Bug 137689 - gl.detachShader breaks shader program
Summary: gl.detachShader breaks shader program
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: WebGL (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified OS X 10.9
: P2 Normal
Assignee: Dean Jackson
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2014-10-14 04:50 PDT by huscar
Modified: 2017-09-11 22:55 PDT (History)
10 users (show)

See Also:


Attachments
Testcase (2.79 KB, text/html)
2017-08-23 18:15 PDT, Dean Jackson
no flags Details
Testcase (2.66 KB, text/html)
2017-09-07 17:57 PDT, Dean Jackson
no flags Details
Patch (14.96 KB, patch)
2017-09-08 19:05 PDT, Dean Jackson
sam: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description huscar 2014-10-14 04:50:14 PDT
Calling gl.detachShader after linking results in a program that doesn't draw anything (without error messages to the console). AFAIK this should not happen and using detachShader and deleteShader is encouraged to reduce memory usage.
Here's a good test from when firefox had similar problems https://bugzilla.mozilla.org/show_bug.cgi?id=867253
Comment 1 Radar WebKit Bug Importer 2017-08-22 16:57:50 PDT
<rdar://problem/34025056>
Comment 2 Dean Jackson 2017-08-23 18:15:28 PDT
I am unable to reproduce this. See the attached test case which both detaches and deletes the shaders after linking the program, but before they are used in a draw call (and before they are used to get attribute and uniform locations).
Comment 3 Dean Jackson 2017-08-23 18:15:51 PDT
Created attachment 318952 [details]
Testcase
Comment 4 Dean Jackson 2017-08-23 18:17:38 PDT
Also, the original demo on flohofwoe doesn't seem to work any more, in any browser. It also downloads rather than running directly.
Comment 5 Dean Jackson 2017-09-07 15:20:07 PDT
Oops. My test case is bad. This is really broken.
Comment 6 Dean Jackson 2017-09-07 17:57:14 PDT
Created attachment 320217 [details]
Testcase

Better testcase. The triangle should be red.
Comment 7 Dean Jackson 2017-09-08 19:05:03 PDT
Created attachment 320329 [details]
Patch
Comment 8 Sam Weinig 2017-09-08 20:20:05 PDT
Comment on attachment 320329 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=320329&action=review

> Source/WebCore/ChangeLog:20
> +        This fixes the bug, but the whole area is still a bit messy. For one,
> +        we're keeping around all the shader information even after it is
> +        no longer used.

Seems like we should have a bug tracking this issue.

> Source/WebCore/platform/graphics/opengl/GraphicsContext3DOpenGLCommon.cpp:996
> +    LinkedShaderMap::iterator result = m_linkedShaderMap.find(program);

Use auto here?

> Source/WebCore/platform/graphics/opengl/GraphicsContext3DOpenGLCommon.cpp:1053
> +    LinkedShaderMap::iterator result = m_linkedShaderMap.find(program);

auto?
Comment 9 Dean Jackson 2017-09-09 15:52:52 PDT
(In reply to Sam Weinig from comment #8)
> 
> Seems like we should have a bug tracking this issue.
> 

https://bugs.webkit.org/show_bug.cgi?id=98204
Comment 10 Dean Jackson 2017-09-09 15:55:00 PDT
Committed r221831: <http://trac.webkit.org/changeset/221831>