New Firmware Release Candidate v1.20.0

[brotherdust]

@iwahdan
Getting build errors on release-candidate:HEAD (2/27/19):

LINK build/WIPY/release/application.elf
lib/libbt.a(bt.o):(.literal.btdm_check_and_init_bb+0x4): undefined reference to `btdm_rf_bb_init_phase2'
lib/libbt.a(bt.o): In function `btdm_check_and_init_bb':
/Users/iwahdan/esp/pycom-esp-idf/components/bt/bt.c:1165: undefined reference to `btdm_rf_bb_init_phase2'
collect2: error: ld returned 1 exit status
make: *** [application.mk:632: build/WIPY/release/application.elf] Error 1

Any thoughts here?
Thanks!

[crumble]

@brotherdust

Any thoughts here?

Do you use an old esp-idf branch? The pycom-esp-idf master is an old version used by the stable firmware. If you want to compile 1.20 you have to switch to v3.1.something. I am on the 3.1.1 branch.

[Xykon]

@brotherdust The official IDF version for release-candidate branch is idf_v3.1 (see Jenkinsfile)

Also make sure to update submodules in both the IDF and the firmware repo (see README.md)

[robert-hh]

@xykon When switching between master and release-candidate branch, one not only has to change the branches in esp-idf (and not forget to do the submodule update), but also mpy-cross has to be rebuilt.
What I do:

• start with cleaning both the esp32 target and the mpy-cross target,
• then switch branches & update submodules,
• build mpy-cross
• build the ESP32 binary

This is all a little bit tedious, so I avoid to do it.

[brotherdust]

@robert-hh, @Xykon : Thanks for the tips! Changing from idf v3.1.1 to 3.1 did the trick!

[snock]

It looks like UART.readall() is missing therefore uModbus does not work.

>>> print(dir(UART))
['__class__', '__name__', 'EVEN', 'ODD', 'any', 'deinit', 'init', 'read', 'readinto', 'readline', 'sendbreak', 'wait_tx_done', 'write']
>>> print(os.uname())
(sysname='GPy', nodename='GPy', release='1.20.0.rc7', version='v1.9.4-2833cf5 on 2019-02-08', machine='GPy with ESP32')

[robert-hh]

@snock uart.read() without argument does the same as uart.readall(). But it would have been better to keep readall() as an alias, especially since it is still documented.

[Xykon]

@robert-hh said in New Firmware Release Candidate v1.20.0:

This is all a little bit tedious, so I avoid to do it.

I remember we had a discussion about this on GitHub around Christmas last year...

Ideally the Makefile should know which branch of pycom-esp-idf should be used and clone / checkout / submodule update as necessary. It could also run make -C ../mpy-cross all before building the firmware as it should only rebuild if necessary. However checking all this before every build would probably be very annoying if it increases the build time significantly especially if you just made a small change between builds. Maybe adding this as make prepare would be better... plus a small post git checkout script that creates a file .force_prepare that would tell make to run the prepare step first so you don't forget.

Of course you can always keep two separate firmware directories, one for the master / stable and one for the release-candidate branch. You can also have a shallow clone of the IDF in each directory (git clone --depth=1 --recursive -b [master | idf_v3.1] https://github.com/pycom/pycom-esp-idf.git esp-idf) and then use a relative IDF_PATH (export IDF_PATH=../esp-idf).

We also plan to change the way we use branches in this repository. Rather than master being the latest stable release, we will create a branch for each major release version. If we did this today, we would have branches called release/v1.18 for stable and release/v1.20 for the release candidate. The master branch would always have the latest code so we can publish hot fixes without creating a full new release for the software server and firmware updater.

[iwahdan]

This post is deleted!

[JamieLanger]

Was in the watch out for the below upgrades lately. Glad
Added support for setting mac addr for STA and AP
Added API for sending Raw 80211 messages