Running `oldconfig' after any of the following configuration change:
either trivial addition, such as:
config A
bool "A"
choice
prompt "Choice ?"
depends on A
config CHOICE_B
bool "Choice B"
config CHOICE_C
bool "Choice C"
endchoice
or more tricky change:
OLD KCONFIG | NEW KCONFIG
|
| config A
| bool "A"
|
choice | choice
prompt "Choice ?" | prompt "Choice ?"
|
config CHOICE_C | config CHOICE_C
bool "Choice C" | bool "Choice C"
|
config CHOICE_D | config CHOICE_D
bool "Choice D" | bool "Choice D"
endchoice |
| config CHOICE_E
| bool "Choice E"
| depends on A
| endchoice
will not cause the choice to be considered as NEW, and thus not be
asked. The cause of this behavior is that choice's novelty are computed
statically right after the saved configuration has been read. At this
point, the new dependency's value is still unknown and asserted to be
`no'. Moreover, no update to this decision is made afterward.
Correct this by dynamically evaluating a choice's novelty, and removing the
static evaluation.
Reported-and-tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Make conf_write_symbol() grammar agnostic to be able to use it from different
code path. These path pass a printer callback which will print a symbol's name
and its value in different format.
conf_write_symbol()'s job become mostly only to prepare a string for the
printer. This avoid to have to pass specialized flag to generic
functions
Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
[mmarek: rebased on top of de12518 (kconfig: autogenerated config_is_xxx
macro)]
Signed-off-by: Michal Marek <mmarek@suse.cz>
* 'kconfig' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild-2.6:
nconf: handle comment entries within choice/endchoice
kconfig: fix warning
kconfig: Make expr_copy() take a const argument
kconfig: simplify select-with-unmet-direct-dependency warning
kconfig: add more S_INT and S_HEX consistency checks
kconfig: fix `zconfdebug' extern declaration
kconfig/conf: merge duplicate switch's case
kconfig: fix typos
kbuild/gconf: add dummy inline for bind_textdomain_codeset()
kbuild/nconf: fix spaces damage
kconfig: nuke second argument of conf_write_symbol()
kconfig: do not define AUTOCONF_INCLUDED
kconfig: the day kconfig warns about "select"-abuse has come
This is an attempt to simplify the expressing printed by kconfig when a
symbol is selected but still has direct unmet dependency.
First, the symbol reverse dependency is split in sub-expression. Then,
each sub-expression is checked to ensure that it does not contains the
unmet dependency. This removes the false-positive symbols and fixed symbol
which already have the correct dependency. Finally, only the symbol
responsible of the "select" is printed, instead of its full dependency tree.
CC: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Arnaud Lacombe <lacombar@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
When expanding a parameterised string we may run out of space, this
triggers a realloc. When computing the new allocation size we do not
allow for the terminating '\0'. Allow for this when calculating the new
length.
Signed-off-by: Andy Whitcroft <apw@canonical.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After fixing a use-after-free bug in kconfig, a 'make defconfig' or
'make allmodconfig' fills the screen with warnings that were not
detected before. Given that we are close to the release now, disable the
warnings temporarily and deal with them after 2.6.36.
Signed-off-by: Michal Marek <mmarek@suse.cz>
Following sample Kconfig generated a segfault:
config FOO
bool
select PERF_EVENTS if HAVE_HW_BREAKPOINT
config PERF_EVENTS
bool
config HAVE_HW_BREAKPOINT
bool
depends on PERF_EVENTS
Fix by reverting back to a valid property if there was no
property on the stack of symbols.
The above pattern were seen in sh Kconfig.
A fix for the Kconfig file has been sent to the sh folks.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
savedefconfig will save a minimal config to a file
named "defconfig".
The config symbols are saved in the same order as
they appear in the menu structure so it should
be possible to map them to the relevant menus
if desired.
The implementation was tested against several minimal
configs for arm which was created using brute-force.
There was one regression related to default numbers
which had their valid range further limited by another symbol.
Sample:
config FOO
int "foo"
default 4
config BAR
int "bar"
range 0 FOO
If FOO is set to 3 then BAR cannot take a value higher than 3.
But the current implementation will set BAR equal to 4.
This is seldomly used and the final configuration is OK,
and the fix was non-trivial.
So it was documented in the code and left as is.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Move logic to determine default for a choice to
a separate function.
