jtag serial drive write function loses characters when rx ring buffer… (IDFGH-10925)

[jjsch-dev]

upon examining the code, I noticed that the usbjtag_tx_char_via_driver function contains a comparison that checks whether an immediate send operation succeeds. However, the return value from usb_serial_jtag_write_bytes is always 1, which prevents the intended logic to try blocking for 50 mS, but also a return that always skyp the bloking try.

[CLAassistant]

All committers have signed the CLA.

[ginkgm]

Hi @jjsch-dev

usb_serial_jtag_write_bytes never returning 0 is a real issue. We will fix this.

The API will return 0 when write fails. Change this to -1 is a breaking change that we can't do.

From the VFS side, losing data when ringbuffer full is a expected behavior. We can't detect the connection to host, so we use the fifo overflow as indication.

[jjsch-dev]

Hi @ginkgm, thanks for the explanation, if you want I'll close the PR.

[jjsch-dev]

Hi @ginkgm, I modified the return of usb_serial_jtag_write_bytes to return 0 on failure and the size on success, and reverted the change to usbjtag_tx_char_via_driver.

[chipweinberger]

@jjsch-dev thanks for creating the PR!

[jjsch-dev]

@chipweinberger, thanks for the comment.

[mythbuster5]

@jjsch-dev Yes, this time changes is absolutely correct, I already have an internal PR to fix this, and that will be merged soon. Thanks for your contribution.

[jjsch-dev]

@mythbuster5, thanks for the confirmation, I will wait the merge.