so i have now isolated this minimum websocket sketch. when i send JSON data of 2930 bytes, it works. when i send 2931 bytes the browser receives garbage. :-D
looks a little bit like if (date length ~ < 2 * MTU) { okish }
that is so because that is the maximum size usually for the first outgoing window
in order for it to send larger packets, you need to split them into MTU chunks and send when it's possible
there is a bit more to that if you want to get the transmission as a single message on the other end
i have hoped, that arduinoWebSockets would do this for me :-(
well large data on the ESP has always been a problem, because you do not have infinite ram
everything is doable though ;)
unfortunatelly Links have not exposed the required method for you to contruct the data
I'll look in a minute what I have exposed in my lib and show you how you can do it with that
i have opend an issue on github: Links2004/arduinoWebSockets#85
yes, i know about that upper limits, but as long as ram is available... didn't expect to things bailing out at 3KB
arduinoWebSockets apparently supports your async TCP lib, but last time i tried that it didn't help
i was not talking about the async tcp lib but the async web server
which has web sockets also
limitation is ok, you just need to think different about it and it can be done
async web server?
yes
this one?
yes
checking it out
there isn't currently an example of how to send large data messages but I can help
thats nice. thank you!
i currently look at the API,
i see AsyncResponseStream *response = ... i guess response handles its internal buffer with something like String. correct?
this is not for the web socket though
the AsyncResponseStream just exposes a Print interface so you can print, println and printf to the output
yes, but if i will swtch to async tcp i will have to migrate my current framework over to it completely.
yes
this is my current http implementation, will it be much work to port over to async webserver? what do you think?
it will be actually pretty easy
then i could get rid of the extra websockets lib as well
pick a callback and I'll translate it to async for you so you can see how it will go
:-D
API is mostly the same, but you do need to carry the request you are answering to instead of calling the server methods
static void handle_info_cb(AsyncWebServerRequest *request) {
#ifndef RELEASE
String webpage;
if (!setup_complete()) return;
if (!authenticated()) return;
if (!webpage.reserve(4*1024)) {
log_print(F("HTTP: failed to allocate memory\n"));
}
html_insert_page_header(webpage);
html_insert_page_body(webpage);
html_insert_page_menu(webpage);
html_insert_info_content(webpage);
html_insert_page_footer(webpage);
request->send(200, "text/html", webpage);
system_count_net_traffic(webpage.length());
#endif
}
:D that was too easy
htat is because of my great software design!!! ;-)
static bool authenticated(AsyncWebServerRequest *request) {
String header;
if (request->hasHeader("Cookie")){
String cookie = request->header("Cookie");
String name = "GENESYS_SESSION_KEY=";
if (cookie.indexOf(name + session_key) != -1) {
return (true);
}
}
AsyncWebServerResponse *response = request->beginResponse(301);
response->addHeader("Location","/login");
response->addHeader("Cache-Control"," no-cache");
request->send(response);
system_count_net_traffic(header.length());
return (false);
}
ahh for auth
here is a more interesting example (need to change the call to authernticated in handle_info_cb
IC
looks straight forward
OTA could be different i guess though?
the biggest doifference is that you are always working with the current request in the callback as opposed to with the server
the regular web server is blocking, so you are always responding only to one client
in async you can be responding to 5 clients at the same time
indeed, only limited by free ram i guess
ram and lwip setting
AsyncWebServerResponse does it have something like length() ?
so i know how much bytes it sent?
many
depends on the response
but why? don't you know how much you sent?
not really. look at the last function call in the 2 mothods you ported for me
in the first it was easy
but the second is complicated
its no deal breaker, but ... i'd prefer to have an API for it in the network stack anyway. this system_count_net_traffic() is a workaround, because there is no byte count in the LWIP for an PHY
oh, you want to count all bytes?
then you need to go inside lwip for that
else you are missing most of the actual bytes that go through the phy
also the response is something that is managed in the background once you "send" it
it's queued for data when possible and freed on completion
yes, i know, it is a crude way, but LWIP does only count packets and not bytes AFAIK
there is a stats API, but i couldn't figure out how i could get to the bytes sent
maybe i could use that linker trick to wrap send and receive in my own send and receive functions, once i figure out, whick one i have to wrap
that would be the best way
i did this with malloc/free... to keep track of heap
for malloc/free you have the source in the core compiled so no need for wrapping ;)
core/umm_malloc/*
this would suck, because i would have to change each time i pull the newest versions
and everybody else who wants to use genesys as a boilerplate firmware
the wrapper works without any changes to the core
only to the build process
so not usable through the IDE
that's true, but i can't stand the arduino IDE in the first place
:-)
🤔