Messages

4.9 also miscompiled 5D3 and M builds last time I checked, same symptoms, ie. the led stays on and nothing else happens and neither of them has Digic 4.

Don't know if it's feasible as it's probably a bit more work to integrate, but LuaJIT 2 is quite a bit faster even with JIT off than the standard Lua interpreter so it might also be a good alternative for these fairly performance constrained devices Also it has integrated FFI interface (http://luajit.org/ext_ffi.html), which makes working with C even easier.

EOS M

Phew, glad it worked out

Any updates regarding FRSP for EOS-M or is that ever not going to be a possiblity?

FRSP should work fine for M in the branch since August last year, at least it did back then

There is a feature in ML called movie restart that tries to restart recording after it stops at 30 minutes. It causes about a second worth of dropped frames though. Can be found under movie -> movie tweaks when the camera is in movie mode.

It uses dcraw and exiftool utilities from the system, so those need to be installed as well.

Exiftool creates a temporary file, then renames it; not sure how to avoid this behavior. If this is too slow, exiv2 is much faster (but the syntax is different).

Also with -overwrite_original_in_place option?

Out of curiosity tried this on 5D3.123 (ProcessPathForFurikake is at 0xFF3CCC14) and it seemed to work fine and visibly faster than without the module. Also looked up the function from EOSM.202 (0xFF425150) but had problem with task creation.

whenever I try to burn me video ML raw red dots appear on the images someone been the same

http://es.zimagez.com/zimage/capturadepantalla2014-08-12alas125742.php

These are the "Pink Dots" aka AF pixels that are on the sensors, similar to what 650D and 700D have. There are ways to remove them, search for PinkDotRemover or use chroma smoothing options to mlv_dump (-cs2x2 or 3x3) or similar option in the converter you use.

Dual ISO images usually require some exposure adjustment due to the way cr2hdr reconstructs and scales the 16-bit data from two 14-bit images. The main benefit is that the shadow areas are cleaner and hold better when exposure is increased. So some exposure/highlight/shadow adjustment is almost always needed.

Just received a nice EOS-M kit from my wife 

EOS-M crash here :

https://drive.google.com/file/d/0B5V0Ovu4aQeQOERTZ0VUUmV1WDA/edit?usp=sharing

Must remove battery to stop it.
No log written on the SD.

(And the ML menu disappear after 5-7 sec. but i think you know it)

The default memory amounts are a bit too much for M without changing parameters to free up some memory. Try setting S3 JPEG and multishot-NR mode and you should have enough, at least mine does. Check from the memory available screen in debug menu how much you have, you need more than 64 megs free, with the settings above I get more than 90, but only 23 with my defaults.

Without either of the memory test loops I can now (after the recursive lock fix) run all tests successfully. The first memory test gives 400 failures:

[FAIL] p = (void*)_shoot_malloc(24*1024*1024) => 0x0
[FAIL] UNCACHEABLE(p) => 0x40000000

Reducing that allocation size to 20 megs or switching the camera to small JPEG makes it work. Debug menu gives maximum contiguous block as 20MB with my normal RAW settings, 29 megs with S3 JPEG.

The second memory test loop partially hangs camera, doesn't give the stack overflow now, camera can be shut down, but leaves the led on, I'll double check the stubs just in case and add some more prints.

The stubs all seem to be correct and if I put the camera in multishot-NR mode with S3 JPEG, I get enough memory to pass both memory tests (and all other tests as well, 4307 passed, 0 failed.).  So no other problems than the RecursiveLock stubs it seems.

Stack overflow errors are not funny. Can you try to locate the test that causes it?

Without either of the memory test loops I can now (after the recursive lock fix) run all tests successfully. The first memory test gives 400 failures:

[FAIL] p = (void*)_shoot_malloc(24*1024*1024) => 0x0
[FAIL] UNCACHEABLE(p) => 0x40000000

Reducing that allocation size to 20 megs or switching the camera to small JPEG makes it work. Debug menu gives maximum contiguous block as 20MB with my normal RAW settings, 29 megs with S3 JPEG.

The second memory test loop partially hangs camera, doesn't give the stack overflow now, camera can be shut down, but leaves the led on, I'll double check the stubs just in case and add some more prints.

