 04:10:56PM +0100, Richard W.M. Jones wrote:
The pipeline failed after you pushed this:
https://gitlab.com/nbdkit/libnbd/-/pipelines/932589424
but I think it's an unrelated OCaml failure. I'll take a proper look
at it tomorrow.
The error is:
FAIL: test_140_explicit_close.bc
================================
libnbd: debug: nbd1: nbd_create: opening handle
libnbd: debug: nbd1: nbd_set_debug: enter: debug=true
libnbd: debug: nbd1: nbd_set_debug: leave: ret=0
libnbd: debug: nbd1: nbd_set_debug_callback: enter: debug=<fun>
Fatal error: Fatal error during lock: Resource deadlock avoided
FAIL test_140_explicit_close.bc (exit status: 134)
FAIL: test_140_explicit_close.opt
=================================
libnbd: debug: nbd1: nbd_create: opening handle
libnbd: debug: nbd1: nbd_set_debug: enter: debug=true
libnbd: debug: nbd1: nbd_set_debug: leave: ret=0
libnbd: debug: nbd1: nbd_set_debug_callback: enter: debug=<fun>
Fatal error: Fatal error during lock: Resource deadlock avoided
FAIL test_140_explicit_close.opt (exit status: 134)
This is with OCaml 5.0 which was just added to Fedora Rawhide. This
version of OCaml is much more stringent about checking that locks are
used correctly.
The thing is that we fixed this already ...
https://gitlab.com/nbdkit/libnbd/-/commit/2c909bc3b8594943c4306ec78e8c133...
I am not able to reproduce this locally (with the same version of
OCaml), and the artifacts that we collect from the CI system do not
include coredumps or stack traces.
If anyone is able to reproduce this problem with the latest libnbd let
me know, especially if you can collect a stack trace.
Rich.
--
Richard Jones, Virtualization Group, Red Hat
http://people.redhat.com/~rjones
Read my programming and virtualization blog:
http://rwmj.wordpress.com
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW