--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOMQ%2Bw8n%2BHzV0qmjihQ5%2BMMCncpCkQ9cERFDdt6S1HStt22PUA%40mail.gmail.com.
I recently spent several days fixing backdrop-filter support for BlinkGenPropertyTrees, not realizing that backdrop-filter was only partially implemented and was not actively being worked on.I support option #2 but would like to expand on what it would look like for a new feature with WPT tests. I think the pattern would be to add WPT tests as usual, but mark them as failing by default. Developers of a cool new feature would then add third_party/WebKit/LayoutTests/virtual/enable-cool-feature which would run specific subdirectories of WPT tests with the in-development feature.
--On Thu, Sep 13, 2018 at 8:08 AM Chris Harrelson <chri...@chromium.org> wrote:Hi all,--When a RuntimeEnabledFeature is marked as "test" or "experimental", it is automatically turned on for unittests and layout tests.There have been several instances lately where this situation caused significant extra work for developers working on unrelated features, and reduced quality of testing. WebKit is also discussing the same issue.Here are some options for how to adjust the current situation:1. Never test what we do not ship.2. Put strict limits on the time a feature is allowed to be tested-but-not-shipped. Proposal: should ship in the same Chrome major version in which tests were flipped to test the feature. Any earlier testing should be done with techniques such as virtual layout test suites or RuntimeEnabledFlag override/TEST_P patterns in unittests.3. Separate "test" from "experimental". "experimental" would merely mean --enable-experimental-web-platform-features turns on the feature. "test" means on for testing. They can be turned on separately.I am leaning towards adopting #2, #3. Opinions?Examples:* In several instances I know of for just my team, a developer got confused about the actual launch state of features, and incorrectly treated an "experimental" feature as blocking the launch of a second feature (instead, it should have been a best-effort fix/discuss with feature owner). In some cases, the "experimental" feature was actually abandoned or on indefinite hold looking for staffing.* We are not testing the code configuration we ship. This can cause us to miss bugs. Not sure if I have seen an instance of this recently though.* Running content_shell with --run-web-tests mysteriously changes behavior due to turning on certain features, leading to lots of delay while debugging. This has bit my team at least twice in the last month or so.* There are a number of abandoned features in the "experimental" or "test" state, and turning them off takes work because it causes some tests to fail. (happened to me a few months ago)Chris
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOMQ%2Bw8n%2BHzV0qmjihQ5%2BMMCncpCkQ9cERFDdt6S1HStt22PUA%40mail.gmail.com.
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAJgFLLt59a-521Sv%2B0mNWK_FHUheVn4Bo1XUqp1mdNEDt0HC4Q%40mail.gmail.com.
Hi all,When a RuntimeEnabledFeature is marked as "test" or "experimental", it is automatically turned on for unittests and layout tests.There have been several instances lately where this situation caused significant extra work for developers working on unrelated features, and reduced quality of testing. WebKit is also discussing the same issue.Here are some options for how to adjust the current situation:1. Never test what we do not ship.
2. Put strict limits on the time a feature is allowed to be tested-but-not-shipped. Proposal: should ship in the same Chrome major version in which tests were flipped to test the feature. Any earlier testing should be done with techniques such as virtual layout test suites or RuntimeEnabledFlag override/TEST_P patterns in unittests.
3. Separate "test" from "experimental". "experimental" would merely mean --enable-experimental-web-platform-features turns on the feature. "test" means on for testing. They can be turned on separately.
--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAHOQ7J-cUsUbgRX%3DHACckRx9CLSj4gAk4xG6XZWP8m1jSrpv2Q%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOMQ%2Bw8n%2BHzV0qmjihQ5%2BMMCncpCkQ9cERFDdt6S1HStt22PUA%40mail.gmail.com.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAFpjS_04EW%3Dk3mc1cXz2G4R-6me9hW5hJPGr-1UXS6GN814hsQ%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/87a7okli01.fsf%40bud.servebeer.com.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAARdPYdmXCGGyzZ_bLHqn1KN%3DYqZ81DSPKGVahkzr1Pz9weJvQ%40mail.gmail.com.
