Bug 91017 - arb_framebuffer_no_attachments.arb_framebuffer_no_attachments-atomic fails intermittently
Summary: arb_framebuffer_no_attachments.arb_framebuffer_no_attachments-atomic fails in...
Status: NEW
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: git
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Ian Romanick
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-06-18 03:53 UTC by Mark Janes
Modified: 2016-11-02 05:31 UTC (History)
4 users (show)

See Also:
i915 platform:
i915 features:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Mark Janes 2015-06-18 03:53:37 UTC
spec.arb_framebuffer_no_attachments.arb_framebuffer_no_attachments-atomic fails on byt and bsw, beginning with mesa 83199998310591b9162ab12e922ed79ee235b5c8

Standard Output

/tmp/build_root/m32/lib/piglit/bin/arb_framebuffer_no_attachments-atomic -auto -fbo

Standard Error

compare_counter (subtest glViewport): expected 4, got 0
Comment 1 Kevin Rogovin 2015-06-18 11:41:15 UTC
This is NOT a framebuffer_no_attachment bug. 

Changing the test to NOT bind a framebuffer, essentially commenting out from tests/spec/arb_framebuffer_no_attachments/atomic.c the block that binds an attachmentless FBO: 


	glBindFramebuffer(GL_FRAMEBUFFER, fbo);

	/* Setup default width and height. */
	glFramebufferParameteri(GL_FRAMEBUFFER,
		GL_FRAMEBUFFER_DEFAULT_WIDTH, piglit_width);
	glFramebufferParameteri(GL_FRAMEBUFFER,
		GL_FRAMEBUFFER_DEFAULT_HEIGHT, piglit_height);

	/* Check that fbo is marked complete. */
	if (glCheckFramebufferStatus(GL_FRAMEBUFFER) !=
		GL_FRAMEBUFFER_COMPLETE)
		piglit_report_result(PIGLIT_FAIL);

	prog = piglit_build_simple_program(vs_source, fs_source);

	/* Check that there are no errors. */
	if (!prog || !piglit_check_gl_error(GL_NO_ERROR))
		piglit_report_result(PIGLIT_FAIL);

produces the exact same error, i.e. this error is independent of FBO no attachments and for that matter FBO's.

The bug source is likely to be a cache flush issue that manifests on atoms() only. Changing the reset_count() implementation to instead use glBufferData (creating a new BO) makes the test pass.
Comment 2 Kevin Rogovin 2015-06-18 11:58:59 UTC
atoms() meaning atoms(byt, bsw)
Comment 3 Kevin Rogovin 2015-06-18 12:01:15 UTC
Arr.. and the block to comment out is exactly this:

         glBindFramebuffer(GL_FRAMEBUFFER, fbo);

	/* Setup default width and height. */
	glFramebufferParameteri(GL_FRAMEBUFFER,
		GL_FRAMEBUFFER_DEFAULT_WIDTH, piglit_width);
	glFramebufferParameteri(GL_FRAMEBUFFER,
		GL_FRAMEBUFFER_DEFAULT_HEIGHT, piglit_height);

	/* Check that fbo is marked complete. */
	if (glCheckFramebufferStatus(GL_FRAMEBUFFER) !=
		GL_FRAMEBUFFER_COMPLETE)
		piglit_report_result(PIGLIT_FAIL);

(For whatever reason I included in the past the next 4 lines).

 -Kevin
Comment 4 Mark Janes 2015-06-19 19:53:46 UTC
This bug is intermittent.  I am disabling the test on the mesa-ci jenkins system.
Comment 5 Mark Janes 2015-06-19 19:58:11 UTC
This test failure reproduced on BDWGT2 during this morning's test run.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.