Commit c79b1a38aa ("samples: display:
Convert driver and lvgl sample.yaml to use depends_on") started using
depends_on in the sample YAML for a lvgl application instead of
platform_allow.
This breaks build testing with twister and overflow checking enabled
on more resource constrained platforms. The test case's .config
ends up with:
CONFIG_LVGL_HOR_RES_MAX=320
CONFIG_LVGL_VER_RES_MAX=240
CONFIG_LVGL_VDB_SIZE=64
CONFIG_LVGL_BITS_PER_PIXEL=24
And lib/gui/lvgl/lvgl.c, where we allocate a buffer of size:
(CONFIG_LVGL_BITS_PER_PIXEL *
((CONFIG_LVGL_VDB_SIZE * CONFIG_LVGL_HOR_RES_MAX *
CONFIG_LVGL_VER_RES_MAX) / 100)
/ 8)
Require 147456 bytes to build the sample, ultimately overflowing RAM
if you run something along the lines of:
twister -T samples -p <constrained_platform> --overflow-as-errors
This is a reasonable test to be doing to make sure that sample RAM
requirements do not get too big for a subset of platforms that are of
interest, and it no longer works.
To fix it, add a min_ram line for this case so that we can still run
overflow tests on a large set of samples without fine-grained special
casing or creating an ever-growing list of platform excludes for this
test.
Signed-off-by: Martí Bolívar <marti.bolivar@nordicsemi.no>