bevel_factor_mapping_start/end in python loop #40223

New Issue

Lukas Treyer · 2014-05-16T09:41:55+02:00

Lukas Treyer commented

2014-05-16 09:41:55 +02:00

System Information
OSX 10.9, Macbook Pro Retina

Blender Version
Broken: 2.70.5 3889483

Short description of error
When I try to set the bevel_factor_mapping_values from python, Blender crashes.

Could not unmap memory
blender(50999,0x116936000) malloc: ***error for object 0x11740c030: incorrect checksum for freed object - object was probably modified after being freed.*** set a breakpoint in malloc_error_break to debug
Abort trap: 6

Exact steps for others to reproduce the error
Open the attached .blend file run the code (should be there already)ZHTraffic.blend:

import bpy 

for o in bpy.data.objects:
    o.data.bevel_factor_mapping_start = 'SPLINE'
    o.data.bevel_factor_mapping_end = 'SPLINE'

**System Information** OSX 10.9, Macbook Pro Retina **Blender Version** Broken: 2.70.5 3889483 **Short description of error** When I try to set the bevel_factor_mapping_values from python, Blender crashes. ``` Could not unmap memory blender(50999,0x116936000) malloc: ***error for object 0x11740c030: incorrect checksum for freed object - object was probably modified after being freed.*** set a breakpoint in malloc_error_break to debug Abort trap: 6 ``` **Exact steps for others to reproduce the error** Open the attached .blend file run the code (should be there already)[ZHTraffic.blend](https://archive.blender.org/developer/F89191/ZHTraffic.blend): ``` import bpy for o in bpy.data.objects: o.data.bevel_factor_mapping_start = 'SPLINE' o.data.bevel_factor_mapping_end = 'SPLINE'

Lukas Treyer commented

2014-05-16 09:41:55 +02:00

Changed status to: 'Open'

Campbell Barton was assigned by Lukas Treyer

2014-05-16 09:41:55 +02:00

Lukas Treyer commented

2014-05-16 09:41:55 +02:00

Added subscriber: @cnd

Campbell Barton commented

2014-05-16 10:12:04 +02:00

crash is caused by segcount = (int)(bl->nr / cu->resolu) when bl->nr isnt dividing evenly
29 / 12 - in this case.

Also noticed calc_bevfac_mapping assumes cu->resolu, but resolu_ren could be used.
use_render_resolution arg should be passed by caller.

crash is caused by `segcount = (int)(bl->nr / cu->resolu)` when bl->nr isnt dividing evenly `29 / 12` - in this case. Also noticed `calc_bevfac_mapping` assumes cu->resolu, but resolu_ren could be used. `use_render_resolution` arg should be passed by caller.

Lukas Treyer commented

2014-05-16 11:36:52 +02:00

I think I could fix it: bevfac_for_poly_splines.diff

I think I could fix it: [bevfac_for_poly_splines.diff](https://archive.blender.org/developer/F89203/bevfac_for_poly_splines.diff)

Bastien Montagne commented

2014-05-16 11:49:19 +02:00

Added subscriber: @mont29

blender-admin commented

2014-05-16 17:01:43 +02:00

This issue was referenced by 1923a8f23a

This issue was referenced by 1923a8f23a174c343601771ab5136f2a7d38fc47

Campbell Barton commented

2014-05-16 17:03:28 +02:00

Changed status from 'Open' to: 'Resolved'

Campbell Barton closed this issue

2014-05-16 17:03:28 +02:00

Campbell Barton commented

2014-05-16 17:03:28 +02:00

Closed by commit 1923a8f23a.

Closed by commit 1923a8f23a.

Campbell Barton commented

2014-05-16 17:12:33 +02:00

Changed status from 'Resolved' to: 'Open'

Campbell Barton reopened this issue

2014-05-16 17:12:33 +02:00

Campbell Barton commented

2014-05-16 17:12:33 +02:00

Re-opening: curve_crash.blend

vector bezier handles make the segment count incorrect (since they only use one segment).

Re-opening: [curve_crash.blend](https://archive.blender.org/developer/F89225/curve_crash.blend) vector bezier handles make the segment count incorrect (since they only use one segment).

Lukas Treyer commented

2014-05-18 19:24:02 +02:00

inspired by "curve_to_displist()" I could fix the vector bezier handles issue.
Additionally I fixed a problem that became only visible with poly splines: swapping start and end factors occurred sometimes in the wrong place because last- and firstblend were not checked in the if-statement on line 1522. bevfac_for_poly_splines_2.diff

inspired by "curve_to_displist()" I could fix the vector bezier handles issue. Additionally I fixed a problem that became only visible with poly splines: swapping start and end factors occurred sometimes in the wrong place because last- and firstblend were not checked in the if-statement on line 1522. [bevfac_for_poly_splines_2.diff](https://archive.blender.org/developer/F89373/bevfac_for_poly_splines_2.diff)

Campbell Barton commented

2014-05-22 08:21:12 +02:00

@cnd, the patch isnt applying on a recent build, would you mind updating the diff?

