Reverting pipe stuff #18360
Reverting pipe stuff #18360
Conversation
jenshannoschwalm
added
bugfix
This reverts commit 48282d4.
This reverts commit e8c2b95.
- added/fixed comment about position - some constify while being here - a hard-to-read static function name was renamed
As the hash is based upon position in the pipe - this is not iop_order - we have to find and use that.
As dt_dev_pixelpipe_cache_hash() requires the piece position in the pipe - not the iop_order of the module in question - we have to use that.
- for applied roi_in/out changes - for applied styles
Only enabled pieces/modules should be integrated into the hash
jenshannoschwalm
force-pushed
the
reverting_pipe_stuff
branch
from
bfafa32 to
014d699
Compare
|
Tested 10 minutes and no issues with the scenario I had reported. Lot better to me. Let's see what @wpferguson has to report with its torture test :) |
|
First test, processed 50 images without any problems. I have a couple more sets of images to process. Back to tortur... err, testing. |
|
Also tested a lot more concentrating on cache correctness (masks and pixelpipe cache hits) and couldn't spot any regression. |
|
Last force-push just fixed clearing |
|
@TurboGit in case this stays all good and the way to go, do you want me to create a 5.0 specific PR? |
jenshannoschwalm
removed
the
wip
Yes please! Are you on the #darktable-dev channel in https://matrix.to/#/#_oftc_#darktable:matrix.org ? (I'm asking because I've given the plan for 5.0.1 release there - BTW, this is important for all devs to be on this channel now that there is no more dev mailing-list). |
|
No I wasn't until now. Is the dev channel something different from plain darktable? |
Yes, that's another channel that I'm using to prepare the releases. |
|
Small correction, #darktable-dev can be joined at https://matrix.to/#/#darktable-dev:matrix.org |
|
Finished testing. Processed 3 sets of images with no problems. By the last set I was just editing and not worrying about problems, so I was going as fast as I normally do. 🎆 It works for me 😄 Thanks |
There was a problem hiding this comment.
Thanks a lot @jenshannoschwalm for the very hard work on this issue.
VERY BIG THANKS TO YOU! btw, after all this time you kept hammering and me trying to isolate the issue a) this now seems a cleaner solution and b) i still could not spot the code part where the translation to iop_order fails :-( |
@wpferguson @TurboGit @kofa73
This would be the proposal based on basic_hash being calculated on position instead of iop_order.