image > copy, adds 3 white pixels on the left, deletes 3 pixels on the right #118351

New Issue

LOIC BRAMOULLE · 2024-02-15T22:58:27+01:00

LOIC BRAMOULLE commented

2024-02-15 22:58:27 +01:00

System Information
Operating system: Windows-10-10.0.19045-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3090/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 546.01

Blender Version
Broken: version: 3.6.4, branch: blender-v3.6-release, commit date: 2023-09-25 13:24, hash: 21bfc5e7fe3f
Worked: (newest version of Blender that worked as expected)

Short description of error
image > copy, adds 4 white pixels on the left, deletes 4 pixels on the right

Exact steps for others to reproduce the error
1 - render anything.
2 - in render result go Image > copy, then paste in photoshop.
3 - you will have a 3 pixels line added on the left, and 3 pixel line missing on the right, shifting the whole image 3 pixels to the right basically.

**System Information** Operating system: Windows-10-10.0.19045-SP0 64 Bits Graphics card: NVIDIA GeForce RTX 3090/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 546.01 **Blender Version** Broken: version: 3.6.4, branch: blender-v3.6-release, commit date: 2023-09-25 13:24, hash: `21bfc5e7fe3f` Worked: (newest version of Blender that worked as expected) **Short description of error** image > copy, adds 4 white pixels on the left, deletes 4 pixels on the right **Exact steps for others to reproduce the error** 1 - render anything. 2 - in render result go Image > copy, then paste in photoshop. 3 - you will have a 3 pixels line added on the left, and 3 pixel line missing on the right, shifting the whole image 3 pixels to the right basically.

LOIC BRAMOULLE added the

labels 2024-02-15 22:58:28 +01:00

Pratik Borhade commented

2024-02-16 07:08:23 +01:00

Hi, thanks for the report. I'm not able to confirm this in 4.1/3.6
Not sure if render/output setting would make the difference here. Can you attach .blend where problem is visible?
Did you try pasting the image in other software? I tried in ms-paint and blender itself btw :)

Hi, thanks for the report. I'm not able to confirm this in 4.1/3.6 Not sure if render/output setting would make the difference here. Can you attach .blend where problem is visible? Did you try pasting the image in other software? I tried in ms-paint and blender itself btw :)

Pratik Borhade added

Status

Needs Information from User

and removed

Status

Needs Triage

labels 2024-02-16 07:08:33 +01:00

LOIC BRAMOULLE commented

2024-02-16 15:44:22 +01:00

Hi Pratik, damn, must come from my blender environment maybe, it happens on any blend file, and any program when pasting, apart from blender itself if I paste back as a new image. (but the 3 pixels are in PS, paint ect..)
Not a showstopper I can just save + open images if it's only me.
It's pretty strange it's actually the 3 far right pixels of the images that are slapped on the far left, effectlively sliding the image 3 pixel to the right. (so not a white line, but a re-order.)

Hi Pratik, damn, must come from my blender environment maybe, it happens on any blend file, and any program when pasting, apart from blender itself if I paste back as a new image. (but the 3 pixels are in PS, paint ect..) Not a showstopper I can just save + open images if it's only me. It's pretty strange it's actually the 3 far right pixels of the images that are slapped on the far left, effectlively sliding the image 3 pixel to the right. (so not a white line, but a re-order.) ![image](/attachments/e9276b78-fb13-44e9-a90f-d8b6c25c4ab3)

image.png

23 KiB

TheRedWaxPolice commented

2024-02-16 16:17:55 +01:00

Yep, I can confirm the bug in 4.2 blender-4.2.0-alpha+main.d803b6acca19-windows.amd64-release

And yes, if you copy and then paste it in blender itself, the problem doesn't occur...

Yep, I can confirm the bug in 4.2 `blender-4.2.0-alpha+main.d803b6acca19-windows.amd64-release` ![image](https://projects.blender.org/attachments/e2680f98-7981-470b-af55-c713b6f51524) And yes, if you copy and then paste it in blender itself, the problem doesn't occur...

image.png

252 KiB

👍 1

LOIC BRAMOULLE commented

2024-02-16 22:08:37 +01:00

if it helps, sometimes it's also 2 pixels vertically:

if it helps, sometimes it's also 2 pixels vertically: ![image](/attachments/02cf7fee-fce5-433c-9c3f-8e7fabffbc2b)

image.png

17 KiB

Jesse Yurkovich commented

2024-02-17 02:43:53 +01:00

I'll confirm. I can reproduce in Krita, but not in Gimp - by default at least.

When Blender puts the image on the clipboard it does so in 2 different formats: the Windows DibV5 bitmap format and a regular PNG format which applications will then choose from on their own. If I change the code to only use PNG then Krita and Gimp work fine. If I use just DibV5 then both Krita and Gimp show the shifted pixels.

