Hello Bruce, On Tue, Jul 18, 2023 at 11:10 AM Bruce Richardson <bruce.richard...@intel.com> wrote: > > On Thu, Jun 29, 2023 at 11:39:53AM +0200, David Marchand wrote: > > On Wed, Jun 28, 2023 at 12:19 PM David Marchand > > <david.march...@redhat.com> wrote: > > > > > > On Fri, Jun 23, 2023 at 5:07 PM Bruce Richardson > > > <bruce.richard...@intel.com> wrote: > > > > > > > > While the build system will skip building most libs and drivers when a > > > > dependency is missing for a component, for DLB2 driver, the > > > > "static_rte_eventdev" object is referenced inside the meson.build file > > > > itself, which will cause crashes if it doesn't exist i.e. if eventdev is > > > > disabled. Prevent this issue by skipping processing the file if no > > > > eventdev. [The build system will still report missing dependency, as the > > > > dependency is set by default for all eventdev drivers] > > > > > > > > Signed-off-by: Bruce Richardson <bruce.richard...@intel.com> > > > > > > Could we evaluate the class "std_deps" before jumping to each driver > > > meson.build? > > > > Hum, with my suggestion, we lose the opportunity for drivers to > > rewrite completely their "deps". > > I doubt we have cases where it really matters, but if this revealed to > > be necessary, such driver may be directly referenced in > > drivers/meson.build like we do for common/cnxk & friends. > > > > To illustrate the idea, I pushed your series along patches of mine > > (target is v23.11) in my github repo. > > https://github.com/david-marchand/dpdk/commit/enable_libs~8 > > > I don't think we should need to worry too much about the case of a driver > needing to rewrite deps. The standard deps should really just be the > minimal deps for a class. > > However, for the implementation of that check, I'm not sure I like the > approach of checking std_deps in a loop for each driver. Firstly, the logic > inside the per-driver block is already pretty complicated, but secondly, I > think we can skip that loop entirely if the standard deps for a class are > not met. I think instead it would be better to process std_deps value once > immediately after the "subdir(class)" line, and if not all deps are > present, just loop through all the driver directories for that class > immediately and mark them as unbuildable, before just moving on to the next > class.
Skipping this loop makes the change smaller and the logic in drivers/meson.build easier to understand. We lose the exhaustive summary for disabled content (before, every disabled driver was logged with the associated reason), but we can still log the reason why a class is disabled. Your suggestion looks good to me, let me send a patch. -- David Marchand
