lib: libc: Kconfig: Correct dependencies for MINIMAL_LIBC_SUPPORTED #81410
Conversation
The MINIMAL_LIBC_SUPPORTED option is supposed to indicate that the minimal C library implementation can be chosen and many tests use this option as a filter for scenarios where the MINIMAL_LIBC option is enabled. On the other hand, the REQUIRES_FULL_LIBC option tells that the minimal C library is insufficient and it prevents the MINIMAL_LIBC option from being enabled. But it should also cause that MINIMAL_LIBC_SUPPORTED is no longer enabled, otherwise for targets requiring a complete C library implementation, tests using the filtering mentioned above will not be actually performed with the minimal C library. This commit removes the above discrepancy by moving the dependency on !REQUIRES_FULL_LIBC from MINIMAL_LIBC to MINIMAL_LIBC_SUPPORTED. Signed-off-by: Andrzej Głąbek <[email protected]>
anangl
requested review from
vanwinkeljan,
tejlmand,
carlescufi,
mmahadevan108 and
aescolar
zephyrbot
added
area: C Library
zephyrbot
requested review from
cfriedt,
keith-packard,
nashif and
stephanosio
| @@ -28,6 +28,7 @@ config FULL_LIBC_SUPPORTED | |||
| config MINIMAL_LIBC_SUPPORTED | |||
| bool | |||
| depends on !NATIVE_APPLICATION | |||
| depends on !REQUIRES_FULL_LIBC | |||
There was a problem hiding this comment.
I don't think this is correct. MINIMAL_LIBC_SUPPORTED indicates the target itself supports the library.
REQUIRES_FULL_LIBC indicates an application or feature requires a full libc. Which has nothing to do with the target supporting or not minimallibc.
There was a problem hiding this comment.
Well, nrf9131ek/nrf9131/ns is an example of a target that selects REQUIRES_FULL_LIBC (because of NPM1300_CHARGER) and this generates a build warning when the libraries.libc.common.minimal scenario is executed for the tests/lib/c_lib/common test.
There was a problem hiding this comment.
should that board kconfig.defconfig have a
config MINIMAL_LIBC_SUPPORTED
default n
?
There was a problem hiding this comment.
This would only solve the problem for this one board, but similar scenario may occur for others. I wanted to provide a generic solution and I thought that such dependency would be fine, as I saw that PICOLIBC_SUPPORTED has dependency on !REQUIRES_FULL_LIBCPP:
Lines 45 to 48 in 5f418f5
There was a problem hiding this comment.
That pico one is different, as you can't build from the source module today if you need the c++ std lib.
Anyhow, to me it seems we should just set MINIMAL_LIBC_SUPPORTED to n for that board, and others which may just not be able to be built with minimal.
But if others think otherwise I'm not going to oppose this change.
The MINIMAL_LIBC_SUPPORTED option is supposed to indicate that the minimal C library implementation can be chosen and many tests use this option as a filter for scenarios where the MINIMAL_LIBC option is enabled.
On the other hand, the REQUIRES_FULL_LIBC option tells that the minimal C library is insufficient and it prevents the MINIMAL_LIBC option from being enabled. But it should also cause that MINIMAL_LIBC_SUPPORTED is no longer enabled, otherwise for targets requiring a complete C library implementation, tests using the filtering mentioned above will not be actually performed with the minimal C library.
This commit removes the above discrepancy by moving the dependency on !REQUIRES_FULL_LIBC from MINIMAL_LIBC to MINIMAL_LIBC_SUPPORTED.