Document potential regression bug introduced to embedded_hal SpiBus implementation in v1.0.0-rc.0

[renkenono]

Hi,

Thank you for your amazing work on esp-hal!

I recently updated my project to esp-hal v1.0.0-rc.0 using esp-generate and I noticed that my SPI-based logic no longer works. It relies on embedded_hal v1.0.0 to call embedded_hal::spi::SpiBus::transfer_inplace() which seems to be implemented differently from esp_hal::spi::master::Spi::transfer(). The fix was adding a missing call to setup full duplex communication

diff --git a/esp-hal/src/spi/master.rs b/esp-hal/src/spi/master.rs
index 18754fae..f31d3a6b 100644
--- a/esp-hal/src/spi/master.rs
+++ b/esp-hal/src/spi/master.rs
@@ -2759,6 +2759,7 @@ mod ehal1 {
         }
 
         fn transfer_in_place(&mut self, words: &mut [u8]) -> Result<(), Self::Error> {
+            self.driver().setup_full_duplex()?;
             self.driver().transfer(words)
         }

This was already fixed in the current state of the main branch, see

esp-hal/esp-hal/src/spi/master.rs

Line 2678 in d0a2d49

fn transfer_in_place(&mut self, words: &mut [u8]) -> Result<(), Self::Error> {

. However, I have yet to try it because updating my project would require some time.

The reason I created a discussion instead of an issue is because it's already fixed, I am not sure what the best approach would be here. My aim is to simply document this issue so that other users of embedded_hal would be aware of it and seek an alternative solution instead.

With the current state of esp-hal's main branch, I don't see the point in adding a test for this since the API is very well abstracted now.

Again, thanks!