On Fri, Jan 23, 2026 at 10:59:23AM -0500, Patrick Robb wrote: > On Fri, Jan 23, 2026 at 10:43 AM Bruce Richardson > <[1][email protected]> wrote: > > On Fri, Jan 23, 2026 at 10:35:53AM -0500, Patrick Robb wrote: > > On Fri, Jan 23, 2026 at 4:02 AM Bruce Richardson > > <[1][2][email protected]> wrote: > > > > On Thu, Jan 22, 2026 at 01:59:19PM -0500, Patrick Robb wrote: > > > Thanks David for the CC. We need to rebuild all of our > DPDK > > > container images for the distro build testing etc. which > takes > > some > > > time. We should be ready early next week. > > > > > Is rebuilding all container images something that needs to be > done > > every > > time we do a minimum meson version bump? If so, we probably > need to > > look to > > take steps to make things easier to do. > > > > Right now we do build new images any time any DPDK dependencies > are > > updated, including meson. That's done with the dpdk-ci template > engine > > (an application that templates out Dockerfiles for DPDK CI > > testing): > [2][3]https://git.dpdk.org/tools/dpdk-ci/tree/containers > > However, to simplify we can just move to installing meson at > runtime in > > all cases. We already have to do this sometimes (like when > testing with > > an LTS that requires a different meson) so doing it in all > cases > > wouldn't be too disruptive. And, of course the main point is > that that > > would allow us to accomodate a version bump like this without > > rebuilding the CI images. Let me know if you have an opinion. > In any > > case I will chat with the CI Lab students about it. > > > Having it installed at runtime would be definitely good. > The other thing I was considering is whether we actually need to use > the > minimum meson version for all build testing. If one or two > containers have > the minimum version that should be fine, and others can use the > distro > supplied version so long as its >= minimum. > > Okay, both of the options above sound good. From my end if I had to > choose I would just do the first solution (install meson at runtime in > all cases) because it keeps our process uniform across distro testing > and there is no "special" container. But, the second option sounds good > too so please let me know what you think is best. >
Go with your runtime solution, it's fine - and possibly better.
