Merge #131

131: Poll for completion on SPI write. r=therealprof a=newAM

Closes #130

---

Continuing discussion from #130 :

The suggested implementation was replacing this:
```rust
 self.check_send().ok(); 
 Ok(()) 
 ```
 
 With this:
 ```rust
 nb::block!(self.check_send())
 ```
 
The concern I have with that is it seems the last `check_send().ok()` was intended to clear the `OVR` error bit that may be set due to dropped reads.

Additionally, it does not seem that this works as intended, the CS still goes high before the write is completed with consecutive write calls.





Co-authored-by: Alex M <[email protected]>

Author: bors[bot]

src/spi.rs

@@ -400,7 +400,7 @@ where
             nb::Error::Other(Error::ModeFault)
         } else if sr.crcerr().bit_is_set() {
             nb::Error::Other(Error::Crc)
-        } else if sr.txe().bit_is_set() {
+        } else if sr.txe().bit_is_set() && sr.bsy().bit_is_clear() {
             return Ok(());
         } else {
             nb::Error::WouldBlock
@@ -482,7 +482,7 @@ where
         }
 
         // Do one last status register check before continuing
-        self.check_send().ok();
+        nb::block!(self.check_send()).ok();
         Ok(())
     }
 }
@@ -526,7 +526,7 @@ where
         }
 
         // Do one last status register check before continuing
-        self.check_send().ok();
+        nb::block!(self.check_send()).ok();
         Ok(())
     }
 }