600D (digic IV?) Raw zebras go nuts.

Started by 1%, June 08, 2013, 07:09:04 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

1%

Raw zebras are going nuts on 600D in movie mode... made worse with a fully manual lens.

Underexpo zebras have a bunch of boxes that show up on screen... in zoom mode only underexposed zebras show up and don't redraw as often. Zoomed out the overexpo zebaras have problems like the underexposed ones. In photo mode they seem OK.

Some videos:




1%

Also zebras or histo + magic zoom = EVF state assert.

1%

Eliminated a lot of things... its something with cbr vsync for modules... even with old_state 5 vsync taken out it still crashes while recording. Will try a copy&restart boot to make sure its not memory and then I guess patch the assert and see if behavior changes...

btw.. reloc =y is broken for individual makefiles.

Somewhat fixed:

This broke it, other cameras seemed fine:
https://bitbucket.org/hudson/magic-lantern/commits/7bc96dff0edb9a01d342156d1b880de5239cbe07

still crashes out on 720P.. when you path the assert 1190 the camera gets really slow with MZ running

a1ex

are you sure? that changeset only draws zebra less often...

edit: I recommend disabling everything from features.h, maybe some things from internals.h, until you isolate the problem. You can also prevent certain tasks from starting, in boot-hack.c.

You shouldn't disable asserts in cases like this. If you do that, you are only hiding the error instead of solving it (it's like treating the symptom and not the cause), and you don't even understand what's going on.

This practice is dangerous and sooner or later it may brick the camera. Seriously.

1%

Reverting that changeset which I know was supposed to slow them down "fixes" it. Still draws excessively in movie mode but not crashy.

Maybe CPU + memory isn't enough to run module/zebra/mz/histogram? The patch was only to see what would happen, it didn't stay. Still asserts on 60P for some reason just with modules, but only at full FPS.

Assert 1190 says something about 32 frames continuous... its in the evfreadoutdoneInterrupt function.

Also strange that in photo mode it doesn't redraw too much. Has issues with display filters too, if you turn something on that uses it, the filter gets flashy and out of sync.

Also noticed its mainly happening to under-exposure zebra (in a scene with both, over not flashing)... like it goes back and forth between detecting the left skip for black level and not detecting it. Maybe needs another video.

1%

Here is some log:


#  1097: 25118.262 [LV] ERROR [EVF] NotifyHeadErrorCBR
#  1098: 25118.374 [LV] ERROR [PATH] VRAM[W](0xc2385c0)   YUV[W](0x6000080)
#  1099: 25118.453 [LV] ERROR [PATH] FENC[W](0xf574580) QUARK[W](0x0)
#  1100: 25118.530 [LV] ERROR [PATH] FLIC[W](0xb6fc300)  TA10[R](0x5763c0)
#  1101: 25118.599 [LV] ERROR [PATH] JPEG[R](0x729fc0)  JPEG[W](0x0)
#  1102: 25139.348 [LV] ERROR [PATH] Path Incomplete[101]
#  1103: 25139.579 [LV] ERROR [PATH] (Rec_HD)CRAW[W](0x46332200->0x1400180)
#  1104: 25139.758 [LV] ERROR [PATH] (Rec_HD)VRAM[W](0x4c233800->0xc2d5920)
#  1105: 25140.161 [LV] ERROR [EVF] 32 frames continuous EngineError!!!
#  1106: 25151.233 [STARTUP] ERROR #ASSERT : Evf\EvfState.c, Task = Evf
#  1107: 25151.347 [STARTUP] ERROR #ASSERT : Line 1190
#  1108: 25151.403 [STARTUP] ERROR #ASSERT : 0
#  1109: 25151.504 [STARTUP] #ASSERT : Time 2013/6/9 2:50:13
#  1110: 25152.075 [STARTUP] startupErrorRequestChangeCBR (0x1d)
#  1111: 25152.243 [STARTUP] startupErrorRequestChangeCBR : ErrorSend (101, ABORT)
#  1112: 25158.494 [LV] ERROR [EVF] Blanking DeadLine


Latest seems to have gotten slightly better but... here is underexpo fluctuating

.

FS magic zoom + raw zebras does this:

a1ex

Ouch...

The error may indicate something in the vsync functions; maybe disable CONFIG_STATE_OBJECT_HOOKS first, then narrow down in the vsync calls?

1%

It does seem vsync related... I tried turning off all vsync lite stuff and evf sync... will try without state object hooks.