Loading complete image after border render crashes Blender #25775

New Issue

Maurice Raybaud · 2011-01-23T16:43:00+01:00

Maurice Raybaud commented

2011-01-23 16:43:00 +01:00

%%%When using POV-Ray exporter, border render generates a padded full sized image, where blender internal expects a cropped tile. It crashes.

An old functionality for partial image reading removed by Brecht Van Lommel, and which he would not disagree to see restored might provide a way to read an image that is different from the expected file size and allow the POV rendered image to load back to Blender when using border render. %%%

%%%When using POV-Ray exporter, border render generates a padded full sized image, where blender internal expects a cropped tile. It crashes. An old functionality for partial image reading removed by Brecht Van Lommel, and which he would not disagree to see restored might provide a way to read an image that is different from the expected file size and allow the POV rendered image to load back to Blender when using border render. %%%

Maurice Raybaud commented

2011-01-23 16:43:00 +01:00

Changed status to: 'Open'

Campbell Barton commented

2011-01-23 16:45:37 +01:00

%%%even if old functionality is added back the crash should be fixed,%%%

Campbell Barton commented

2011-02-08 11:57:14 +01:00

%%%note, the offset value in do_merge_tile() can become negative which causes a crash and points to the renderlayer created from python being incorrect however the values here don't make any sense to me since the max values are always smaller then the min values even when using blender internal.

assigning to Ton.%%%

%%%note, the offset value in do_merge_tile() can become negative which causes a crash and points to the renderlayer created from python being incorrect however the values here don't make any sense to me since the max values are always smaller then the min values even when using blender internal. assigning to Ton.%%%

Ton Roosendaal commented

2011-02-19 12:13:23 +01:00

%%%I'll pass it on to Brecht, getting this tested will just give me too much overhead :)%%%

Bastien Montagne commented

2011-05-12 10:48:05 +02:00

%%%(below might be somewhat redundant info…)

As far as I know, the problem is that, in RE_engine_begin_result (source/blender/render/intern/source/pipeline.c), when the external render tries to get a RenderResult corresponding to the border-rendered region (e.g. [100, 50, 300, 200]), the current re->result rectx/y is [200, 150] (i.e. the size of the “render display” automatically set by Blender), which will obviously gives wrong results (and segfaults at the end), due to the tests and clamping done on the given coordinates and dimensions.

Commenting out these tests (as in attached patch) makes the border render work with external engines and py API (at least, with the tested POV Ray addon).

However, I do not truly understand why the problem does not arise with internal render? And I’m not sure at all erasing those tests are a good thing (I don’t expect it to break anything, but I guess it introduces possible problems if render scripts make mistakes at this point!). It might be better to invalidate the resizing of the render display when using external engines? Or otherwise solve that mismatch? Thing is, I’m not enough at ease with this part of the code yet to propose a better solution…%%%

%%%(below might be somewhat redundant info…) As far as I know, the problem is that, in RE_engine_begin_result (source/blender/render/intern/source/pipeline.c), when the external render tries to get a RenderResult corresponding to the border-rendered region (e.g. [100, 50, 300, 200]), the current re->result rectx/y is [200, 150] (i.e. the size of the “render display” automatically set by Blender), which will obviously gives wrong results (and segfaults at the end), due to the tests and clamping done on the given coordinates and dimensions. Commenting out these tests (as in attached patch) makes the border render work with external engines and py API (at least, with the tested POV Ray addon). However, I do not truly understand why the problem does not arise with internal render? And I’m not sure at all erasing those tests are a good thing (I don’t expect it to break anything, but I guess it introduces possible problems if render scripts make mistakes at this point!). It might be better to invalidate the resizing of the render display when using external engines? Or otherwise solve that mismatch? Thing is, I’m not enough at ease with this part of the code yet to propose a better solution…%%%

Ton Roosendaal commented

2011-06-19 13:44:35 +02:00

%%%Internal render system uses own buffers (per tile) and copies this to the final 'result' with a function respecting the border and actual image size.
To me it seems the external engines should do same?%%%

%%%Internal render system uses own buffers (per tile) and copies this to the final 'result' with a function respecting the border and actual image size. To me it seems the external engines should do same?%%%

Campbell Barton commented

2011-10-07 00:48:27 +02:00

%%%@Ton, this isn't about the tile system, even though the bug may be here, the problem is that blender crashes when reading an un-cropped border render from an external engine.%%%

Brecht Van Lommel commented

2011-12-20 13:26:09 +01:00

%%%Fix in svn, thanks for the report.%%%

Brecht Van Lommel commented

2011-12-20 13:26:10 +01:00

Changed status from 'Open' to: 'Resolved'

Brecht Van Lommel closed this issue

2011-12-20 13:26:10 +01:00

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

Loading complete image after border render crashes Blender #25775