[BUG] cmake savedefconfig failure after https://github.com/apache/nuttx/pull/13815

[xuxin930]

Description / Steps to reproduce the issue

Important

cmake savedefconfig failure after #13815

Steps to reproduce the problem

1. config with CMake

cmake -B build -DBOARD_CONFIG=qemu-armv7a:nsh

2. change any config with menuconfig

cmake --build build -t menuconfig
# For example  Disable CONFIG_EXAMPLES_HELLO

3. savedefconfig use cmake

cmake --build build -t savedefconfig

Issue: Neither the BINARY directory nor the defconfig in the config directory has been modified.
Expect: Both are modified to the latest

This issue did not exist before this PR: #13815

On which OS does this issue occur?

[OS: Linux]

What is the version of your OS?

Ubuntu 20.04

NuttX Version

12.5.1

Issue Architecture

[Arch: all]

Issue Area

[Area: Build System], [Area: Kconfig]

Verification

• I have verified before submitting the report.

[xuxin930]

hi @simbit18 can you take a look at this issue?

[simbit18]

hi @xuxin930 sure !!!

[simbit18]

@xuxin930 Clarification

3. savedefconfig use cmake
cmake --build build -t savedefconfig
Issue: Neither the BINARY directory nor the defconfig in the config directory has been modified.
Expect: Both are modified to the latest

Why should savedefconfig replace the defconfig file in the boards/arm/qemu-armv7a/configs/nsh path?
Its task (as in compiling with make) should only be to recreate the defconfig file in the cmake buid path.
https://github.com/ulfalizer/Kconfiglib/blob/master/savedefconfig.py

Perhaps you should have a target with a different name like replacedefconfig.

It is also better to rename the current savedefconfig with refreshsilent

[xuxin930]

yes @simbit18
Thank you for your reminder. You are right.
We just need to ensure that the defconfig saved in BINARY_DIR is what we expect.

[simbit18]

@xuxin930 I also created replacedefconfig simbit18@2061d3a
Do I need it or remove it?

[xuxin930]

@xuxin930 I also created replacedefconfig simbit18@2061d3a Do I need it or remove it?

I think rename is unnecessary, as it has different semantics from Kconfig.
How about keeping the name savedefconfig?

[simbit18]

@xuxin930 OK then I rename current savedefconfig to refreshsilent
and my replacedefconfig to savedefconfig confirm ?

[xuxin930]

This seems to solve the problem I reported.

because after modifying the config through menuconfig on the Makefile base,
then executing savedefconfig, we can get the modified defconfig

[simbit18]

OK I add PR.
#14401

[raiden00pl]

Why should savedefconfig replace the defconfig file in the boards/arm/qemu-armv7a/configs/nsh path?
Its task (as in compiling with make) should only be to recreate the defconfig file in the cmake buid path.
https://github.com/ulfalizer/Kconfiglib/blob/master/savedefconfig.py

@simbit18 look at this PR discussion: #11416
FYI @anchao

[xuxin930]

Why should savedefconfig replace the defconfig file in the boards/arm/qemu-armv7a/configs/nsh path?
Its task (as in compiling with make) should only be to recreate the defconfig file in the cmake buid path.
https://github.com/ulfalizer/Kconfiglib/blob/master/savedefconfig.py

@simbit18 look at this PR discussion: #11416 FYI @anchao

thanks @raiden00pl
our original intention is to be consistent with the behavior of Makefile.
That is, only save the defconfig of BINARY_DIR, not the original defconfig of the board.

[simbit18]

@raiden00pl
refreshsilent (PR #14401) behaves with the same logic as the tools/refresh.sh script used for the refresh stage (make build) on GitHub.

See https://github.com/apache/nuttx/blob/master/tools/refresh.sh

example nucleo-l152re/nsh

./tools/refresh.sh --silent nucleo-l152re/nsh

• rm -f SAVEconfig
• mv .config SAVEconfig
• cp -a boards/arm/stm32/nucleo-l152re/configs/nsh/defconfig .config
• make olddefconfig
• make savedefconfig
• diff boards/arm/stm32/nucleo-l152re/configs/nsh/defconfig defconfig

has been merge with this PR #13815

without generating this error on github
#13815 (comment)