search

letscontrolit / ESPEasy

Easy MultiSensor device based on ESP8266/ESP32
http://www.espeasy.com
Other
3.3k stars 2.22k forks source link

[Test Build] Upcoming X-mas build, testers wanted #4917

Closed TD-er closed 11 months ago

TD-er commented 11 months ago

We're approaching Christmas and I know I mentioned there will be a X-mas release, but I'm not 100% sure it will be ready. This all depends on the amount of critical bugs found by all of you. @fly74 @ghtester @chemmex @alex-kiev @iz8mbw @Wookbert @Dickesplumpel

Please test the latest build (same as this GH Actions build ) and let me know what critical issues you may find.

The ESP32 builds with "LittleFS" are all running the latest ESP-IDF 5.1 code (upto commits of a few days ago from Espressif) New platforms are ESP32-C2 and ESP32-C6. Those only have a 'custom' and 'normal' build right now as I plan on adding a 2-step OTA for those so these can all have a small partition for the OTA and a large partition for the build. This concept allows for all 4M boards to run a "Max" build and thus will reduce the number of build files significantly. This will later also be made available for all other ESP32-variants including some steps to migrate existing setups.

All ESP32-builds still running SPIFFS are based on ESP-IDF 4.4

I've been working on this migration to IDF5.x for over 2 months full time, as a lot had to be changed and upstream bugs had to be fixed for Arduino and ESP-IDF. So I would be pleasantly surprised if no critical bugs were found.

If you plan on migrating a build using SPIFFS to one using LittleFS, please keep in mind you will loose the settings. So please backup your settings and restore them when making this migration.

There is also some extra control of persistent settings like WiFi credentials etc. which will be stored outside the file system. So you could first update to a SPIFFS build, then check all the boxes on the "Factory Reset" page, save and wait for at least 5 minutes to let ESPEasy know the WiFi connection is considered stable. Then extra WiFI info will be stored in the NVS partition. If you then migrate to a LittleFS build, your WiFi will be kept (and all other things checked on that factory reset page)

Dickesplumpel commented 11 months ago

Esp8266-4M1M seems to work generaly perfect but I have trouble with dummy device or with taskvalueset, even if "taskvalueset to all plugins" is enabled taskvalueset to dummy device seems not working anymore.

fly74 commented 11 months ago

Esp8266-4M1M seems to work generaly perfect but I have trouble with dummy device or with taskvalueset, even if "taskvalueset to all plugins" is enabled taskvalueset to dummy device seems not working anymore.

Would be important to me.

TD-er commented 11 months ago

Hmm sounds like a bug which needs to be fixed before any release (and thus exactly the reason I asked for testing :) ) Happy to hear about this bug :)

TD-er commented 11 months ago

That was easy to fix, or a stupid bug... ;)

When the build is ready, I will upload it to the webflasher.

TD-er commented 11 months ago

Latest build with fix is now uploaded to the web flasher.

Dickesplumpel commented 11 months ago

Taskvalueset works! 👍 Maybe tomorrow I have a closer look at ESP32-S2 and ESP32 builds.

TD-er commented 11 months ago

There is still one issue I'm looking into right now and that's when calling wifidisconnect, the ESP sometimes reboots with a crash. At least when using LittleFS builds on the ESP32-C6.

fly74 commented 11 months ago

That was easy to fix, or a stupid bug... ;)

When the build is ready, I will upload it to the webflasher.

Looks good in my landscape.

fly74 commented 11 months ago

Houston there is a problem:

I have a Screenshot 2023-12-23 222254

wich is filled remotely with MMddHHmm to compare in a rule, if the time is longer than 6 minutes away.

let,1,[D1#Dummy]+6

if [var#1]>%sysmonth%%sysday_0%%syshour_0%%sysmin_0%
...
else

in this release it always comes to the else path, even if D1 is not longer than 6 minutes away.

2nd if I try to change a rule, save does nothing. (MS Edge)

tonhuisman commented 11 months ago

2nd if I try to change a rule, save does nothing. (MS Edge)

What build environment is that on? (as in: Works on my ESP32 machine 😛)

fly74 commented 11 months ago

2nd if I try to change a rule, save does nothing. (MS Edge)

What build environment is that on? (as in: Works on my ESP32 machine 😛)

In MS Edge?

tonhuisman commented 11 months ago

no, ESP8266, ESP32, etc. Normal, Collection, Display?

