Fix T84167: Saving half-float EXR might result in NaN pixels
ClosedPublic
Actions

Authored by Sergey Sharybin (sergey) on Jan 13 2021, 2:59 PM.

Details

Reviewers

Brecht Van Lommel (brecht)

Commits

rB53bd58993e2f: Fix T84167: Saving half-float EXR might result in NaN pixels

Summary

Clamp value to the HALF_MAX.

Clamp value to the -HALF_MAX .. HALF_MAX.

The non-clamped values were causing NaN and inf values saved to
the file, which was the root cause of glare node giving unexpected
result.

The nan/inf on overflow is something mentioned in the half data
type in OpenEXR header.

Diff Detail

Repository: rB Blender

Event Timeline

Sergey Sharybin (sergey) requested review of this revision.Jan 13 2021, 2:59 PM

Sergey Sharybin (sergey) created this revision.

Harbormaster completed remote builds in B12167: Diff 32730.Jan 13 2021, 2:59 PM

I guess this needs to clamp to -HALF_MAX..HALF_MAX? Large negative numbers are not as likely, but still possible.

This revision now requires changes to proceed.Jan 13 2021, 3:03 PM

Clamp to -HALF_MAX .. HALF_MAX

Harbormaster completed remote builds in B12168: Diff 32739.Jan 13 2021, 3:20 PM

Sergey Sharybin (sergey) edited the summary of this revision. (Show Details)Jan 13 2021, 3:20 PM

Brecht Van Lommel (brecht) accepted this revision.Jan 13 2021, 3:21 PM

This revision is now accepted and ready to land.Jan 13 2021, 3:21 PM

Closed by commit rB53bd58993e2f: Fix T84167: Saving half-float EXR might result in NaN pixels (authored by Sergey Sharybin (sergey)). · Explain WhyJan 14 2021, 3:19 PM

This revision was automatically updated to reflect the committed changes.

Sergey Sharybin (sergey) added a commit: rB53bd58993e2f: Fix T84167: Saving half-float EXR might result in NaN pixels.

Revision Contents
Changeset List

Path

Size

source/

blender/

imbuf/

intern/

openexr/

openexr_api.cpp

15 lines

Diff 32775

View Options

source/blender/imbuf/intern/openexr/openexr_api.cpp

Fix T84167: Saving half-float EXR might result in NaN pixelsClosedPublicActions