Tools: Testbench: Change file IPC3 init to be process compatible #9532
Conversation
|
@singalsu in the future we could probably get rid of type completely for IPC4. IPC3 could have a lookup able that converts type to UUID for module lookup. |
This allows the component load to pass the type check. The existing components in process.c with module adapter API are changed to load with type SOF_COMP_MODULE_ADAPTER. The new components DRC, MULTIBAND_DRC, and TDFB they are also added to be able to run them in testbench with the type check in place. Signed-off-by: Seppo Ingalsuo <[email protected]>
This change is needed to change file component type to SOF_COMP_MODULE_ADAPTER to be able to handle the file configuration without exceptions in comp_specific_builder(). Signed-off-by: Seppo Ingalsuo <[email protected]>
Fixing Volume and SRC to pass the new component type check would require large changes to IPC, possibly in kernel side too so these tests are left out for now with IPC3 testbench. The test can be restored with IPC4 testbench for all components after it is ready for use. Signed-off-by: Seppo Ingalsuo <[email protected]>
singalsu
force-pushed
the
testbench_fix_file_comp_type
branch
from
86f2e8c
to
185dbaa
Compare
What do you think of this version? The testbench CI step works (should -- checked so far only locally) but I've had to remove Volume and SRC component tests since there is currently no solution to load them to IPC3 testbench. |
singalsu
requested review from
ranj063,
RanderWang,
marcinszkudlinski,
pblaszko,
lgirdwood,
plbossart,
mmaka1,
lbetlej,
dbaluta and
kv2019i
as code owners
|
@lgirdwood @cujomalainey #9518 revert is now merged so this is no more needed. I'll close this for now since this patch adds no value. But I have the work in my git tree so we can re-introduce the check later as better prepared. |
|
@singalsu please keep the work around as the fix is going to be applied, but we need to fix stuff before it lands. |
OK, I'll keep the draft open, so also other can use these changes. |
I'm having doubts for this approach to avoid revert of #9518. There might be need to change some component's IPC structures, e.g. volume, to be able to initialize with process type IPC as SOF_COMP_MODULE_ADAPTER.