we have to add a patch to usb-device because the SpinalHDL UDC
needs a little extra handholding on EP0 to switch IN and OUT modes
(vs e.g. STM32 where EP0 has IN and OUT descriptors).
a small bug was found in the device core that is now fixed. the
device core was not computing the target for the SETUP transaction
correctly if the HEAD of the OUT descriptor was set in advance of
the SETUP transaction. we now see much more consistent SETUP
transactions, and even the first IN response.
the next challenge is figuring out why the second IN response doesn't
seem to be issued.
we're now two packets into the setup.
i think i see one fundamental mismatch between how the driver
behaves and how the hardware runs. it has to do with where the
ep0 data gets deposited for a SETUP or OUT transaction --
the HEAD of the list can't be allocated the transaction is
completely ready to go.
will bang on this more tomorrow, but this is progress at least.
interrupts seem to behave much more like i think they should.
i think the interrupt unit on the spinalHDL block is not made to
have the whole set enabled/disabled frequently, it just wants
to have the bits banged to clear the interrupts. enabling and
disabling interrupts as a whole causes some erratic operations.
NB: the reference linux driver only enables them once and never
turns them off.
also, realized just now that OUT is from the direction of the host.
so OUT on the device is incoming data. oops.
things are moving in the right direction, i'm seeing the next
IN packet being issued by the USB driver on the precursor
side, but it's not being sent by the hardware. yet.
change the notification modal to accept the primary message and an
Option of independent text to display as a qrcode
pub fn show_notification(&self, notification: &str, qrtext:Option<&str>) -> Result<(), xous::Error> {}
Instead of having a single, shared buffer, simply stack-allocate the log
buffers.
This paves the way in the future to allow for buffers that are
larger-sized and span multiple pages. But most importantly it alleviates
inter-thread locking and contention when passing messages.
Signed-off-by: Sean Cross <sean@xobs.io>
Previously, we would reject buffers that were the maximum allowed size.
This is reported in #141.
Allow these maximally-sized buffers to be printed, since they don't end
up causing a problem.
Signed-off-by: Sean Cross <sean@xobs.io>
- track field heights using the drawing height directly
- use an &mut reference in the text processing section so changes stick
- other minor cosmetic improvements