[2/3] Regression: huffmantable is bad for last decode about rawspeed HOT 7 CLOSED

HURRAY!
Thanks for checking!

from rawspeed.

There are no more problems with e9a586f in rawspeed. Thanks!

from rawspeed.

@exaexa hi. (+ others affected, who tried this)
With this fix, did you ever seen any NEF's still failing to decode?
Or does this appear to fully resolve the situation?

from rawspeed.

With this fix

Which one, the workaround or the "unfucking" commit from May?

I'm pretty sure there weren't any extra issues with the workaround and I think that I'm already on a newer version without the workaround and with the commit. Going to verify ASAP.

from rawspeed.

Negative, the problem is still present (tested on darktable+rawspeed from git; rawspeed commit 2bc2bf5 is there)

You can get an affected file for testing here: http://e-x-a.org/stuff/_DSC0193.NEF (not sure if the previous sample I posted to pixls got fixed or not, but it seems more like not).

from rawspeed.

With this fix

Which one, the workaround or the "unfucking" commit from May?

The this one in this issue:

Changing

rawspeed/src/librawspeed/decompressors/HuffmanTable.h

Line 309 in d41fd6d

bs.fill(32);

-bs.fill(32);
+bs.fill(maxCodePlusDiffLength());

seems to help somewhat, but i don't know the full solution yet.

from rawspeed.

@exaexa please do tell if any of the raws are still broken after this.

from rawspeed.