Hi everybody,
I've noticed that commit 4213609cc7fb78f84b2ea63f4a5691b60d01c248
changes how `uclass.c:uclass_find_device_by_name` function compares the
names for finding devices.
Because of this I have an issue when having a DTS and trying to find a
device by name.
The `lists.c:lists_bind_fdt` function will bind a device to a device
node tree and use the node name as device name.
```
lists.c:
int lists_bind_fdt(..)
...
name = ofnode_get_name(node);
...
```
Because of this devices will end up having names like: i2c@021a0000,
pfuze100@8...
When using `uclass.c:uclass_find_device_by_name` to get a device, before
the change, this was not an issue since `strncmp` was used with the
length of the name making a call like
`uclass_find_device_by_name(pfuze100)` successfully even if the device
name was `pfuze100@8`.
```
if (!strncmp(dev->name, name, strlen(name))) {
```
However after the change (which I think is correct), the check fails,
resulting in not finding the expected device, since the device name has
also the suffix `@...` included.
What's the proper fix here:
1. Changing the `lists.c:lists_bind_fdt` to drop the `@` suffix?
Note: Could this be a problem when having the same node names but
different addressees in DTS {e.g.foo@1, foo@2}
2. When using `uclass.c:uclass_find_device_by_name` the user should take
in consideration that `@..` suffix needs to be added.
Note: This could be a problem when same code is used with different
DTSs.
3. Changing the comparation in `uclass.c:uclass_find_device_by_name` to
ignore the part after `@` for device names.
4. What else?
Regads,
Nandor
_______________________________________________
U-Boot mailing list
[email protected]
https://lists.denx.de/listinfo/u-boot
