apps: fix distclean for partial builds and residual .kconfig files

[Zepp-Hanzj]

Summary

Directory.mk's distclean only traverses CLEANSUBDIRS — directories with .built, .depend, or .kconfig markers. When a build fails before these markers are written (e.g. Python configure succeeds but compilation fails), distclean skips the directory entirely, leaving behind build/, config.site, and other artifacts.

Fix: use SUBDIRS (all directories with Makefiles) for distclean. It is idempotent, so calling it on un-built directories is harmless.

Changes

File	Change
Directory.mk	distclean traverse SUBDIRS instead of CLEANSUBDIRS; inline target definitions with TOPDIR; - prefix for error resilience
Application.mk	distclean: add $(call DELFILE, .kconfig)
audioutils/lame/Makefile	guard cd $(DST_PATH) against already-deleted directory
crypto/wolfssl/Makefile	distclean: -> distclean:: (conflicted with Application.mk)

Companion PR

Requires nuttx#19310 which passes TOPDIR in SDIR_template / MAKE_template.

Verification

The bug: when a build fails between context:: and depend:: phases, .built and .depend markers are never written. The old distclean uses CLEANSUBDIRS which requires these markers, so it skips the directory and leaves artifacts behind.

To reproduce, simulate a partial build by creating the residual files a failed Python configure would leave, without .built/.depend markers:

$ cd ~/workspace/nuttx-apps/interpreters/python
$ rm -rf build install .kconfig .built .depend
$ mkdir -p build/target/{Modules,Objects,Parser,Python}
$ touch build/target/{Makefile,pyconfig.h,config.log}
$ touch config.site Setup.local .kconfig
$ test -f .built && echo "exists" || echo "not exists"
not exists
$ test -f .depend && echo "exists" || echo "not exists"
not exists

Run distclean through the NuttX build system (note: apps is a symlink to nuttx-apps):

$ cd ~/workspace/nuttx
$ make -C apps/interpreters/python distclean TOPDIR=$(pwd) APPDIR=$(pwd)/apps V=1
make: Entering directory '/home/hanzj-mi/workspace/nuttx-apps/interpreters/python'
rm -rf  /home/hanzj-mi/workspace/nuttx-apps/interpreters/python/build
rm -rf  /home/hanzj-mi/workspace/nuttx-apps/interpreters/python/install
rm -rf  Python
rm -f  v3.13.0.zip
rm -f  romfs_cpython_modules.img
rm -f  romfs_cpython_modules.h
rm -f  config.site
rm -f  Setup.local
rm -f  .built
rm -f *.o *.a *~ .*.swp  /home/hanzj-mi/workspace/nuttx/apps/libapps.a /home/hanzj-mi/workspace/nuttx/apps/libapps.a.lock
rm -f  .kconfig
rm -f  Make.dep
rm -f  .depend
make: Leaving directory '/home/hanzj-mi/workspace/nuttx-apps/interpreters/python'

Verify all residuals are cleaned:

$ cd ~/workspace/nuttx-apps/interpreters/python
$ test ! -d build && echo "build/ cleaned" || echo "build/ still exists"
build/ cleaned
$ test ! -f config.site && echo "config.site cleaned" || echo "config.site still exists"
config.site cleaned
$ test ! -f Setup.local && echo "Setup.local cleaned" || echo "Setup.local still exists"
Setup.local cleaned
$ test ! -f .kconfig && echo ".kconfig cleaned" || echo ".kconfig still exists"
.kconfig cleaned

Without this patch, make distclean produces no output (CLEANSUBDIRS is empty, nothing is traversed) and all residuals remain.

Fixes #2895

[Zepp-Hanzj]

Hi, the CI failure on Linux (arm-02) appears to be a timeout issue rather than a build failure caused by this PR.

Looking at the arm-02 logs:

• The job ran for ~2h27m before being terminated
• All builds that completed showed successful Normalize (no compilation errors)
• The testbuild.sh script was cut short mid-way through the testlist (last completed: teensy-lc/nsh)
• The cat: x264/config.h: No such file or directory and choosenim: command not found messages are pre-existing CI environment issues, not related to this PR
• All other failing jobs show cancelled status (cascade from arm-02)

Could someone re-run the CI? The timeout seems to be an infrastructure issue. Thank you!

[xiaoxiang781216]

let extend SDIR_template instead duplicating the code