These messages should be directed to stderr.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Ulf Magnusson <ulfalizer@gmail.com>

Pull Kconfig updates from Masahiro Yamada:
 "A pretty big batch of Kconfig updates.

  I have to mention the lexer and parser of Kconfig are now built from
  real .l and .y sources. So, flex and bison are the requirement for
  building the kernel. Both of them (unlike gperf) have been stable for
  a long time. This change has been tested several weeks in linux-next,
  and I did not receive any problem report about this.

  Summary:

   - add checks for mistakes, like the choice default is not in choice,
     help is doubled

   - document data structure and complex code

   - fix various memory leaks

   - change Makefile to build lexer and parser instead of using
     pre-generated C files

   - drop 'boolean' keyword, which is equivalent to 'bool'

   - use default 'yy' prefix and remove unneeded Make variables

   - fix gettext() check for xconfig

   - announce that oldnoconfig will be finally removed

   - make 'Selected by:' and 'Implied by' readable in help and search
     result

   - hide silentoldconfig from 'make help' to stop confusing people

   - fix misc things and cleanups"

* tag 'kconfig-v4.16' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (37 commits)
  kconfig: Remove silentoldconfig from help and docs; fix kconfig/conf's help
  kconfig: make "Selected by:" and "Implied by:" readable
  kconfig: announce removal of oldnoconfig if used
  kconfig: fix make xconfig when gettext is missing
  kconfig: Clarify menu and 'if' dependency propagation
  kconfig: Document 'if' flattening logic
  kconfig: Clarify choice dependency propagation
  kconfig: Document SYMBOL_OPTIONAL logic
  kbuild: remove unnecessary LEX_PREFIX and YACC_PREFIX
  kconfig: use default 'yy' prefix for lexer and parser
  kconfig: make conf_unsaved a local variable of conf_read()
  kconfig: make xfgets() really static
  kconfig: make input_mode static
  kconfig: Warn if there is more than one help text
  kconfig: drop 'boolean' keyword
  kconfig: use bool instead of boolean for type definition attributes, again
  kconfig: Remove menu_end_entry()
  kconfig: Document important expression functions
  kconfig: Document automatic submenu creation code
  kconfig: Fix choice symbol expression leak
  ...

Reverse dependency expressions can get rather unwieldy, especially if
a symbol is selected by more than a handful of other symbols. I.e. it's
possible to have near endless expressions like:
   A && B && !C || D || F && (G || H) || [...]

Chop these expressions into actually readable chunks:
   - A && B && !C
   - D
   - F && (G || H)
   - [...]

I.e. transform the top level OR tokens into newlines and prepend each
line with a minus. This makes the "Selected by:" and "Implied by:" blurb
much easier to read. This is done only if there is more than one top
level OR. "Depends on:" and "Range :" were deliberately left as they are.

Based on idea from Paul Bolle.

Suggested-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Petr Vorel <petr.vorel@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Many of these functions are quite the head scratchers if you don't know
what they're trying to do. Document them.

Also make it clear which functions rewrite expressions in-place and
which return new expressions. This prevents memory errors.

No functional changes. Only comments added.

Signed-off-by: Ulf Magnusson <ulfalizer@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Only the E_NOT operand and not the E_NOT node itself was freed, due to
accidentally returning too early in expr_free(). Outline of leak:

	switch (e->type) {
	...
	case E_NOT:
		expr_free(e->left.expr);
		return;
	...
	}
	*Never reached, 'e' leaked*
	free(e);

Fix by changing the 'return' to a 'break'.

Summary from Valgrind on 'menuconfig' (ARCH=x86) before the fix:

	LEAK SUMMARY:
	   definitely lost: 44,448 bytes in 1,852 blocks
	   ...

Summary after the fix:

	LEAK SUMMARY:
	   definitely lost: 1,608 bytes in 67 blocks
	   ...

