On Sat, Oct 17, 2026 at 02:32:09PM +0000, [email protected] wrote: >> +dry-run := $(findstring n,$(firstword -$(MAKEFLAGS))) > >[Severity: Medium] >Could this dry-run detection logic produce a false positive for common make >arguments? > >If MAKEFLAGS starts with a long option containing 'n' (such as Kbuild's >frequent default --no-print-directory), -$(MAKEFLAGS) evaluates to something >like ---no-print-directory, and findstring erroneously matches the letter 'n'. [...] >Is it possible to use the standard Kbuild idiom here, like >$(findstring n,$(filter-out --%,$(MAKEFLAGS))), to properly check the flags?
MAKEFLAGS does not start with a long option. With make 4.0 and later, which the kernel requires, its first word holds the single letter options and the long options follow after a space. With no single letter options, -$(MAKEFLAGS) starts with "- --no-print-directory" and the first word is "-". The top level Makefile detects -s the same way. Under kbuild the first word is "-rR", as kbuild adds -rR to MAKEFLAGS. I checked with make 4.4.1, calling Makefile.kbuild from a parent that adds -rR --no-print-directory: dry-run is empty for a plain, -j, -k and -s build, and is "n" only with -n. -- Thanks, Sasha
