A lot of bugs on a EOS 70D, some major too

[lizard™]

Looking at the bug tracker having currently over 400 open issues, I figured I might create a topic first. Maybe some of these are already well-known, or easy to fix for myself. If I should open a bug report for any of those or all of those, just let me know.

So I got my new EOS 70D and installed ML on it (magiclantern-lua_fix.2018Sep17.70D112.zip) and at first glance, everything was fine. Until I noticed a bunch of bugs, some of them really annoying:

Please note that within "everything was fine" and "everything was ok" and "all hope is lost" I didn't do anything, not even taking pictures. Still just playing around with the menus, but didn't change any settings (except for focus peaking and zebras, those being on or off don't affect these bugs though).


1) Some menu items (of ML) disappear sometimes. E.g. the lua scripts and the modules tab just disappeared earlier on me. Had to restart the camera around 6 times until it re-appeared.

2) The LCD is sometimes kind-of f*&ked. Currently I only have this in the arkanoid game (which worked fine until yesterday!) but it happened in the menu before too.



Note that this is only on the LCD itself. Taking a screenshot of this (using the ML VRAM screenshot feature) shows the expected normal image.

3) When in the ML liveview and half-pressing the shutter button, I get these super annoying overlays from Canon:



This is how it should look:



4) Changing to any other display theme results in heavy, constant flickering: (GIF didn't catch it, but it's NOT only in the ML menu. It's everywhere.)



5) Not able to change anything in the ML menu at all. After pressing set, I can still cycle through the menu points instead of changing the option:



6) ML overlays do not appear (at all) while Wi-Fi is enabled, even when Wi-Fi isn't in use.

7) USB connection doesn't work at all anymore. "dmesg -w" on the computer shows nothing. Camera on or off, SD card in or out - Same results. The same machine and cable worked fine two days ago in both darktable and the file manager. Will try on another machine soon, but worked on current 2 days ago. No changes (not even a reboot) have been made to PC in-between.
Was due to Wi-Fi being on, not a ML bug. Is there a way to have Wi-Fi enabled and still keep USB port functionality using ML?

8) Shutter speed changes are sometimes ignored by the ML overlay. In normal overlay (canon liveview) it works fine. In ML sometimes I get this weird behaviour:
I decrease the shutter speed by one stop: It works. Camera's top LCD shows the change, so does ML.
I increase it by one stop: Camera top LCD shows the change, ML doesnt. I increase it another stop, ML shows it.
So basically ML's overlay is skipping one stop, but only when I increase the shutter speed. Decreasing works fine all the time.


Happened since the beginning, always: 3), 4), 6)
Happened since yesterday, but constant: 1), 2), 7)
Happens sometimes, since beginning: 5), 8)

I'm sure there's been more that I forgot by now. I'll update as I find more. On a side note: Is there something I should backup just in case ML ends up bricking the camera itself?
I love ML so far but these bugs are really annoying, especially 5), 4) and 2) being quite major issues. If any more information is needed to debug, please ask.

[a1ex]

1) by default, modules are not loaded after a bad shutdown, and you get this explanation on the screen (unless you start the camera with the main display off...)

2) if it doesn't happen with this build, I think I know what the problem is; assumed only 5D3.123 was affected; 5D3 1.1.3 and 700D were not affected, 6D was reported to be clean as well, so I've assumed all other models are OK.

After going to LiveView and back, or after reviewing an image and going back, it should be OK. Please confirm, so I can re-enable the workaround.

3) hard to fix, all cameras do the same.

4) was not aware of this; none of my cameras do this and unable to reproduce in the emulator; may require rewriting that feature from scratch.

5) left/right/scrollwheel

6) not aware of this, no way to test/debug, sorry; enabling wifi in the emulator doesn't break ML menu...

7) ?!

8.) video of the issue?

9) the installer tells you what files to back up.

[lizard™]

1) But one of the "modules" that disappear, is the module menu itself. As in, I can't load or select any modules at all. Is that expected to happen? And what would one do to fix that? I tried restarting it 6 times until it came back..

2) Doesn't really let me install it. I've unpacked it and placed it on the SD card, camera won't turn on at all anymore. Tried another SD card with the newest experimental build again and after turning it off/on a couple times it worked. Switched cards again, doesn't work. Switched cards back, doesn't turn on at all anymore. Without an SD card it doesn't turn on either at first, but after trying around 10 times with battery out/in a couple times it works. I doubt this is normal. Not sure exactly on how to load the other ML firmware, or how to turn on the camera reliably at all at this point.

4) I'll try this again with the other ML image once it works.

5) Is this expected? Because this doesn't always happen and seems like a bug to me. Why does it just happen sometimes? It gets really annoying with menus like "Picture Style" for example.

7) Sorry, it was due to Wi-Fi being enabled. Is there any reason behind canon disabling the port while Wi-Fi is enabled, even when not in use?

8) I can record one, but it's really just that it's always skipping one stop when increasing the shutter speed, nothing else happens and there's no weird screen behaviour or anything.