fly74 commented 11 months ago 
<html>
<body>
<!--StartFragment-->
Firmware
--
Build:⋄ | 20231223 - Mega32
System Libraries:⋄ | ESP32 SDK 4.4.6
Git Build:⋄ | HEAD_40df5ee
Plugin Count:⋄ | 48 [Normal]
Build Origin: | GitHub Actions
Build Time:⋄ | Dec 23 2023 19:15:05
Binary Filename:⋄ | ESP_Easy_mega_20231223_normal_ESP32_4M316k
Build Platform:⋄ | Linux-6.2.0-1018-azure-x86_64-with-glibc2.35
Git HEAD:⋄ | HEAD_40df5ee
System Status
Syslog Log Level: | None
Serial Log Level: | None
Web Log Level: | None
Enable Serial Port Console: | true
Console Serial Port: | HW Serial0 RX:3 TX:1 @ 115201 baud
Network Services
Network Connected: | ✔
NTP Initialized: | ✔
ESP Board
ESP Chip ID: | 16169629 (0xF6BA9D)
ESP Chip Frequency: | 240 MHz
ESP Crystal Frequency: | 40 MHz
ESP APB Frequency: | 80 MHz
ESP Chip Model: | ESP32-D0WDQ6
ESP Chip Features: | Wi-Fi bgn / BLE
ESP Chip Revision: | 1.0
ESP Chip Cores: | 2
ESP Board Name: | Espressif Generic ESP32 4M Flash ESPEasy 1810k Code/OTA 316k FS

<!--EndFragment-->
</body>
</html>
tonhuisman commented 11 months ago

And what build did that still work on?, as it seems to be related to parsing the if condition, and that hasn't recently changed, AFAICS. if I change your rule to:

Let,1,[D1#Dummy]+6
Let,2,%sysmonth%%sysday_0%
  If [var#1]>%v2%%syshour_0%%sysmin_0%

it works as intended, though I can see what you expect

fly74 commented 11 months ago

If I could save the rule I would test it.

tonhuisman commented 11 months ago

Can you hard-refresh the browser page, so it can re-fetch the js code?

tonhuisman commented 11 months ago

or watch the log in a second browser tab (or serial log) to see if it throws an error when saving the rules file?

TD-er commented 11 months ago

What kind of "sensor type" have you used for the dummy? This looks like the float resolution, which is essentially 6 decimals and you're using 8 decimals here. This is exactly why I added other integer types to the dummy.

fly74 commented 11 months ago

Can you hard-refresh the browser page, so it can re-fetch the js code?

Firefox don't work too. But download and edit and upload works. @tonhuisman The rulechange works. Thx.

So only the save would be interesting.

Browserdebug: Screenshot 2023-12-23 232120

fly74 commented 11 months ago

What kind of "sensor type" have you used for the dummy? This looks like the float resolution, which is essentially 6 decimals and you're using 8 decimals here. This is exactly why I added other integer types to the dummy.

Screenshot 2023-12-23 232326

TD-er commented 11 months ago

What I meant is this selection on the Dumy task: image

You're referring to the value of the dummy task value, which is by default a float. (Single/dual/quad etc are all float)

fly74 commented 11 months ago

What I meant is this selection on the Dumy task: image

You're referring to the value of the dummy task value, which is by default a float. (Single/dual/quad etc are all float)

But Ton's rulechange works with single :/

tonhuisman commented 11 months ago

Seems to be a parsing issue, as I got an error with the original rule that didn't have the %sysmonth% part replaced

Edit: 2011416: Calculate: Unknown token input: %sysmonth%232333 = 0

TD-er commented 11 months ago

But Ton's rulechange works with single :/

Yep, it looks like he is splitting the value into separate floats. But depending how the value is being set via remote calls, this will still fail on new updates.

If you set the new value via an event value, then you can split it with enough resolution just fine. But if it is stored in the dummy as a single value, then a new update may not appear to change the task value.

Edit: This has always been a problem, that's exactly why I added these other options to store int values and even int64 types if really needed.

Edit2: Any integer value outside the range of -8388608 ... 8388607 will be stored with reduced resolution as a float simply can't represent those values without loss of resolution.

fly74 commented 11 months ago

But Ton's rulechange works with single :/

Yep, it looks like he is splitting the value into separate floats. But depending how the value is being set via remote calls, this will still fail on new updates.

If you set the new value via an event value, then you can split it with enough resolution just fine. But if it is stored in the dummy as a single value, then a new update may not appear to change the task value.

Edit: This has always been a problem, that's exactly why I added these other options to store int values and even int64 types if really needed.

Edit2: Any integer value outside the range of -8388608 ... 8388607 will be stored with reduced resolution as a float simply can't represent those values without loss of resolution.

@TD-er I can't recognize the problem. The 8 digit are set remotely to Screenshot 2023-12-23 233736 to single.

TD-er commented 11 months ago

Hmm tested here and indeed, you're still in the proper range. Just looked here and it seems I was off by a factor of 2 for the range. Strange as I thought it had a range of 23 bits, but apparently it allows upto 2^24 -1.

TD-er commented 11 months ago

Just had a chat with Ton, apparently 3 variables replaced in a single line seems to work, but 4 doesn't work. Have to dive into this a bit further... So let me get a fresh beer and have a look :)

fly74 commented 11 months ago

Hmm tested here and indeed, you're still in the proper range. Just looked here and it seems I was off by a factor of 2 for the range. Strange as I thought it had a range of 23 bits, but apparently it allows upto 2^24 -1.

