The recent of-platdata implementation has caused an occasional error in CI, possibly due to the transition between !OF_PLATDATA_INST and OF_PLATDATA_INST.
The problems seems to be due to a generated file (C or object file) not being regenerated when the setting changes.
This series take two steps aimed at correct this problem.
Firstly it makes the set of files generated (in each case) mutually exclusive, except for the header files which remain common. This means that the build will fail if new files are not generated when the setting changes.
Secondly it removes the old generated files before building new ones, since that could trip things up if the flag changes back again in a subsequent build.
In addition, dtoc is currently running on every of-platadata build, even if nothing has changed. Also pylibfdt is always built due to a change in file naming with Python 3. Both of these problems are fixed.
Changes in v2: - Adjust tag so this patch doesn't got to dtc list - Add a comment to the endif - Only remove the old (and to be unused) files, so avoid confusing make
Simon Glass (6): libfdt: Tidy up pylibfdt build rule Makefile: Avoid running dtoc every time Makefile: Depend only on required of-platdata files dtoc: Only generate the required files Makefile: Use a variable for generated of-platdata headers Makefile: Remove old of-platdata files before regenerating
scripts/Makefile.spl | 43 ++++++++++++++++++++++++----------- scripts/dtc/pylibfdt/Makefile | 8 +++++-- tools/dtoc/dtb_platdata.py | 23 +++++++++++++++---- tools/dtoc/test_dtoc.py | 2 +- 4 files changed, 56 insertions(+), 20 deletions(-)