On Tue, Nov 04, 2025 at 07:10:54AM +0800, Sune Brian wrote: > > Because ".config" isn't valid, you need to have $(objdir)/.config be > > checked, and in turn that has to depend on the appropriate target (we > > have some stage that says that we are configured, I forget which). > > ``` > make clean && make mrproper > > make > *** > *** Configuration file ".config" not found! > *** > *** Please run some configurator (e.g. "make oldconfig" or > *** "make menuconfig" or "make xconfig"). > *** > make: *** No rule to make target '.config', needed by > 'include/config/auto.conf'. Stop. > ``` > > Still not getting this, sorry a bit.
You need to do some builds with "make O=/tmp/somedir". But also, I'm trying to make sure we avoid race conditions and that we don't try this logic until after the step where we know we have configured the system. > > > > spin with "make O=/tmp/somedir". It also looks like it's going to impact > > > > every board, and either fail or add some build time? Make sure both say > > > > "sandbox" and "qemu_arm" both build at least, if not a full CI run. > > > > > > For build time -> There could be 4 line of code to check. > > > > I do forget off-hand which methods of doing "something" in a Makefile > > are cheap and which result in firing off a subshell. But for a handy > > feature for one SoC I want to make sure we don't slow down say the world > > build of all 1400 configurations. For a maybe useful example, see how > > the "checkgcc6" target is handled, and maybe we can move all of this to > > live under arch/arm/mach-socfpga/config.mk and so only get invoked at > > all there. > > If config.mk can also do this job why not just move it. Yes, I'm saying move this logic to that file where I think it should be workable, still. > > Brian -- Tom
signature.asc
Description: PGP signature
