1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
|
// Copyright 2020 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef UI_BASE_TEST_SKIA_GOLD_MATCHING_ALGORITHM_H_
#define UI_BASE_TEST_SKIA_GOLD_MATCHING_ALGORITHM_H_
#include <string>
namespace base {
class CommandLine;
}
namespace ui {
namespace test {
// Describes what algorithm to use to match golden images.
// Fuzzy and Sobel algorithms have adjustable parameters:
//
// Fuzzy has the max number of different pixels, the max per-channel delta sum
// (i.e. how much a pixel can differ by and still be considered the same), and
// how many border pixels to ignore. The number of ignored border pixels is
// mainly to make the Sobel filter work better, as if an edge goes all the way
// to the border of the image, the pixels along the border won't be blacked
// out due to the Sobel filter.
//
// Sobel has the parameters of fuzzy + the edge threshold, which determines how
// much of the image should be blacked out based on the generated Sobel filter.
//
// In general, it works like this:
// 1. Try an exact comparison check against all approved images for the test,
// returning success if a hash matches.
// 2. Get the most recent image for the trace (test + any reported JSON keys,
// e.g. OS/hardware information)
// 3. If the algorithm is Sobel, generate a Sobel filter and black out parts
// of the image based on the provided edge threshold.
// 4. Do a fuzzy comparison of the images.
//
// To determine the suitable parameter values using historical data for a test:
// https://cs.chromium.org/chromium/src/content/test/gpu/gold_inexact_matching/determine_gold_inexact_parameters.py
class SkiaGoldMatchingAlgorithm {
public:
SkiaGoldMatchingAlgorithm();
virtual ~SkiaGoldMatchingAlgorithm();
// Append the algorithm parameter to |cmd|.
virtual void AppendAlgorithmToCmdline(base::CommandLine& cmd) const;
// The algorithm name for commandline.
virtual std::string GetCommandLineSwitchName() const = 0;
};
class ExactSkiaGoldMatchingAlgorithm : public SkiaGoldMatchingAlgorithm {
public:
ExactSkiaGoldMatchingAlgorithm();
~ExactSkiaGoldMatchingAlgorithm() override;
void AppendAlgorithmToCmdline(base::CommandLine& cmd) const override;
protected:
std::string GetCommandLineSwitchName() const override;
};
class FuzzySkiaGoldMatchingAlgorithm : public SkiaGoldMatchingAlgorithm {
public:
FuzzySkiaGoldMatchingAlgorithm(int max_different_pixels,
int pixel_delta_threshold,
int ignored_border_thickness = 0);
~FuzzySkiaGoldMatchingAlgorithm() override;
void AppendAlgorithmToCmdline(base::CommandLine& cmd) const override;
protected:
std::string GetCommandLineSwitchName() const override;
private:
int max_different_pixels_{0};
int pixel_delta_threshold_{0};
int ignored_border_thickness_{0};
};
class SobelSkiaGoldMatchingAlgorithm : public FuzzySkiaGoldMatchingAlgorithm {
public:
SobelSkiaGoldMatchingAlgorithm(int max_different_pixels,
int pixel_delta_threshold,
int edge_threshold,
int ignored_border_thickness = 0);
~SobelSkiaGoldMatchingAlgorithm() override;
void AppendAlgorithmToCmdline(base::CommandLine& cmd) const override;
protected:
std::string GetCommandLineSwitchName() const override;
private:
int edge_threshold_{0};
};
// With this algorithm, an image is regarded to match a golden image when:
// 1. the image is an exact match, and
// 2. the set of golden images for the test must only contain a single image.
// This algorithm is used in Ash pixel testing to facilitate the gold image
// revision update.
// The main difference between this and `ExactSkiaGoldMatchingAlgorithm` is:
// * ExactSkiaGoldMatchingAlgorithm
// Your test can match multiple valid Gold images. For a test run, if the
// pixel output is one of the Gold images, we consider it's valid and test
// would pass.
// Any committers can approve a new Gold image. The new Gold image would be
// added to the valid Gold images set.
//
// * PositiveIfOnlyImageAlgorithm
// There is only 1 valid Gold image. No one can approve a new Gold image if
// there is already a Gold image. In order to updating the Gold image, you
// need to add a revision to the screenshot name, so it's treated as a new
// test from Gold's perspective.
//
// Say you have a button pixel test with screenshot name `pwd_manager_button1`.
// With ExactSkiaGoldMatchingAlgorithm:
// When a cl updates all button styles, the cl owner may approve all new
// styles via Skia Gold, gets a button style owner to approve it, and merges
// the change.
// In this case, the feature owner(pwd manager UI owner) would not get
// notified. If you as feature owner don't want this behavior, you can use
// PositiveIfOnlyImageAlgorithm:
// Your screenshot name would be `pwd_manager_button1.rev0`. When a cl updates
// all button styles, the cl owner need to edit your test case, change the
// screenshot name to `pwd_manager_button1.rev1`, and seek feature owner's
// approval for the test case change. In this way, feature owner would know
// the UI changes.
//
// This algorithm would increase the Gold image maintenance cost. Toolkit owners
// who regularly update styles might be hindered as they need to edit all test
// cases. So use your best judgement for the algorithm you choose.
class PositiveIfOnlyImageAlgorithm : public SkiaGoldMatchingAlgorithm {
public:
PositiveIfOnlyImageAlgorithm();
~PositiveIfOnlyImageAlgorithm() override;
private:
// SkiaGoldMatchingAlgorithm:
void AppendAlgorithmToCmdline(base::CommandLine& cmd) const override;
std::string GetCommandLineSwitchName() const override;
};
} // namespace test
} // namespace ui
#endif // UI_BASE_TEST_SKIA_GOLD_MATCHING_ALGORITHM_H_
|
