On Sat, 07 Nov 2026, [email protected] wrote: > > > > +checkconfig := $(PYTHON3) -B $(srctree)/tools/kconfig-checker/check.py > > +list_checkconfig_rules = \ > > + echo -n >&2 "Available RULES: "; \ > > + $(checkconfig) list-rules >&2; \ > > + echo >&2 > > + > > +PHONY += checkconfig > > +checkconfig: > > + $(Q)if [ -z "$(RULES)" ]; then \ > > + echo >&2 "Usage: make RULES=<profile> checkconfig"; \ > > + $(list_checkconfig_rules); \ > > + exit 2; \ > > + fi > > + $(Q)if ! $(checkconfig) list-rules | grep -qw "$(RULES)"; then \ > > + echo >&2 "Unknown RULES profile = $(RULES)"; \ > > + $(list_checkconfig_rules); \ > > + exit 2; \ > > + fi > > + $(Q)$(checkconfig) check -r "$(RULES)" -c "$(KCONFIG_CONFIG)" > ... > > When either module is missing, check.py writes the error to stderr, > produces no stdout, and exits with os.EX_UNAVAILABLE. Because the POSIX > shell pipeline ignores the exit status of $(checkconfig) list-rules, > grep -qw exits 1 on the empty stdout and the negated condition enters the > branch. That reports a valid setting such as RULES=preempt_rt as an > unknown profile and runs $(list_checkconfig_rules), which invokes check.py > a second time and prints a duplicate module error on the same line after > "Available RULES: ".
It works, so there is no reason to complicate things. For example, invoking the make target from a container with old Python: $ make checkconfig Usage: make RULES=<profile> [OVERLAY=<overlay>] checkconfig Available RULES: Error: Python 3.9 or later is required make[2]: *** [scripts/kconfig/Makefile:138: checkconfig] Error 2 make[1]: *** [/home/darwi/linux/Makefile:772: checkconfig] Error 2 make: *** [Makefile:248: __sub-make] Error 2 and: $ make RULES=nonExistent checkconfig Error: Python 3.9 or later is required Unknown RULES profile = nonExistent Available RULES: Error: Python 3.9 or later is required make[2]: *** [scripts/kconfig/Makefile:139: checkconfig] Error 2 make[1]: *** [/home/darwi/linux/Makefile:772: checkconfig] Error 2 make: *** [Makefile:248: __sub-make] Error 2 The duplicate error, which is a single line, is no big deal. IMO, it's better than complicating the Makefile for that single edge case. > > In addition, grep -qw treats $(RULES) as a regular expression with word > boundaries against the comma-separated output of list-rules rather than > matching an exact profile name, and spawns Python twice on every valid > invocation even though check.py check -r already validates the profile > name. > That's good catch; will fix. Thanks, Ahmed

