1609805 - Support a new reftest kind, for verifying rasterizer accuracy with different tile sizes

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, enhancement)

Comment 1

[Glenn Watson [:gw]]

Attachment: Bug 1609805 - Support a new reftest kind, for verifying rasterizer accuracy.

This patch introduces a new reftest (syntax ** or !* in reftest files).

This type of test renders a single input file multiple times, at a range
of different picture cache tile sizes. It then verifies that each of the
images matches (or doesn't).

This can be used to verify rasterizer accuracy when drawing primitives
with different tile sizes and/or dirty rect update strategies.

One of the included tests in this patch fails the accuracy test - the
intent is to fix this inaccuracy in a follow up patch and then be able to
mark it pixel exact.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/918e156c75aa
Support a new reftest kind, for verifying rasterizer accuracy. r=nical,Bert

Comment 3

[Andrei Ciure[:aciure]]

https://hg.mozilla.org/mozilla-central/rev/918e156c75aa

Comment 4

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

There is a permanent Wrench failure since this landed, see bug 1610251. That bug is about turning the task's exit status from success to fail if it encounters a failure. Can you fix the reported failure with a follow-up commit or in a new bug?

Error message:

Parsed log: https://treeherder.mozilla.org/logviewer.html#?job_id=285556958&repo=autoland
Full log: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/QK1niWDET7iPC3afOzarag/runs/0/artifacts/public/logs/live_backing.log
Reftest URL: https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/QK1niWDET7iPC3afOzarag/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1

Either the task should exit as failed (so Treeherder will show it as failed) or shouldn't print TEST-UNEXPECTED-FAIL (which can confuse people looking at a run which actually failed, because of a different reason.

[task 2020-01-19T21:43:12.994Z] REFTEST reftests\transforms\big-axis-aligned-scale.yaml == reftests\transforms\big-axis-aligned-scale-ref.yaml
[task 2020-01-19T21:43:13.095Z] REFTEST ** reftests\tiles\rect.yaml
[task 2020-01-19T21:43:13.301Z] REFTEST ** reftests\tiles\simple-gradient.yaml
[task 2020-01-19T21:43:13.609Z] REFTEST !* reftests\tiles\prim-suite.yaml
[task 2020-01-19T21:43:13.881Z] REFTEST TEST-UNEXPECTED-FAIL | !* reftests\tiles\prim-suite.yaml | image comparison
[task 2020-01-19T21:43:13.881Z] REFTEST TEST-END | !* reftests\tiles\prim-suite.yaml
[task 2020-01-19T21:43:13.883Z] REFTEST INFO | 274 passing, 0 failing
[task 2020-01-19T21:43:13.956Z] WARN: rx::HLSLCompiler::compileToBinary(259):
[task 2020-01-19T21:43:13.956Z] C:\fakepath(1983,18-133): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(1989,62-116): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2006,84-93): warning X3556: integer divides may be much slower, try using uints if possible.
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2009,50-59): warning X3556: integer modulus may be much slower, try using uints if possible.
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2029,33-90): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(1823,1-6): warning X4000: use of potentially uninitialized variable (dyn_index_vec4_float4)
[task 2020-01-19T21:43:13.956Z]
[task 2020-01-19T21:43:13.956Z] WARN: rx::HLSLCompiler::compileToBinary(259):
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2049,62-116): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2043,18-133): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2072,87-96): warning X3556: integer divides may be much slower, try using uints if possible.
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2075,50-59): warning X3556: integer modulus may be much slower, try using uints if possible.
[task 2020-01-19T21:43:13.956Z] C:\fakepath(2091,33-90): warning X3571: pow(f, e) will not work for negative f, use abs(f) or conditionally handle negative values if you expect them
[task 2020-01-19T21:43:13.956Z] C:\fakepath(1748,1-6): warning X4000: use of potentially uninitialized variable (dyn_index_vec4_float4)
[task 2020-01-19T21:43:13.956Z]
[task 2020-01-19T21:43:14.049Z]
[task 2020-01-19T21:43:14.049Z] z:\task_1579466173\build\src\gfx\wr\wrench>if 0 NEQ 0 EXIT /b 0
[task 2020-01-19T21:43:14.049Z]
[task 2020-01-19T21:43:14.049Z] z:\task_1579466173\build\src\gfx\wr\wrench>popd
[task 2020-01-19T21:43:14.050Z]
[task 2020-01-19T21:43:14.050Z] z:\task_1579466173\build\src\gfx\wr>pushd examples
[task 2020-01-19T21:43:14.051Z]
[task 2020-01-19T21:43:14.051Z] z:\task_1579466173\build\src\gfx\wr\examples>cargo check --verbose

Comment 5

[Glenn Watson [:gw]]

Sure, I'll create a patch to fix this today.