On Mon, Jun 22, 2020 at 02:05:02PM -0500, Eric Blake wrote:
On 6/22/20 10:49 AM, Richard W.M. Jones wrote:
>VDDK creates background threads. fork kills these, resulting in the
s/kills/strands/
>library hanging or failing completely in certain configurations.
>---
> plugins/vddk/vddk.c | 28 ++++++++++++++++++++++------
> 1 file changed, 22 insertions(+), 6 deletions(-)
>
ACK.
Should we also modify our tests/libvixDiskLib.so dummy library to
also spawn a thread, where .pread then checks for something provided
by that helper thread, for testsuite coverage of this? That's a
rather complicated change to make, so I'm not insisting, but if
you're up to it, the more our dummy library behaves like the real
thing, the harder it is to introduce further regressions.
Yup that's a good idea.
Another problem with testing is that (real) VDDK either doesn't create
a thread or doesn't use a thread when it is accessing a local file, so
we never hit this case in our real VDDK test case. However testing
against a VMware server is not appropriate either.
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