publishing an intermediate release 3.1.0?

blackmoon wrote: Don't know if I can post this here, if not I'll move.

For future it will be better to open a new thread. I can assure I read anything...

Was playing whit devo10 and latest commit compiled this evening with VM, when the tx rebooted.

I was using my old model.ini files for Dev 3.0, maybe it's why...

I changed to model menu to load a new model, nothing on screen, hit ext or ent (don't remember well) key, reboot...

With the help of the map file and the errors.txt I have seen that the trouble happened while refreshing the screen when drawing a label (probably the headline "Press ENT to load"). The chain is: main - GUI_DrawScreen - GUI_DrawObjects - GUI_DrawObject - GUI_DrawLabel - open_font - devo_fread - _read_r - GUI_DrawLabelHelper - LCD_GetStringDimensions.
I have looked into the code of screen/lcd_string.c but I can't imagine what could be wrong with the code. I guess this method was not changed with your patch so I analyzed the assembler code from my objs/devo10/lcd_string.o, but nothing found what might go wrong (as far as I can see - I have not looked for the memory addresses where variables were stored).
I have no idea. Everything looks fine and the possibility that something strange happens is quite low.

Is this behavior reproducible? Can you try (leave the empty page for a few seconds; maybe reboot occurs now without button pressed)? Send me the errors.txt if it's different.
There are only few objects to be drawn and the only one where the size is calculated is "Press ENT to load". Which language did you use? And what is the translation for this string in your language file (please look at your tx since you did not refresh the language files)?

richardclli
Did you replace the filesystem completely with the new one? I found that replacing the whole filesystem is essential when upgrading from 3.0 to 3.1, however you may reuse the old model files afterwards.

I replace the all filesystem expect tx.ini (was lazy), then replaced the model files with my own.

After your post I remembered that on the early days I had some issues when I didn't replaced the tx.ini file in between releases.

But I still fail to see how that text file could trigger a reboot, after all it is only a text file...

I have now replaced it by the one provided in the build and calibrated my tx again.

rbe2012 wrote: Is this behavior reproducible? Can you try (leave the empty page for a few seconds; maybe reboot occurs now without button pressed)? Send me the errors.txt if it's different.
There are only few objects to be drawn and the only one where the size is calculated is "Press ENT to load". Which language did you use? And what is the translation for this string in your language file (please look at your tx since you did not refresh the language files)?

I will try to reproduce, and the language used was English.

Even before I pressed ENT there was nothing displayed on screen, not really nothing, it had a video invert box 5mm - 1cm wide(full black)...

I did had some glitches prior the reboot, the press on navigation buttons would make the choice jump two rows (i.e. model1 to model3 not model2)

Maybe something about the debounce method , or just my devo hardware that's faulty...

It could be a zillion things

blackmoon wrote: I replace the all filesystem expect tx.ini (was lazy), then replaced the model files with my own.

rbe2012 wrote: Is this behavior reproducible? Can you try (leave the empty page for a few seconds; maybe reboot occurs now without button pressed)? Send me the errors.txt if it's different.
There are only few objects to be drawn and the only one where the size is calculated is "Press ENT to load". Which language did you use? And what is the translation for this string in your language file (please look at your tx since you did not refresh the language files)?

Unfortunately me again. Last weekend I was able to confirm blackmoons issue. beta11 with new filesystem and tx.ini calibrated on 3.1. German language - tried to load a new model but the screen was totally empty. I do not know if I pressed a button but the Tx returned with the beep to main screen. Shut down Tx going directly to the load screen . All ok. Switched to English same reset as before but I also was not able to reproduce this behavior reliable. So, you know, then it is not an issue .

EDIT: error.txt added

Didn't had to test to much.

Full life battery, power on...

model already selected : super CP

model setup --> load : FBL100 ppm

servo monitor played 2 min : all ok

model setup ---> load : hubsan X4

model setup ---> load : nothing...

screen shows :
[/tr]
[/tr]

model setup	Below screen is blank with a scroll bar at right side
played with all buttons, ENT, UP DN R L, switches pots and sticks; expect EXT key	nothing happens on screen, let it settle for 5 min no reboot
Press EXT key once	tx stays on Model menu with black box in middle of screen
Press EXT again	Top screen change to Main menu with a black box wider in middle of it
Press EXT once again	Only one character shown a È (Alt+ 0200) to the top left of the screen	Then TX reboot one second latter	after reboot tx shows bind screen and when ENT pressed Main screen is shown, with last model Hubsan X4

There is definitely something going on... should I open a thread for this ?

or file a ticket on bitbucket ?

Could it be related with the use of model.ini files from deviation 3.0 trunk ?

I was using deviation-devo10-v3.0.0-ddf6350 nightly and all my models where done with that one.

Just tried two times my devo 7e with deviation-devo7e-v3.0.0-ddf6350 same steps no reboot...