No functional changes.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Consider following kconfig file:
config TEST1
bool "test 1"
depends on TEST2
config TEST2
bool "test 2"
depends on TEST1
Previously kconfig would report:
foo:6:error: found recursive dependency: TEST2 -> TEST1 -> TEST2
With the following patch kconfig reports:
foo:5:error: recursive dependency detected!
foo:5: symbol TEST2 depends on TEST1
foo:1: symbol TEST1 depends on TEST2
Note that we now report where the offending symbols are defined.
This can be a great help for complex situations involving
several files.
Patch is originally from Roman Zippel with a few adjustments by Sam.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
When we add a new config symbol save the file/line
so we later can refer to their location.
The information is saved as a property to a config symbol
because we may have multiple definitions of the same symbol.
This has the side-effect that a symbol always has
at least one property.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
This makes it so "make oldconfig" really prompts for any choice where
options that previously weren't visible just became so. Previously one
would have to remember to go over all choice values and check whether
some that previously couldn't be selected now can be.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
The "select" statement in Kconfig files allows the enabling of options
even if they have unmet direct dependencies (i.e. "depends on" expands
to "no"). Currently, the "depends on" clauses are used in calculating
the visibility but they do not affect the reverse dependencies in any
way.
The patch introduces additional tracking of the "depends on" statements
and prints a warning on selecting an option if its direct dependencies
are not met.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
A symbol's value won't be recalc-ed until we save config file or
enter the menu where the symbol sits.
So If I enable OPTIMIZE_FOR_SIZE, and search FUNCTION_GRAPH_TRACER:
Symbol: FUNCTION_GRAPH_TRACER [=y]
Prompt: Kernel Function Graph Tracer
Defined at kernel/trace/Kconfig:140
Depends on: ... [=y] && (!X86_32 [=y] || !CC_OPTIMIZE_FOR_SIZE [=y])
...
From the dependency it should result in FUNCTION_GRAPH_TRACER=n,
but it still shows FUNCTION_GRAPH_TRACER=y.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
While looking for something else I noticed that the symbol
hash function used by kconfig is quite poor. It doesn't
use any of the standard hash techniques but simply
adds up the string and then uses power of two masking,
which is both known to perform poorly.
The current x86 kconfig has over 7000 symbols.
When I instrumented it showed that the minimum hash chain
length was 16 and a significant number of them was over
30.
It didn't help that the hash table size was only 256 buckets.
This patch increases the hash table size to a larger prime
and switches to a FNV32 hash. I played around with a couple of hash
functions, but that one seemed to perform best with reasonable
hash table sizes.
Increasing the hash table size even further didn't
seem like a good idea, because there are a couple of global
walks which walk the complete hash table.
I also moved the unnamed bucket to 0. It's still the longest
of all the buckets (44 entries), but hopefully it's not
often hit except for the global walk which doesn't care.
The result is a much nicer distribution:
(first column bucket length, second number of buckets with that length)
1: 3505
2: 1236
3: 294
4: 52
5: 3
47: 1 <--- this is the unnamed symbols bucket
There are still some 5+ buckets, but increasing the hash table
even more would be likely not worth it.
This also cleans up the code slightly by removing hard coded
magic numbers.
I didn't notice a big performance difference either way
on my Nehalem system, but I presume it'll help somewhat
on slower systems.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
As choice dependency are now fully checked, it's quite easy to add support
for named choices. This lifts the restriction that a choice value can only
appear once, although it still has to be within the same group,
but multiple choices can be joined by giving them a name.
While at it I cleaned up a little the choice type logic to simplify it a
bit.
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Properly check the dependency of choices as a group.
Also fix that sym_check_deps() correctly terminates the dependency loop
error check (otherwise it would continue printing the dependency chain).
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
> The attached .config (with current -git) results in a compile
> error since it contains:
>
> CONFIG_X86=y
> # CONFIG_EMBEDDED is not set
> CONFIG_SERIO=m
> CONFIG_SERIO_I8042=y
>
> Looking at drivers/input/serio/Kconfig I simply don't get how this
> can happen.
You've hit the rather subtle rules of select vs default. What happened is
that SERIO is selected to m, but SERIO_I8042 isn't selected so the default
of y is used instead.
We already had the problem in the past that select and default don't work
well together, so this patch cleans this up and makes the rule hopefully
more straightforward. Basically now the value is calculated like this:
(value && dependency) || select
where the value is the user choice (if available and the symbol is
visible) or default.
In this case it means SERIO and SERIO_I8042 are both set to y due to their
default and if SERIO didn't had the default, then the SERIO_I8042 value
would be limited to m due to the dependency.
I tested this patch with more 10000 random configs and above case is the
only the difference that showed up, so I hope there is nothing that
depended on the old more complex and subtle rules.
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Tested-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