I like option 2).I think we shouldn't make something "experimental" that isn't tested.
On Fri, Sep 14, 2018 at 5:18 PM Jochen Eisinger <joc...@chromium.org> wrote:I like option 2).I think we shouldn't make something "experimental" that isn't tested.I think Option 3 doesn't encourage it. Rather, experimental features should be surely tested in other ways; enabling it explicitly in each tests, or having virtual test suites.By this way, the configuration of *main* tests would match what we are shipping.
On Fri, Sep 14, 2018 at 10:34 AM Hayato Ito <hay...@chromium.org> wrote:On Fri, Sep 14, 2018 at 5:18 PM Jochen Eisinger <joc...@chromium.org> wrote:I like option 2).I think we shouldn't make something "experimental" that isn't tested.I think Option 3 doesn't encourage it. Rather, experimental features should be surely tested in other ways; enabling it explicitly in each tests, or having virtual test suites.By this way, the configuration of *main* tests would match what we are shipping.I think in practice, this will decrease the test coverage for experimental features, as we won't be able to duplicate all tests for all features, and we'll increase the risk of random things breaking once we try to ship the feature.For non-experimental features, we also get coverage from canary / dev, so I'm less worried about them, assuming that implementing 2) also ensured that features aren't experimental for ages.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CALjhuifs12vzQuPLpxV6yX0xZcSb2LxUk-B2kJJ7%2BDO3iSi6QA%40mail.gmail.com.To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.
On Fri, Sep 14, 2018 at 10:34 AM Hayato Ito <hay...@chromium.org> wrote:On Fri, Sep 14, 2018 at 5:18 PM Jochen Eisinger <joc...@chromium.org> wrote:I like option 2).I think we shouldn't make something "experimental" that isn't tested.I think Option 3 doesn't encourage it. Rather, experimental features should be surely tested in other ways; enabling it explicitly in each tests, or having virtual test suites.By this way, the configuration of *main* tests would match what we are shipping.I think in practice, this will decrease the test coverage for experimental features, as we won't be able to duplicate all tests for all features, and we'll increase the risk of random things breaking once we try to ship the feature.
On Thu, Sep 13, 2018 at 8:08 AM Chris Harrelson <chri...@chromium.org> wrote:Hi all,When a RuntimeEnabledFeature is marked as "test" or "experimental", it is automatically turned on for unittests and layout tests.There have been several instances lately where this situation caused significant extra work for developers working on unrelated features, and reduced quality of testing. WebKit is also discussing the same issue.Here are some options for how to adjust the current situation:1. Never test what we do not ship.Amen, I am also bothered by this, and have wasted time on it.2. Put strict limits on the time a feature is allowed to be tested-but-not-shipped. Proposal: should ship in the same Chrome major version in which tests were flipped to test the feature. Any earlier testing should be done with techniques such as virtual layout test suites or RuntimeEnabledFlag override/TEST_P patterns in unittests.I would like to also investigate whether there's something lighter-weight than a virtual test suite to accomplish this. Also, if we created a virtual test suite for *every* feature marked "test" or "experimental", it would greatly increase the number of tests we run. At the least, maybe we should have a virtual test suite that turns on all of the "experimental" features at once?
3. Separate "test" from "experimental". "experimental" would merely mean --enable-experimental-web-platform-features turns on the feature. "test" means on for testing. They can be turned on separately.I feel like one of these should imply the other, but I'm not sure which way it should go.
--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAHOQ7J-cUsUbgRX%3DHACckRx9CLSj4gAk4xG6XZWP8m1jSrpv2Q%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/860e08ff-a68a-4af0-9b22-8dc9fa546e6b%40chromium.org.