On Fri, Nov 10, 2023 at 08:12:53AM -0600, Eric Blake wrote:
 On Fri, Nov 10, 2023 at 02:00:22PM +0000, Daniel P. Berrangé wrote:
 > > > 
 > > > Pipeline #1067529654 (
https://gitlab.com/nbdkit/libnbd/-/pipelines/1067529654 ) triggered by Eric Blake (
https://gitlab.com/ebblake )
 > > > had 1 failed job.
 > > > 
 > > > Job #5504447391 ( 
https://gitlab.com/nbdkit/libnbd/-/jobs/5504447391/raw
)
 > > > 
 > > > Stage: builds
 > > > Name: x86_64-fedora-rawhide-clang-prebuilt-env
 > > 
 > > This was a timeout in building the rust bindings on rawhide; unrelated
 > > to my patch (which was merely adding a new unit test).  Not sure why
 > > rust timed out; it looks like timing out eats the log file that would
 > > have let us see what the rust build was doing.
 > 
 > I'd suggest running  'make check V=1' in the CI system so it is always
 > printed
 
 The 'make' portion was already verbose; where we were silent was
 during the 'make check' portion.  Adding V=1 didn't make that more
 verbose; probably something we need to tweak in our rust/run-tests.sh
 script.
 
 As it is, I just noticed that Makefile.am passes CARGO=$(CARGO) in the
 TESTS_ENVIRONMENT, yet run-tests.sh hard-codes a call to 'cargo' on
 $PATH instead of using ${CARGO}. 
At any rate, it looks like the Rust timeout on rawhide is transient -
when I manually re-kicked that pipeline, it passed on the second
attempt.  I'm not sure how we can go about making things less likely
to hang with transient Rust failures.
-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.
Virtualization:  
qemu.org | 
libguestfs.org