On Mon, Sep 27, 2021 at 4:08 AM Peter Maydell <peter.mayd...@linaro.org> wrote:
> On Mon, 27 Sept 2021 at 10:59, Daniel P. Berrangé <berra...@redhat.com> > wrote: > > If we look at linux-user/meson.build though things are more complex. > > There we have alot of sub-dirs, and meson.biuld in those dirs adds > > generators for various files. So conceivably skipping linux-user > > will mean we won't auto-generate files we don't need on non-Linux. > > The top level meson.build doesn't call process on the > syscall_nr_generators[] list unless we're building linux-user > targets, so I don't think we will auto-generate anything we > don't need. > The problem is that I'm trying to add some os-specific files to the bsd-user in a patch set I sent off. bsd-user compiles for different BSDs, and I'd like to start pulling in per-bsd files as I'm upstreaming. To do that, I have this construct in the bsd-user/meson.build file: # Pull in the OS-specific build glue, if any if fs.exists(targetos) subdir(targetos) endif primarily because the builds failed on Linux when it tried to find the non-existent bsd-user/linunx directory. The question came up on how to cope with this situation, and Philippe sent off this series as an alternative to that construct. The whole reason why we descend into the linux-user directory on BSD and into the bsd-user directory on linux is unclear to me as well. So this is preparing the ground for my future work of upstreaming. I'm agnostic as to how it's resolved, but it needs to be resolved. Warner