Blackmoon, could you please post all your model.inis and, if exisiting the model icons. Then some other guys would be able to confirm or can have a look at the ini files.

rbe already made an ticket for this issue https://bitbucket.org/rbe2012/deviation-3.1/issue/63/devo10-reboot-by-blackmoon but you cannot attach files if you have not opened this ticket.

@rbe: Sounds different to my experience but happened on the same place.

kreidler wrote: @rbe: Sounds different to my experience but happened on the same place.

Your errors.txt (without further investigation) looks as if you entered an endless loop somewhere.
I must look for the elf file to see when it happened. You haven't built it by your own and have the map file for me?

EDIT: If your build was downloaded from my repo (this is what I guess) I have the map file.
What I have seen so far: the last things done were to remove all (old) elements from the scroll list and draw actual entries.
I don't have yet found a reason why the addresses appear more than once in the backtrace - this should only happen in recursions.
I will investigate the GUI_Scrollable later...

Tactic SLT - please build and test from my fork at bitbucket.org/rivig/deviation-3.1-slt . Or I can build Devo 10 and 7E builds later.

Hisky - I would like to fortify this a bit - add frequency hopping generator so that there could be many TXs at a time active.

It requires some testing and I am not able to do so this week end.

vlad_vy wrote: RBE, why not include latest developed protocols (HiSky and Tactic Anylink) to v310 release? It change nothing except protocols list.

I can take a look. As far as I remember are the protocol implementations quite independent from each other. If others are not affected we could do that, but as victzh wrote they should be tested (I don't want to create doubts about the stability of deviation all over only because a single protocol does not work properly - and I don't want do publish 3.1.x short time after 3.1...).

Does anybody know if the modular approach for Devo7e could be transferred to the other tx? Might be a problem of resources which are esp. on Devo12 somehow limited (or why should we put the media folder in the lib?)...
If so, we could possibly exchange simply the compiled protocol implementations instead of the complete firmware.

I will be able to test SLT (and hopefully, fixed Hisky) next week only.

I will try to build my SLT later today for people to test. Anybody interested?

Modular approach can be applied to anything, but I am not sure how it works - are the relocation addresses stable from build to build or are they calculated from the build as a whole? If latter, we're probably screwed a little - no separate updates of the modules, only build as a whole. I need to look at how it works finally.

...but when they are modular, you can place only those you need in the file system. And if we find a method to build them independent from the firmware we have not to worry too much about integrating improvements - we would not need to wait for the next deviation release...

rbe2012 wrote: EDIT: If your build was downloaded from my repo (this is what I guess) I have the map file.
What I have seen so far: the last things done were to remove all (old) elements from the scroll list and draw actual entries.
I don't have yet found a reason why the addresses appear more than once in the backtrace - this should only happen in recursions.
I will investigate the GUI_Scrollable later...

Sorry for the reply. Too busy at the moment.
Yepp, origin beta 11 from your repo.

But also still waiting for blackmoons response.

Sorry but like you to busy with others things.

The three model I was playing with and my models icons.

But I don't see how these could induce a reboot.

blackmoon wrote:

File Attachment:

File Name: model_ini.zip
File Size:2 KB

File Attachment:

File Name: modelico.zip
File Size:3 KB

Copied all three models and all icons (one hubsan not working ) followed your instruction posted earlier but I had no success. If you can repeat the crash any time when you follow your post I would need at least the models 11 and 12 if their name is not NONE.

a small thing i noticed today was when using the simple mixer and heli going into the dual rates and expo settings the small dot on the curve no longer follows the stick input. it still works as expected on the pitch and throttle curves but not the dual rates and expo screen. the fw was built today from rbe2012's repo with the hisky patch added

if there is anything you need from me to help look at this just let me know

HappyHarry wrote: if there is anything you need from me to help look at this just let me know

Only a small detail... which tx did you use? Devo10 I guess?
Can you publish a model config where I can see what does not work?

sorry it's a devo7e, and it happens even with a fresh reset config so i think it's not model config related, the actual dual rates and expo still works, it's just that when setting the rates there is a small cluster of pixels that move up and down the curve following the movement of the stick. this doesn't happen anymore on the dual rates screen, but still works ok on the throttle and pitch curve screens

kreidler wrote: Copied all three models and all icons (one hubsan not working ) followed your instruction posted earlier but I had no success. If you can repeat the crash any time when you follow your post I would need at least the models 11 and 12 if their name is not NONE.

My all model folder, after 15 it's the default model file.

Pay attention, some of the files are not usable, some were copied from the forum but don't work, like a cyclic template for a fbl unit...

Has I use exclusively my devo7E for some time now, I didn't bother to update the files on the devo10.

And your right hubsanX4_2.bmp won't work (to many letters ?)

blackmoon wrote: ...hubsanX4_2.bmp won't work (to many letters ?)

Only 8 characters are allowed. I am actually not sure if they can be uppercase.