[PATCH v2] drm/panel: himax-hx8279: Always initialize goa_{even,odd}_valid in hx8279_check_goa_config()

[AngeloGioacchino Del Regno]

Il 23/04/25 19:41, Nathan Chancellor ha scritto:
> Clang warns (or errors with CONFIG_WERROR=y):
> 
>    drivers/gpu/drm/panel/panel-himax-hx8279.c:838:6: error: variable 'goa_even_valid' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
>      838 |         if (num_zero == ARRAY_SIZE(desc->goa_even_timing))
>          |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    drivers/gpu/drm/panel/panel-himax-hx8279.c:842:23: note: uninitialized use occurs here
>      842 |         if (goa_odd_valid != goa_even_valid)
>          |                              ^~~~~~~~~~~~~~
>    drivers/gpu/drm/panel/panel-himax-hx8279.c:838:2: note: remove the 'if' if its condition is always true
>      838 |         if (num_zero == ARRAY_SIZE(desc->goa_even_timing))
>          |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>      839 |                 goa_even_valid = false;
>    drivers/gpu/drm/panel/panel-himax-hx8279.c:818:36: note: initialize the variable 'goa_even_valid' to silence this warning
>      818 |         bool goa_odd_valid, goa_even_valid;
>          |                                           ^
>          |                                            = 0
> 
> Even though only the even valid variable gets flagged, both valid
> variables appear to have the same issue of possibly being used
> uninitialized if the if statement initializing them to false is not
> taken.
> 
> Turn the if statement then variable assignment into a single variable
> assignment, which states that the configuration is valid when there are
> not all zeros, clearing up the warning since the variable will always be
> initialized.
> 
> Fixes: 38d42c261389 ("drm: panel: Add driver for Himax HX8279 DDIC panels")
> Suggested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno at collabora.com>
> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno at collabora.com>
> Signed-off-by: Nathan Chancellor <nathan at kernel.org>

Actually, I prefer this one, as Arnd's commit fixes only one of the two, but this
initializes both :-)

Cheers,
Angelo