On EOS-M the full test causes stack overflow in idle task during the malloc tests, but anything up to that works:

Code Select Expand


[Pass] buf = fio_malloc(0x1000000) => 0x4de1a094
[Pass] FIO_GetFileSize_direct("test.dat") => 0x82000000
[Pass] f = FIO_OpenFile("test.dat", O_RDWR | O_SYNC) => 0x3
[Pass] FIO_SeekSkipFile(f, 0, SEEK_END) => 0x82000000
[Pass] FIO_WriteFile(f, buf, 0x10) => 0x10
[Pass] FIO_SeekSkipFile(f, -0x20, SEEK_END) => 0x81fffff0
[Pass] FIO_WriteFile(f, buf, 0x30) => 0x30
[Pass] FIO_SeekSkipFile(f, 0x20, SEEK_SET) => 0x20
[Pass] FIO_SeekSkipFile(f, 0x30, SEEK_CUR) => 0x50
[Pass] FIO_SeekSkipFile(f, -0x20, SEEK_CUR) => 0x30
[Pass] FIO_GetFileSize_direct("test.dat") => 0x82000020
[Pass] FIO_RemoveFile("test.dat") => 0x0


I've tried as you said, and I've firstly unchecked all of the Overlay tab options, then checked them one by one. It happens when I enable the Magic Zoom (F+HS, Med). I've tried it with all the other options checked and unchecked, and it seems that this is what's causing the problem.
Also when I've changed the picture size from either largest L or RAW to S3 I've got the error message Error 70.
On the previously mentioned picture size, the camera just freezes without giving any error message.

Thanks, didn't get this reproduced with Magic Zoom + raw (though I did get an error message about "Raw error, falling back to YUV" or similar, probably from raw histogram I had enabled), but got it to hang with Magic Zoom + S3, without error message though. I'll enable some debug options in my build and see if I get any meaningful info out of it.

I was using the version labeled 2014Jul18 and previous. On the latest one, labeled 2014Jul22 it seems to be working.
It was the same with some older versions in these past 2 months.

So I guess this issue was fixed now.

EDIT: now sometimes works, sometimes it doesn't.
What specific settings should I tell you?

The difference now is that it works when you do exposure bracketing but if you do just continuous shooting, at some point it freezes.

Ah ok, I tried this as well, took 30 frame "burst" without problems. By info options I meant the options in Overlay tab (spot meter, zebras, focus peaking, histogram, etc.). Also any other settings that might be relevant like whether you shoot raw or jpeg (I used raw) and so on?

Greetings!
It's my fist time posting on this forum, and I'm using ML on an EOS M.
Generally, I can say I'm quite pleased with the features it provides but I'm here to report a bug, since i don't know where else to do so.

So I've tried the following settings:
- manual mode with exposure bracketing of 1 stop
- continuous shooting
- ML screen

When I'm doing this, the camera freezes and it won't respond not even to the shut down button. Sometimes the screen makes some gray and red lines and it stays like that. After removing the battery all is good, but the pictures aren't saved on the memory card.
This only happens with the ML screen. If I try using any other canon screen (by pressing the info button) then nothing like this happens.

I tried reproducing this but couldn't, took quite many brackets without problems (apart from a cosmetic problem of zebras filling the screen after the first frame). I use my own compiled binary, but it should be roughly the same as nightly from yesterday. Which info options you have enabled? I have disabled spot meter, only use focus peaking and zebras, otherwise default options.

Tried finding the CCDWriteEDmacCompleteCBR from EOSM.202 rom, but couldn't. No problems finding it from 5D3.123 rom so I think it's just not there. Any ideas where to look next? I blindly tried the values used for other cameras, but just got error "Exposure too long".

Edit: If I just comment out the check for 15s exposures, I get somewhat correct images with RAW_PHOTO_EDMAC value used by 5D3 and 700D. A bit overexposed but quite good.

Take a look here: http://www.magiclantern.fm/forum/index.php?topic=9581

Main problem: you will need more than 16 bits, so you'll need a floating point DNG. But the only program that can open them is Adobe SDK (not Lightroom nor ACR). In some extreme cases, even the plain dual iso files will push the limit of the 16-bit integer DNGs.

(well, you can also open floating point DNGs with my hacked copy of ufraw, but that's far from a proper solution)

Interesting as Lightroom and ACR can definitely open floating point TIFFs (both 16 and 32 bits per component)...

New XCode doesn't include GCC anymore and uses Clang in its place, which doesn't recognize this GCC option, but it's not needed either, so you can safely remove it from Makefile.cr2hdr.

Has anyone had good or bad experiences with using the modules from the 5D3 1.1.3 nightly builds? Or is that a no-no?

1.1.3 modules don't work with 1.2.3.

MacPorts has gnutar, though iirc bsdtar works just as well when symlinked as far as ML is concerned.

the problem is , new lenses is coming or arrived yet. f.e. the 24-70 II , etc. With the 1.1.3 firmware , the 5DIII does not recognize this new lenses , and we cant use any features on the cam like "automatic vignetting correction", Chromatic aberration correction, etc,etc. That´s a big problem...but...

In-camera lens profiles can be updated/added/removed with EOS Utility regardless of firmware version, I have 1.1.3 and 24-70 II is detected just fine.

Hi, there have been a number of people coming to the ML IRC channel lately having problems not getting their forum activation emails. Any pointers/help we could give them?