Hello Gerhard.
There is another issue for those who do
make HOSTCC=${CROSS_COMPILE}gcc env (needs a HOSTSTRIP spec as well in recent mainline sources)
to get an fw_printenv(1) binary that runs on the target (in contrast to the build machine). Some yocto recipes do this, to not re-invent how to build this tool or how to read and write the environment image.
With your changes, the 'env' make target is gone, instead there is a 'tools-all' target but it has a wider scope, includes the BMP manipulation/conversion stuff and uses HOSTCC to build a bmp_logo(1) tool, which breaks in the mentioned use case (cross compiled bmp_logo(1) executable, empty bmp_logo.h header, aborted build sequence).
I did not know how "env" target have been used. Your comments are highly appreciated!
I will revive "env" target at v5.
But whan I cannot understand is why HOSTCC must be tweaked. (I am also asking this question in "Compiling fw_printenv tool" thread)
If fw_printenv must be always compiled with cross-tools, I'd like to suggest to use $(CC) instead of $(HOSTCC) in tools/env/Makefile.
While testing your series I noticed a probably missing dependency: Running something different from 'make all' (or make without a target spec) after 'make <board>_config' won't work since the fixdeps(1) tool is missing. It's a "byproduct" of 'make all', afterwards other targets can get built separately. Is there a "preparation" step that one needs to take if not calling 'make all' immediately after configuration?
Good catch! I will fix this problem at v5.
Best Regards Masahiro Yamada