Lukas Treyer commented

2014-05-25 02:45:15 +02:00

updating the diff bevfac_for_poly_splines_3.diff

updating the diff [bevfac_for_poly_splines_3.diff](https://archive.blender.org/developer/F90648/bevfac_for_poly_splines_3.diff)

Campbell Barton commented

2014-05-26 04:06:26 +02:00

Checked this patch and there are more issues, I'm testing with valgrind to report problems, while they don't cause crashes they are definitely bugs still.

First, the patch reverts a correction I made previously, when updating patches please check what changed, the reason the patch failed to apply was because a fix was made (check for zero cu->resolu_ren).
There is a mismatch between segcount used to alloc bevp_array, and the index used to read calc_bevfac_spline_mapping (reads past array bounds).
Cyclic curves also look to cause memory read errors with bevp_array, Im not sure this option even makes much sense for cyclic curves?, probably this should just be disabled?
While testing I keep getting errors with rotateBevelPiece, its reading past array bounds (checks for first/last ignore that bevp is offset by start)

I'm thinking maybe we should disable this for release, the more I check this feature the more bug's I keep finding.

Checked this patch and there are more issues, I'm testing with valgrind to report problems, while they don't cause crashes they are definitely bugs still. * First, the patch reverts a correction I made previously, when updating patches please check what changed, the reason the patch failed to apply was because a fix was made (check for zero `cu->resolu_ren`). * There is a mismatch between `segcount` used to alloc `bevp_array`, and the index used to read `calc_bevfac_spline_mapping` (reads past array bounds). * Cyclic curves also look to cause memory read errors with `bevp_array`, Im not sure this option even makes much sense for cyclic curves?, probably this should just be disabled? * While testing I keep getting errors with `rotateBevelPiece`, its reading past array bounds (checks for first/last ignore that bevp is offset by `start`) I'm thinking maybe we should disable this for release, the more I check this feature the more bug's I keep finding.

Campbell Barton removed their assignment 2014-05-26 04:07:32 +02:00

Lukas Treyer was assigned by Campbell Barton

2014-05-26 04:07:32 +02:00

Campbell Barton commented

2014-05-26 04:07:32 +02:00

Added subscriber: @ideasman42

Campbell Barton commented

2014-05-26 04:07:32 +02:00

Re-assigning

Campbell Barton commented

2014-05-26 04:10:30 +02:00

Note, we might be better off to store some info about the spline in BevPoint, having to calculate this after the fact is error prone.

Though Im not completely sure of this either, would have to try it and see if it introduces other problems.

Note, we might be better off to store some info about the spline in `BevPoint`, having to calculate this after the fact is error prone. Though Im not completely sure of this either, would have to try it and see if it introduces other problems.

Lukas Treyer commented

2014-05-26 07:40:38 +02:00

Sorry for urging you to test so much. I am new to C but I will get into the XCode Instruments to deliver a better performance in memory testing next time!
zero check for cu->resolu_ren is not necessary anymore since I get the segcount from nu. It also caused the spline to vanish completely. That might has to do with the array errors you discovered.
Cyclic splines: While I had the idea for this feature thinking about open curves, I think it makes sense for cyclic curves too. It basically always makes sense when you want to animate bevstart/end linearly but having segments not equal in length.

Sorry for urging you to test so much. I am new to C but I will get into the XCode Instruments to deliver a better performance in memory testing next time! zero check for cu->resolu_ren is not necessary anymore since I get the segcount from nu. It also caused the spline to vanish completely. That might has to do with the array errors you discovered. Cyclic splines: While I had the idea for this feature thinking about open curves, I think it makes sense for cyclic curves too. It basically always makes sense when you want to animate bevstart/end linearly but having segments not equal in length.

Lukas Treyer was unassigned by Campbell Barton

2014-05-26 13:34:24 +02:00

Campbell Barton self-assigned this 2014-05-26 13:34:24 +02:00

Campbell Barton commented

2014-05-26 13:34:24 +02:00

Ok. well I found mostly whats needed to fix this and we do a testbuild tomorrow, so Ill finish continue to finish off the fix.

blender-admin commented

2014-05-26 16:00:44 +02:00

This issue was referenced by 91a91b9362

This issue was referenced by 91a91b9362345c2a86543a42981cebc209527998

Campbell Barton commented

2014-05-26 16:02:09 +02:00

Changed status from 'Open' to: 'Resolved'

Campbell Barton closed this issue

2014-05-26 16:02:09 +02:00

Campbell Barton commented

2014-05-26 16:02:09 +02:00

Closed by commit 91a91b9362.

Closed by commit 91a91b9362.

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

bevel_factor_mapping_start/end in python loop #40223