CC @Harley do you remember why DibV5 offsets by + 12 for the pixel data? This seems to be the issue in this case.

I'll confirm. I can reproduce in Krita, but not in Gimp - by default at least. When Blender puts the image on the clipboard it does so in 2 different formats: the Windows DibV5 bitmap format and a regular PNG format which applications will then choose from on their own. If I change the code to only use PNG then Krita and Gimp work fine. If I use just DibV5 then both Krita and Gimp show the shifted pixels. CC @Harley do you remember why DibV5 offsets by `+ 12` for the pixel data? This seems to be the issue in this case.

Jesse Yurkovich added

Status

Confirmed

and removed

Status

Needs Information from User

labels 2024-02-17 02:44:02 +01:00

Jesse Yurkovich added the

Module

User Interface

label 2024-02-17 02:45:04 +01:00

Harley Acheson commented

2024-02-17 03:39:09 +01:00

@deadpin

There is something weird about the specification of that format and I'm probably just doing something wrong.

They define a header with all its members, but then tell you that if one member (bV5Compression) is set to BI_BITFIELDS then there is 3 dword mask values between the header and the bitmap.

When we GET that format from the clipboard I see that we are checking for that state and offsetting by 12 bytes. But when we PUT that format on the clipboard we do set that bV5Compression to BI_BITFIELDS (meaning no compression) and also do that offsetting.

Although when I am looking around for this I am not seeing this in Microsoft docs right now but only references to it in other peoples posts. So perhaps there is a difference with different versions of this structure - we are using V5, as in BITMAPV5HEADER so maybe this extra crap between header and data is only for an older version?

Sorry if you explained it well and I'm not getting it, but are we putting this format on the clipboard and others see it off by three pixels, or are we offsetting by three pixels when we read this format from others? As in are we broken coming in, going out, or both?

@deadpin There is something weird about the specification of that format and I'm probably just doing something wrong. They define a header with all its members, but then tell you that if one member (bV5Compression) is set to BI_BITFIELDS then there is 3 dword mask values between the header and the bitmap. When we GET that format from the clipboard I see that we are checking for that state and offsetting by 12 bytes. But when we PUT that format on the clipboard we do set that bV5Compression to BI_BITFIELDS (meaning no compression) and also do that offsetting. Although when I am looking around for this I am not seeing this in Microsoft docs right now but only references to it in other peoples posts. So perhaps there is a difference with different versions of this structure - we are using V5, as in BITMAPV5HEADER so maybe this extra crap between header and data is only for an older version? Sorry if you explained it well and I'm not getting it, but are we putting this format on the clipboard and others see it off by three pixels, or are we offsetting by three pixels when we read this format from others? As in are we broken coming in, going out, or both?

Harley Acheson commented

2024-02-17 03:54:48 +01:00

The following is a discussion of similar problem with Paint.net with their main dev looking like he read the spec like I did and did it wrong too. This comment from a MS employee looks to explain it well,

The V1 version of that structure had this optional extra 12 bytes, but then those became part of the structure later. The docs just don't explain that well:

https://forums.getpaint.net/topic/122848-pasting-dibv5-image/#comment-610918

But even his comment ends with "Now, even though I claim my interpretation of the structures to be "the right one" it can also be argued that they are both correct. The documentation has been changed so many times with regard to this and it's still broken, some parts of Windows write one variant, other parts write another and it can usually read them both."

The following is a discussion of similar problem with Paint.net with their main dev looking like he read the spec like I did and did it wrong too. This comment from a MS employee looks to explain it well, The V1 version of that structure had this optional extra 12 bytes, but then those became part of the structure later. The docs just don't explain that well: https://forums.getpaint.net/topic/122848-pasting-dibv5-image/#comment-610918 But even his comment ends with "Now, even though I claim my interpretation of the structures to be "the right one" it can also be argued that they are both correct. The documentation has been changed so many times with regard to this and it's still broken, some parts of Windows write one variant, other parts write another and it can usually read them both."

Harley Acheson commented

2024-02-17 04:07:33 +01:00

@deadpin - If I change the code to only use PNG then Krita and Gimp work fine. If I use just DibV5 then both Krita and Gimp show the shifted pixels.

Oh, when I was testing this the latest version of Gimp at the time wouldn't use this format, if I remember right, which is why I started down the road of also placing PNG as that was the only other format supporting transfer of alpha. And I only had a very old version of Photoshop. So for this format I was only testing with Paint and Krita, but (hopefully) just missed noticing it was out by three pixels.

