This prevents segfault when getting filename and lineno in recursive
checks.

If the following snippet is found in Kconfig:

[Test code 1]

config FOO
        bool
        depends on BAR
        select BAR

... without BAR defined; then there is a segfault.

  Kconfig:34:error: recursive dependency detected!
  Kconfig:34:	symbol FOO depends on BAR
  make[4]: *** [scripts/kconfig/Makefile:85: allnoconfig] Segmentation fault

This is because of the following. BAR is a fake entry created by
sym_lookup() with prop being NULL. In the recursive check, there is a
NULL check for prop to fall back to stack->sym->prop if stack->prop is
NULL. However, in this case, stack->sym points to the fake BAR entry
created by sym_lookup(), so prop is still NULL. prop was then referenced
without additional NULL checks, causing segfault.

As the previous email thread suggests, the file and lineno for select is
also wrong:

[Test code 2]

config FOO
       bool

config BAR
       bool

config FOO
       bool "FOO"
       depends on BAR
       select BAR

  $ make defconfig
  *** Default configuration is based on 'x86_64_defconfig'
  Kconfig:1:error: recursive dependency detected!
  Kconfig:1: symbol FOO depends on BAR
  Kconfig:4: symbol BAR is selected by FOO
  [...]

Kconfig:4 should be Kconfig:10.

This patch deletes the wrong and segfault-prone filename/lineno
inference completely. With this patch, Test code 1 yields:

error: recursive dependency detected!
	symbol FOO depends on BAR
	symbol BAR is selected by FOO

Signed-off-by: HONG Yifan <elsk@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

I previously submitted a fix for a bug in the choice feature [1], where
I mentioned, "Another (much cleaner) approach would be to remove the
tristate choice support entirely".

There are more issues in the tristate choice feature. For example, you
can observe a couple of bugs in the following test code.

[Test Code]

    config MODULES
            def_bool y
            modules

    choice
            prompt "tristate choice"
            default A

    config A
            tristate "A"

    config B
            tristate "B"

    endchoice

Bug 1: the 'default' property is not correctly processed

'make alldefconfig' produces:

    CONFIG_MODULES=y
    # CONFIG_A is not set
    # CONFIG_B is not set

However, the correct output should be:

    CONFIG_MODULES=y
    CONFIG_A=y
    # CONFIG_B is not set

The unit test file, scripts/kconfig/tests/choice/alldef_expected_config,
is wrong as well.

Bug 2: choice members never get 'y' with randconfig

For the test code above, the following combinations are possible:

               A    B
        (1)    y    n
        (2)    n    y
        (3)    m    m
        (4)    m    n
        (5)    n    m
        (6)    n    n

'make randconfig' never produces (1) or (2).

These bugs are fixable, but a more critical problem is the lack of a
sensible syntax to specify the default for the tristate choice.
The default for the choice must be one of the choice members, which
cannot specify any of the patterns (3) through (6) above.

In addition, I have never seen it being used in a useful way.

The following commits removed unnecessary use of tristate choices:

 - df8df5e4bc37 ("usb: get rid of 'choice' for legacy gadget drivers")
 - bfb57ef0544a ("rapidio: remove choice for enumeration")

This commit removes the tristate choice support entirely, which allows
me to delete a lot of code, making further refactoring easier.

Note:
This includes the revert of commit fa64e5f6a35e ("kconfig/symbol.c:
handle choice_values that depend on 'm' symbols"). It was suspicious
because it did not address the root cause but introduced inconsistency
in visibility between choice members and other symbols.

[1]: https://lore.kernel.org/linux-kbuild/20240427104231.2728905-1-masahiroy@kernel.org/T/#m0a1bb6992581462ceca861b409bb33cb8fd7dbae

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>

The 'choice' statement is primarily used to exclusively select one
option, but the 'optional' property allows all entries to be disabled.

In the following example, both A and B can be disabled simultaneously:

    choice
            prompt "choose A, B, or nothing"
            optional

    config A
            bool "A"

    config B
            bool "B"

    endchoice

You can achieve the equivalent outcome by other means.

