[DONE] How do I dump 5D3 firmware 1.2.3?

Started by chris_overseas, January 05, 2014, 06:18:47 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.


Quote from: pompeiisneaks on February 05, 2014, 04:39:17 AM
1 is 1080p and had horrid colors all messed up.  2 is just fine, right after the first failure, and 3 is at a lower resolution so that I could record to my slower card.  It also was all messed up.

I've got these artifacts today with silent pictures. Couldn't find a way to reproduce them though.


Good to know I'm not the only one.  a1ex, are you using the 1.2.3 build on 5D3 or is this something global?



I took thousands of silent pictures back then with 1.1.3 (before raw_rec) and I don't remember seeing them.

Can't tell if it's a quirk in 1.2.3 or something introduced in latest ML code (first I need a way to reproduce them, then I can downgrade to 1.1.3 and try the same thing).


chris_miller corrected RAW_TYPE_ADDRESS 1 hour ago, maybe it will help


Didn't help. I've reproduced it with silent pics in burst mode, and sometimes slit-scan.

Went back to 113, everything clean.


As far I remeber this function removed all bad frames on 60D:
/* get exclusive access to our edmac channels */

Maybe they changed something with edmac ?


Some bad pictures were in slit-scan mode, where I'm not doing any kind of EDMAC trickery. That code works with plain old memcpy from Canon's buffer.

I should probably run some tests without any ML code loaded (from a minimal boot-hack).

Jason Montalvo

Quick question, I recently got my camera back from a clean and check from canon where I specifically asked them not to upgrade my firmware. they complied and left it alone, but when I got it back, I had to reload ML on the camera.

Now since i've reloaded ML the ML card has to always be in the camera or I get:

"Update file error. Please check the memory card and reload the battery and try again"

I've tried putting reenabling the boot flag made my sd card bootable again with no such luck.  The reason I discovered was because I was trying to update to 1.2.3

Thanks in advance !


Getting clean hdmi recording and raw recording so far  :D Going to do some test with both at once   :-\


123 clean HDMI works well with ninja2. Tested crop modes at x5 & x10, they record to ninja cropped while recording raw & MLV, but at a reduced frame size. Would be nice to have timecode active to trigger ninja2 when shutter is pressed. More testing to come.


Corrupted frames fixed hopefully; please test and let me know if it actually works.



MA Visuals

Quote from: Audionut on February 12, 2014, 10:04:52 AM
An updated build.  This one comes with the same warranty  as the last build.


Does this build include the pink frame fix Alex recently implemented?



Can someone set up a platform entry 5D3_1.2.3 for these new builds in the nightlies?

While making it user friendly takes ressources from development, it also frees
resources through preventing double and triple questions about ML in first place.

Thanks everybody for your great work.

Also: What needs to be done to solve the bootflag problem?
Laying out the problems regarding porting stubs and record restart
helped chris_miller and viciniusatique to contribute.

Maybe a quick word on this could be helpful.

Regards, Michael Quack


Please don't rush it. If you want to provide feedback on the early builds, please do. If not, stay calm.


I know it's a stupid question but how can I download ML repo files and then update it with modifications from pull request "#354: Ported ML over to 5D3 firmware 1.2.3." using hg ?



Quote from: Audionut on February 13, 2014, 10:27:53 AM
Probably easier just to build straight from the 1.2.3 repo:  https://bitbucket.org/chris_miller/magic-lantern/overview

I'm not sure if all commits are pushed to this repo, for example latest memory backend..


I've just updated it with the latest bleeding-edge stuff.

However, the memory backend is not as rock-solid as on 60D (it fails at g3gg0's tests after less than a minute).


Quote from: a1ex on February 13, 2014, 11:13:57 AM
I've just updated it with the latest bleeding-edge stuff.

However, the memory backend is not as rock-solid as on 60D (it fails at g3gg0's tests after less than a minute).

thx a1ex
is there any impact to MLV_REC ?


I don't know, you should tell me that ;)

What I know right now is that we have a memory leak with repeated malloc(16MB)/free calls, but the underlying call (_shoot_malloc/_shoot_free) is fine. If you want to help me track it down, you are welcome.



I can test something on new firmware after work. Any suggestions ? :)


Anything that was touched in recent changesets :D