> @deadpin - If I change the code to only use PNG then Krita and Gimp work fine. If I use just DibV5 then both Krita and Gimp show the shifted pixels. Oh, when I was testing this the latest version of Gimp at the time wouldn't use this format, if I remember right, which is why I started down the road of also placing PNG as that was the only other format supporting transfer of alpha. And I only had a very old version of Photoshop. So for this format I was only testing with Paint and Krita, but (hopefully) just missed noticing it was out by three pixels.

Harley Acheson commented

2024-02-17 20:05:06 +01:00

@deadpin - If I use just DibV5 then both Krita and Gimp show the shifted pixels.

Yes, seems to test out well. But for testing I have to alter the code to ensure that only the formats I want to test are used, with breakpoints set to confirm.

Load image into Gimp 2.10.36
Copy and Paste to Blender (as DibV5)
Copy and Page to new layer in Gimp (as DibV5)
These layers show 3 pixel shift
Load image into Gimp 2.10.36
Copy and Paste to Blender (as PNG)
Copy and Page to new layer in Gimp (as DibV5)
These layers show 3 pixel shift

But, weirdly, If load an image in Krita, then copy to Gimp and also copy to Blender (as DibV5). Then copy and paste from Blender to a new layer in Gimp (as DibV5) they are perfectly aligned. But then copy and paste from Gimp to Blender and back as new layer and they are misaligned.

So far have managed to recreate the issue with Gimp, but not with Krita. But removing those 12 bytes it fixes the issue with Gimp and does not cause any errors with Krita.

This, and my (new) reading of the spec based on what that MS guy is saying, seems like a lock.

Although I don't have a new version of Photoshop to test DibV5 transfers the error I see with Gimp exactly matches what is described with Photoshop.

> @deadpin - If I use just DibV5 then both Krita and Gimp show the shifted pixels. Yes, seems to test out well. But for testing I have to alter the code to ensure that only the formats I want to test are used, with breakpoints set to confirm. * Load image into Gimp 2.10.36 * Copy and Paste to Blender (as DibV5) * Copy and Page to new layer in Gimp (as DibV5) * These layers show 3 pixel shift * Load image into Gimp 2.10.36 * Copy and Paste to Blender (as PNG) * Copy and Page to new layer in Gimp (as DibV5) * These layers show 3 pixel shift But, weirdly, If load an image in Krita, then copy to Gimp and also copy to Blender (as DibV5). Then copy and paste from Blender to a new layer in Gimp (as DibV5) they are perfectly aligned. But then copy and paste from Gimp to Blender and back as new layer and they are misaligned. So far have managed to recreate the issue with Gimp, but not with Krita. But removing those 12 bytes it fixes the issue with Gimp and does not cause any errors with Krita. This, and my (new) reading of the spec based on what that MS guy is saying, seems like a lock. Although I don't have a new version of Photoshop to test DibV5 transfers the error I see with Gimp exactly matches what is described with Photoshop.

Harley Acheson referenced a pull request that will close this issue

2024-02-17 20:28:06 +01:00

Fix #118351: Pixel Shift with Win32 Clipboard DibV5 Image Transfer #118417

Harley Acheson referenced this issue from a commit

2024-02-17 20:29:53 +01:00

Fix #118351: Pixel Shift with Win32 Clipboard DibV5 Image Transfer

Harley Acheson closed this issue

2024-02-17 20:29:56 +01:00

Blender Bot added

and removed

labels 2024-02-17 20:29:57 +01:00

Pratik Borhade commented

2024-02-19 10:09:46 +01:00

Broken: version: 3.6.4, branch: blender-v3.6-release

Given 3.6 is affected, can we add your fix in 3.6 backporting list? @Harley

And thanks for confirming and fixing this report. I usually don't check bug reports during weekend 😅

> Broken: version: 3.6.4, branch: blender-v3.6-release Given 3.6 is affected, can we add your fix in 3.6 backporting list? @Harley And thanks for confirming and fixing this report. I usually don't check bug reports during weekend 😅

Harley Acheson referenced this issue

2024-02-19 17:54:32 +01:00

Blender LTS: Maintenance Task 3.6 #109399

Harley Acheson commented

2024-02-19 17:56:10 +01:00

@PratikPB2123 - Given 3.6 is affected, can we add your fix in 3.6 backporting list?

I just added this to that list.

> @PratikPB2123 - Given 3.6 is affected, can we add your fix in 3.6 backporting list? I just added this to [that list](https://projects.blender.org/blender/blender/issues/109399).

Philipp Oeser referenced this issue from a commit

2024-03-12 10:44:39 +01:00

Fix #118351: Pixel Shift with Win32 Clipboard DibV5 Image Transfer

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

image > copy, adds 3 white pixels on the left, deletes 3 pixels on the right #118351