On Fri, Jan 24, 2020 at 04:13:14PM +0000, Daniel P. Berrangé wrote:
> On Fri, Jan 24, 2020 at 03:01:51PM +0000, Daniel P. Berrangé wrote:
> > On Fri, Jan 24, 2020 at 02:51:39PM +0000, Richard W.M. Jones wrote:
> > > On Fri, Jan 24, 2020 at 02:35:01PM +0000, Daniel P. Berrangé wrote:
> > > > On Sun, Jan 19, 2020 at 01:04:07PM +0000, Richard W.M. Jones wrote:
> > > > > Basically OCaml 4.10 is much more strict about the difference between
> > > > > string (immutable) and bytes (mutable).
> > > > 
> > > > For all patches
> > > > 
> > > >   Reviewed-by: Daniel P. Berrangé <berra...@redhat.com>
> > > > 
> > > > I presume this will fix the rawhide CI failure we have with ocaml
> > > > now.
> > > 
> > > Well the previous code didn't compile with OCaml 4.10 at all, so I
> > > guess so.  However I wasn't previously aware we even had CI for this.
> > > Do you have a link?
> > 
> > Rawhide was succeeding until most recent build:
> > 
> >   https://ci.centos.org/view/libvirt/job/libvirt-ocaml-build/
> > 
> > which I presume matches when the VM got updated to ocaml 4.10
> 
> While it passes 4.10 and Fedora 30/31 with ocaml 4.9, it fails
> on CentOS 7 with 4.05 and Debian with 4.02

OCaml < 4.06 didn't have the C macro Bytes_val, but we can use
String_val instead with a bit of casting.  (The reason we can't just
use String_val in the code is because in OCaml >= 4.10 that macro was
changed to return a const as strings are now forced to be immutable).
I pushed the obvious patch:

https://libvirt.org/git/?p=libvirt-ocaml.git;a=commit;h=db1e05d99b3cb195b19d531a8832b980e155cb1f

That should fix CentOS 7 (I tested RHEL 7).

I'm fairly sure it will also fix Debian 9 (OCaml 4.02), but I don't
have that convenient at the moment.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-builder quickly builds VMs from scratch
http://libguestfs.org/virt-builder.1.html

Reply via email to