A common solution is to add another option to guard the choice block.
In the following example, you can set ENABLE_A_B_CHOICE=n to disable
the entire choice block:

    choice
            prompt "choose A or B"
            depends on ENABLE_A_B_CHOICE

    config A
            bool "A"

    config B
            bool "B"

    endchoice

Another approach is to insert one more entry:

    choice
            prompt "choose A, B, or disable both"

    config A
            bool "A"

    config B
            bool "B"

    config DISABLE_A_AND_B
            bool "choose this to disable both A and B"

    endchoice

Some real examples are DEBUG_INFO_NONE, INITRAMFS_COMPRESSION_NONE,
LTO_NONE, etc.

The 'optional' property is even more unnecessary for a tristate choice.

Without the 'optional' property, you can disable A and B; you can set
'm' in the choice prompt, and disable A and B individually:

    choice
            prompt "choose one built-in or make them modular"

    config A
            tristate "A"

    config B
            tristate "B"

    endchoice

In conclusion, the 'optional' property was unneeded.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nicolas Schier <n.schier@avm.de>

Commit c8fb7d7e48d1 ("kconfig: fix broken dependency in randconfig-
generated .config") fixed the issue, but I did not add a test case.

This commit adds a test case that emulates the reported situation.
The test would fail without c8fb7d7e48d1.

To handle the choice "choose X", FOO must be calculated beforehand.
FOO depends on A, which is a member of another choice "choose A or B".
Kconfig _temporarily_ assumes the value of A to proceed. The choice
"choose A or B" will be shuffled later, but the result may or may not
meet "FOO depends on A". Kconfig should invalidate the symbol values
and recompute them.

In the real example for ARCH=arm64, the choice "Instrumentation type"
needs the value of CPU_BIG_ENDIAN. The choice "Endianness" will be
shuffled later.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Since commit 3b9a19e08960 ("kconfig: loop as long as we changed some
symbols in randconfig"), conf_set_all_new_symbols() is repeated until
there is no more choice left to be shuffled. The motivation was to
shuffle a choice nested in another choice.

Although commit 09d5873e4d1f ("kconfig: allow only 'config', 'comment',
and 'if' inside 'choice'") disallowed the nested choice structure,
we must still keep 3b9a19e08960 because there are still cases where
conf_set_all_new_symbols() must iterate.

scripts/kconfig/tests/choice_randomize/Kconfig is the test case.
The second choice depends on 'B', which is the member of the first
choice.

With 3b9a19e08960 reverted, we would never get the pattern specified by
scripts/kconfig/tests/choice_randomize/expected_config2.

A real example can be found in lib/Kconfig.debug. Without 3b9a19e08960,
the randconfig would not shuffle the "Compressed Debug information"
choice, which depends on DEBUG_INFO, which is derived from another
choice "Debug information".

My goal is to refactor Kconfig so that randconfig will work more
simply, without using the loop.

For now, let's add a test case to ensure all dependent choices are
shuffled, as it is a somewhat tricky case for the current Kconfig.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

This will help get consistent results for randconfig tests.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

for_all_symbols() iterates in the symbol hash table. The order of
iteration depends on the hash table implementation.

If you use it for printing errors, they are shown in random order.

For example, the order of following test input and the corresponding
error do not match:
 - scripts/kconfig/tests/err_recursive_dep/Kconfig
 - scripts/kconfig/tests/err_recursive_dep/expected_stderr

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Now "modules" is the only member of the "option" property.

Remove "option", and move "modules" to the top level property.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

"defconfig_list" is a weird option that defines a static symbol that
declares the list of base config files in case the .config does not
exist yet.

This is quite different from other normal symbols; we just abused the
"string" type and the "default" properties to list out the input files.
They must be fixed values since these are searched for and loaded in
the parse stage.

It is an ugly hack, and should not exist in the first place. Providing
this feature as an environment variable is a saner approach.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Nesting choice statements does not make any sense.

Commit df8df5e4bc37 ("usb: get rid of 'choice' for legacy gadget
drivers") got rid of the only usecase.

I will turn it into a syntax error. Remove the test in advance.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>