so, the migration to async webserver is done, but the OTA update stack-dumps on me. :-(
clemens@secrets:~/Devel/ESP8266/genesys$ make stack
0x402010cf: __wrap_realloc at ??:?
0x40104c5c: lmacTxFrame at ??:?
0x4020ffcc: yield at ??:?
0x4020f76d: UpdaterClass::_writeBuffer() at ??:?
0x4020f81f: UpdaterClass::write(unsigned char*, unsigned int) at ??:?
0x402102da: String::String(char const*) at ??:?
0x4020de24: handle_update_progress_cb(AsyncWebServerRequest*, String, unsigned int, unsigned char*, unsigned int, bool) at http.cpp:?
This message was deleted
UpdaterClass::_writeBuffer() calls yield(). Isn't that forbidden in AsyncTCP ?
You can not use yield or delay or any function that uses them inside the callbacks
@me-no-dev looks pretty dim. :-(
I know some folks got it to run async
can you try commenting out the yield calls in
write and _writeBuffer
it uploads the file, then reboots, but does not start anymore
either the update is not finished correctly (but on my side) or the update does not work correctly anymore
if the updater said it's good, then it should by all means work
debug enabled on Updater?
trying that
skrew that, it worked, it just booted into bootloader
everything is fine
i tried 5 times now, and it worked every single time
the progress log (in my code) is borked, but this will be easily fixable.
reloading different pages is now faster and more stable, i'm quite happy
btw I do not know if you know, but there is an Async MQTT imlementation now based on AsyncTCP
i guess
yes
i will need TLS though, (even though i don't have it now)
is TLS planned?
click
axtls does not work with async
@me-no-dev : anyway, thank you very much. can you pls announce here, if the asyncUpdater gets merged?
i will pospone my push until it is at least in master
after more testing i have to say async works so much better, it really should be the default!
it is better, but it's not as easy as the regular one :) there are many things that you can not do inside async callbacks that users can do in the callbacks of the included server
esp8266/Arduino#2096 is the PR, it adds
Updater.runAsync(true/false); to switch the modes
@igrr thanks :)
Yeah, making async the default approach is the plan, but it requires some work... we need to make shooting oneself in a foot more complicated. After all, Arduino is about ease of use.
Right now i have a working port of asio, which is the candidate for networking API to be incorporated into C++ STL. It almost fits the bill, but unfortunately its dependence on STL brings some problems.
For example, it would be nice if it took Arduino strings, but of course it takes STL strings instead...
are you planning to ditch lwip all together?
so i have to call Update.runAsync() after Update.begin(). COOL!
thx guys!
works as advertised! :)
i can now push the hardware really hard, i.e. pressing reload on a page that immediately opens a websocket back from the browser as fast as i can. i can trigger an out of memory condition after ~20 reloads and even then it seldomly crashes. before that, it would reboot after 3 reloads at max.
About my NYTimes tests, I'm still pulling my hairs with WTDs, I think will wait the dusts coming down, especially that @vicnevicne got it working on his side.
@me-no-dev it's not a replacement for LwIP (which is a TCP/IP stack), it's an API wrapper around sockets, readv, writev, and select.
so no, not ditching LwIP
@igrr maybe I missread something :) it looked like full-on stack