aboutsummaryrefslogtreecommitdiffstats
path: root/tests/auto/declarative/qmlvisual/TEST_GUIDELINES
diff options
context:
space:
mode:
Diffstat (limited to 'tests/auto/declarative/qmlvisual/TEST_GUIDELINES')
-rw-r--r--tests/auto/declarative/qmlvisual/TEST_GUIDELINES7
1 files changed, 7 insertions, 0 deletions
diff --git a/tests/auto/declarative/qmlvisual/TEST_GUIDELINES b/tests/auto/declarative/qmlvisual/TEST_GUIDELINES
new file mode 100644
index 0000000000..469832f501
--- /dev/null
+++ b/tests/auto/declarative/qmlvisual/TEST_GUIDELINES
@@ -0,0 +1,7 @@
+Guidelines for creating new visual tests:
+
+1. Keep it small. All visual tests should be able to run on a device with a screen of at least 640x360 pixels. Smaller than that is even better, because images of this side need to be processed and saved for every test (and even committed to the repository).
+
+2. Keep it short. It is hoped that these tests can be run regularly, perhaps even for every commit, and if you add up ten seconds for every time someone commits a change to QML then we'll be sitting here for a long time. Completeness is more important than haste, but consider the most time efficient ways to achieve said completeness. Do not forget about snapshot mode (tst_qmlvisual -help for details on -recordsnapshot) when testing that a static scene looks right.
+
+3. Avoid text. Text is relatively unstable due to platform specific peculiarities. If you need to identify an area, consider a unique color as opposed to a unique text label. If you must use Text, TextEdit, or TextInput, use the test-friendlier versions in the 'shared' directory. Also keep in mind that text anti-aliasing is disabled during tests for greater consistency, and you should never use point sizes in tests. Text autotests can thus only test some aspects of the elements.