On Fri, Jul 17, 2020 at 04:18:52PM +0200, Peter Krempa wrote:
On Fri, Jul 17, 2020 at 16:04:16 +0200, Peter Krempa wrote:
> On Thu, Jul 16, 2020 at 11:53:56 +0200, Pavel Hrdina wrote:
>
> I've tried building RPMs both from the pre-patch tree state and after
> your patchset. I've then extracted all the RPMs toghether and compared
> the file lists (minus the 'build-id' directory which differs).
>
> I'm not sure whether it's due to the RPM or regular build process
> though.
I've also started to verify file differences:
config files which are installed in /etc/libvirt from said RPMs were not
post-processed.
E.g. virtnetworkd.conf:
The meson-originated version has:
# Set the name of the directory in which sockets will be found/created.
#
# This setting is not required or honoured if using systemd socket
# activation with systemd version >= 227
#
#unix_sock_dir = "@runstatedir@/libvirt"
while automake:
# Set the name of the directory in which sockets will be found/created.
#
# This setting is not required or honoured if using systemd socket
# activation with systemd version >= 227
#
#unix_sock_dir = "/run/libvirt"
So it's actually the other way around. I just checked it and with meson
it is expanded correctly. With autotools we have @runstatedir@ and
others in the config files. I also checked the installed files in my
fedora system and they also contain not-expanded strings.
So the meson rewrite fixes the issue that we have with autotools.
So I guess no need to do anything here.
Pavel