Image artifacts after updating to 3.0

[mattleibow]

Have you searched the existing issues (both open and closed) in the libjpeg-turbo issue tracker to ensure that this bug report is not a duplicate?
Yes

Does this bug report describe one of the two known and unsolvable issues with the JPEG format?
No

Clear and concise description of the bug:

I am using skia to encode jpegs and all is OK in the 2.x series. However, once updating to 3.x I am getting invalid output files. Now I know this is the worst bug where I say I am using layers of tech and not even tried yet with plain libjpeg-turbo but was wondering if the issues I am getting opened my ring some bells before I start diving into this.

• [BUG] SkiaSharp version 2.88.6 has mosaics when loading images in .NET 7 and the background becomes dark. mono/SkiaSharp#2650
• [BUG] Xamarin : Broken Bitmap.Encode on real devices mono/SkiaSharp#2643
• [BUG] SKBitmap.Encode issue with Skiasharp 2.88.6 on .NET 8.0 MacOS and iOS mono/SkiaSharp#2666

Reports are saying there are some invalid bytes in the files:

For example in this comment: mono/SkiaSharp#2643 (comment)

Steps to reproduce the bug (using only libjpeg-turbo):
Not yet done, but I am working on it

Image(s) needed in order to reproduce the bug (if applicable):
Any image that is read and then re-encoded.

Expected behavior:
No artifacts

Observed behavior:
Artifacts

Platform(s) (compiler version, operating system version, CPU) on which the bug was observed:
Seems to be arm64. Reported on iOS and Android, and I am getting this on my arm64 mac.

libjpeg-turbo release(s), commit(s), or branch(es) in which the bug was observed (always test the tip of the main branch or the latest stable pre-release to verify that the bug hasn't already been fixed):
3.0.0, 3.0.1 and main

If the bug is a regression, the specific commit that introduced the regression (use git bisect to determine this):
TBD

Additional information:

Dodgy report, I know. As a software dev myself I know I just am being a pain. But, I need something to link to and work on as I try get my code reverted. I will try isolate the issue but hopefully the jpeg community has an "aha" moment.

[dcommander]

I am the sole developer of libjpeg-turbo, so there isn't really a "community" to answer bug reports. They all fall upon me. There isn't enough information yet to be actionable on my part. I need to understand how the downstream software is using libjpeg-turbo. There weren't any changes between 2.1.x and 3.0.x that would readily explain this.

[mattleibow]

OK, thanks. I will try get a sample using the layers of APIs on top. Maybe skia has a bug in the way they are using things - or it may be that we are using the wrong 12- or 16- bit something. Or my compiler flags.

[dcommander]

3.0.2 is overdue, so the sooner you can determine whether this is actually a libjpeg-turbo bug, the better.

[dcommander]

@mattleibow Any update on this issue?