Signed-off-by: Ulf Magnusson <ulfalizer@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Since commit 31847b67bec0 ("kconfig: allow use of relations other than
(in)equality") it is possible to use relational operators in Kconfig
statements. However, those operators give unexpected results when
applied to bool/tristate values:

	(n < y) = y (correct)
	(m < y) = y (correct)
	(n < m) = n (wrong)

This happens because relational operators process bool and tristate
symbols as strings and m sorts before n. It makes little sense to do a
lexicographical compare on bool and tristate values though.

Documentation/kbuild/kconfig-language.txt states that expression can have
a value of 'n', 'm' or 'y' (or 0, 1, 2 respectively for calculations).
Let's make it so for relational comparisons with bool/tristate
expressions as well and document them. If at least one symbol is an
actual string then the lexicographical compare works just as before.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Fixes: 31847b67bec0 ("kconfig: allow use of relations other than (in)equality")
Signed-off-by: Michal Sojka <sojkam1@fel.cvut.cz>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>

Over the years I found it desirable to be able to use all sorts of
relations, not just (in)equality. And apparently I'm not the only one,
as there's at least one example in the tree where the programmer
assumed this would work (see DEBUG_UART_8250_WORD in
arch/arm/Kconfig.debug). Another possible use would e.g. be to fold the
two SMP/NR_CPUS prompts into one: SMP could be promptless, simply
depending on NR_CPUS > 1.

A (desirable) side effect of this change - resulting from numeric
values now necessarily being compared as numbers rather than as
strings - is that comparing hex values now works as expected: Other
than int ones (which aren't allowed to have leading zeroes), zeroes
following the 0x prefix made them compare unequal even if their values
were equal.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>

expr_eliminate_dups2() in scripts/kconfig/expr.c applies two invalid
inference rules:

(FOO || BAR) && (!FOO && !BAR) -> n
(FOO && BAR) || (!FOO || !BAR) -> y

They would be correct in propositional logic, but this is a three-valued
logic, and here it is wrong in that it changes semantics. It becomes
immediately visible when assigning the value 1 to both, FOO and BAR:

(FOO || BAR) && (!FOO && !BAR)
-> min(max(1, 1), min(2-1, 2-1)) = min(1, 1) = 1

while n evaluates to 0 and

(FOO && BAR) || (!FOO || !BAR)
-> max(min(1, 1), max(2-1, 2-1)) = max(1, 1) = 1

with y evaluating to 2.

Fix it by removing expr_eliminate_dups2() and the functions that have no
use anywhere else: expr_extract_eq_and(), expr_extract_eq_or(),
and expr_extract_eq() from scripts/kconfig/expr.c

Currently the bug is not triggered in mainline, so this patch does not
modify the configuration space there. To observe the bug consider this
example:

config MODULES
        def_bool y
        option modules

config FOO
        def_tristate m

config BAR
        def_tristate m

config TEST1
        def_tristate y
        depends on (FOO || BAR) && (!FOO && !BAR)

if TEST1 = n
comment "TEST1 broken"
endif

config TEST2
        def_tristate y
        depends on (FOO && BAR) || (!FOO || !BAR)

if TEST2 = y
comment "TEST2 broken"
endif

config TEST3
        def_tristate y
        depends on m && !m

if TEST3 = n
comment "TEST3 broken"
endif

TEST1, TEST2 and TEST3 should all evaluate to m, but without the patch,
none of them does. It is probably not obvious that TEST3 is the same bug,
but it becomes clear when considering what happens internally to the
expression
m && !m":
First it expands to
(m && MODULES) && !(m && MODULES),
then it is transformed into
(m && MODULES) && (!m || !MODULES),
and finally due to the bug it is replaced with n.

As a side effect, this patch reduces code size in expr.c by roughly 10%
and slightly improves startup time for all configuration frontends.

Signed-off-by: Martin Walch <walch.martin@web.de>
Signed-off-by: Michal Marek <mmarek@suse.cz>

Signed-off-by: Michal Marek <mmarek@suse.cz>