Precision limitations on integer values
Integers between 0 and 16777216 can be exactly represented (also applies for negative integers between −16777216 and 0)

That makes it clear. Int is enough to store 8 digit when the left 2 is the month 0..12.

fly74 commented 11 months ago

Just had a chat with Ton, apparently 3 variables replaced in a single line seems to work, but 4 doesn't work. Have to dive into this a bit further... So let me get a fresh beer and have a look :)

But worked in the last official.

TD-er commented 11 months ago

But worked in the last official.

Never underestimate my super powers to introduce sneaky bugs. But you guys managed to find them within hours and it probably took me a lot more to create these bugs. ;)

chromoxdor commented 11 months ago

Ohhh.. here is a testing party going on... i almost missed it. 🙂

fly74 commented 11 months ago

Ohhh.. here is a testing party going on... i almost missed it. 🙂

@chromoxdor It's not to late come in.

TD-er commented 11 months ago

I expect there will be some bugs left for you to find and report :) @fly74 I think I found it... have to test it What was the exact line you used with the variables?

%sysmonth%%sysday_0%%syshour_0%%sysmin_0%

??

chromoxdor commented 11 months ago

@chromoxdor It's not to late come in.

Thanks! But first things first. Where do i find the buffet?

I expect there will be some bugs left for you to find and report :)

I hope i find these before the release...

fly74 commented 11 months ago

I expect there will be some bugs left for you to find and report :)

@fly74

I think I found it... have to test it

What was the exact line you used with the variables?


%sysmonth%%sysday_0%%syshour_0%%sysmin_0%

??

As above 👆. Before Ton fixed it.

TD-er commented 11 months ago

If I place the variables in a different order, they will be replaced without any error:

logentry,"%sysday_0%%syshour_0%%sysmin_0%%sysmonth%"

03:40:23.464 : (305112) Info   : HTTP: logentry,"%sysday_0%%syshour_0%%sysmin_0%%sysmonth%"
03:40:23.471 : (305128) Debug  : HTTP after parseTemplate: logentry,"23232412"
03:40:23.475 : (305096) Debug  : Command: logentry
03:40:23.476 : (305096) Debug  : logentry,"23232412"

Here the order you used:

03:40:01.958 : (312032) Info   : HTTP: logentry,"%sysmonth%%sysday_0%%syshour_0%%sysmin_0%"
03:40:01.963 : (312032) Debug  : HTTP after parseTemplate: logentry,"%sysmonth%232324"
03:40:01.968 : (312016) Debug  : Command: logentry
03:40:01.969 : (312016) Debug  : logentry,"%sysmonth%232324"
03:40:01.969 : (312000) Debug  : Par1: 12232324 Par2: 0 Par3: 0 Par4: 0 Par5: 0
03:40:01.970 : (312016) Info   : 12232324

So it seems like the order makes a difference here, which it obviously shouldn't.

N.B. the reason why this could now be broken is that I have also added quite some improvements in parsing speed and reduction of the bin size. So a lot (!!) of code has changed.

chromoxdor commented 11 months ago

Almost all notes have no proper formatting

Bildschirmfoto 2023-12-24 um 00 26 45
chromoxdor commented 11 months ago

Ok...i need to sleep. So far it looks good here except the notes. @TD-er how much time do i have for testing? I guess you want to release it tomorrow?

TD-er commented 11 months ago

Well if it isn't done, it isn't going to be released. And here in the Netherlands we have 2 Christmas days, so even the 26th is still Christmas :)

TD-er commented 11 months ago

OK these are all fixed now. Even the notes: image

See GH Actions build: https://github.com/letscontrolit/ESPEasy/actions/runs/7311657637

I'm not going to wait till 3:30 am to upload these to the web flasher. So please use the GH Actions build for testing :)

TD-er commented 11 months ago

OK, found 1 possible other issue which I will look into tomorrow... (well after sleeping...) The config command might not work right now or if it does we can speed up command parsing even more and reduce build size along the way. :)

fly74 commented 11 months ago

Anyway merry Christmas!

chemmex commented 11 months ago

Not sure if current code is affected, but it seems that latest ESP32x custom builds do not apply #define DEFAULT_NAME from Custom.h Instead, the name is set to ESPEasy- which is broadcasted as AP SSID. ESP8266 builds are Ok

Dickesplumpel commented 11 months ago

Nothing to test - build failed!

chemmex commented 11 months ago

Nothing to test - build failed!

There are at least ESP32 images in the binaries.zip

TD-er commented 11 months ago

Just added a quick fix, hopefully this is enough to complete the build

Dickesplumpel commented 11 months ago

For me it makes no sense to test when build failed, I wait for next build.

TD-er commented 11 months ago

I have no idea why it doesn't trigger a new build, I will try pushing the commit from my main computer...

iz8mbw commented 11 months ago

Hi guys. Sorry but for now I cannot give my contribution for testing. I'll try to do my best in next days (if I'm still in time).

TD-er commented 11 months ago

No idea why the GH Actions build didn't start, but now it is running... https://github.com/letscontrolit/ESPEasy/actions/runs/7314499893