Re: [PATCH 1/3] perf build: Build the Python extension before pylint consumers

[Leo Yan <leo.yan@arm.com>]

On Wed, Sep 30, 2026 at 09:15:25AM +0000, sashiko-bot@kernel.org wrote:

[...]

> > +ifdef PYLINT
> > +# Build the Python extension before starting sub-makes that lint its users.
> > +$(PERF_IN) $(LIBPERF_TEST_IN): | $(LANG_BINDINGS)
> 
> [Severity: Low]
> Does this order-only dependency block the $(PERF_IN) sub-make from executing
> until $(LANG_BINDINGS) and all its prerequisites (including $(LIBPERF_UTIL))
> are fully built?
> 
> $(PERF_IN) launches the sub-make that compiles all builtin-*.c files.
> Meanwhile, $(LANG_BINDINGS) depends on $(PERFLIBS_PY), which launches the
> sub-make that compiles all util/*.c files.
> 
> Does this completely serialize the compilation of the builtin/ directory
> behind the full compilation of the util/ directory and the Python extension,
> impacting concurrent parallel build efficiency?


> Furthermore, does it force an isolated make perf to needlessly build the
> Python extension when PYLINT is enabled on the host?

Fair point. I will update the Makefile to separate pylint from C
compilation, this can avoid dependency within C files, and can avoid
the dependency between perf binary and python extension.