Another issue I noticed is that when using a battery grip, ML reports "100%" and full battery icon all the time, even when the Canon info screen would show the empty battery warning already. Without using the battery grip it works fine in ML too.

[lizard™]

Ok, about 2): Managed to turn it on using the other SD card with the newest build. Went to firmware installation and waited 60 seconds to uninstall ML again, put ML on SD card again using another card read (first one seemed to mess it up) and it worked. Still weird how the camera behaved. This wasn't the first time I had to try this many times to just turn it on, even with no SD card in it.

Using the build you linked now.

1) The LCD bug is gone. It works fine now.
2) The color scheme bug is gone now. No flickering anymore.
3) The menu bug seems gone now, I can select just normally all the values now and change it with the up/down keys.
4) The shutter speed bug seems to be gone too, however this one appeared randomly before. But so far, it works.

There seems to be a change between those two builds that causes most of these bugs. I'll test further.


Edit: Nevermind on 2) - It worked fine until I restarted the camera. Nothing else was done or changed. It worked perfectly, until reboot. The other three bugs seem still fixed.

Edit 2: 3) isn't fixed either, but I've found out the "random" behaviour: If liveview is disabled, it works fine. But if liveview is enabled, it acts weirdly as in my GIF.

Edit 3: The same seems to apply to 2). If Liveview is turned off, it works perfectly fine. Once I turn on liveview, it acts as weird as in the GIF.

[a1ex]

1) I restart once; no surprises yet...

2) after a failed startup attempt, take the battery out; otherwise the camera will remain stuck even if you flip the power switch, change the card etc.

5) looks OK to me, try the keys suggested earlier

7) no wifi-enabled camera here, no idea, sorry...

8.) pretty sure it's not "really just that"; that's why I've asked for a video - to see how to reproduce and what other features are enabled.

10) also reported on 5D3; no battery grip here to test, but you should be able to get a log file with MPU messages; that log should contain the complete info to reproduce this situation in QEMU, as the battery level is reported by the MPU.

[lizard™]

10) also reported on 5D3; no battery grip here to test, but you should be able to get a log file with MPU messages; that log should contain the complete info to reproduce this situation in QEMU, as the battery level is reported by the MPU.

Which part of the MPU log do you need?

Also I tested around further and I'm able to reproduce the menu bugs (my GIFs) every single time when I enable liveview. I tried to disable overlays completely in ML but the issue still persists as long as liveview is running while being in the ML menu. If liveview is off, it works perfectly fine, no flickering.

The shutter speed issue didn't occur again, I tried several settings to reproduce it but since I changed to the build you've linked earlier I wasn't able to reproduce the issue.

Another issue I noticed: If I use the dark red theme (probably happens with others too but didn't test) and zebras (1% on over- and underexposure, Luma Fast) the zebras will flash constantly all over the Liveview, indicating that the entire image is underexposed. If there IS actual under- or overexposure in the image, the zebras don't work (still flashing all over the LCD, not on top of overexposed image parts).

[a1ex]

I need the entire MPU log, covering a regular startup process with that battery grip. If possible (i.e. if it appears in Canon menu), write down the expected battery percentage, too, so I'll know what to look for.

Confirmed the fast zebras bug.

[lizard™]

I need the entire MPU log, covering a regular startup process with that battery grip. If possible (i.e. if it appears in Canon menu), write down the expected battery percentage, too, so I'll know what to look for.

Confirmed the fast zebras bug.

Here is the entire log: https://pastebin.com/3QTG6kv4

Something weird: The battery % in the canon menu (correctly identifying the battery in the grip) said 80%. I then shut off the camera and loaded the startup-log build onto SD. I then checked the canon menu again (without SD card) and the battery was magically at 99%. I don't know how this happened. After putting the SD card back in I waited until the log was generated, then checked battery % in Canon's menu again: 99% still. After playing a bit of pong it's now at 97%. Very confusing why it reported 80% before. (It should be indeed above 95% as it was fully charged)

This is at the bottom of the log:

Code Select Expand

20.134.998     CtrlSrv:ff19b3ac:83:03: PROP_BATTERY_REPORT copysize (96)<-(96)
20.135.186     CtrlSrv:ff19b55c:83:03: PROP_BATTERY_HISTORY copysize (76)<-(76)

Which confuses me further, but maybe I'm just reading it wrong. From a simple guess I'd say it reported 96%. This was when canon's menu said 99%.

On a side question: I just looked at the ML experiments download page. For the crop_rec_4k module there is no 70D download, however when scrolling down to the manual_lens_info downloads, there is one for the 70D that says "crop_rec_4k" too. Does that mean it includes both? Does that include the lua_fix too? Having this many branches is quite confusing, but I guess making these experimental changes as simple modules won't work because of compability issues with the nightly builds. (?)

[a1ex]

Actually the 70D code compiles in the crop_rec_4k branch, but neither crop_rec nor lossless compression are working, to my knowledge. That 70D build appears to be left over from an older experiment that's waiting for my attention for quite some time...

Will look into the battery log later (there's a bricked 70D waiting to be recovered first - its battery was removed in the middle of a firmware update).