Supporting the instance tree requires a lot of complexity and it can probably be removed without affecting any users. The instance tree was ahead of it's time (many of the same issues are on public-webapps right now!) but it never caught on with users. Chromium's usecounters found no usage of instanceRoot and it was removed entirely in mid-2014[1]. I'm not aware of any web compatibility fallout from removing it (not a single bug, AFAIK). There is support in the SVGWG for removing the instance tree in SVG2[2]. The instance tree was never implemented in Gecko. Removing the instance tree in blink took a few patches but most were mechanical. All the patches are listed on crbug.com/313438 and would be pretty easy to port to WebKit. [1] https://groups.google.com/a/chromium.org/d/msg/blink-dev/mgYCJuzfU0Q/8jdhjY0pNgoJ [2] https://lists.w3.org/Archives/Public/www-svg/2014Jan/0014.html
Sounds good to me.
I'm fine with this as well.
I’m going to take a crack at this.
Investigating crbug.com/313438 further, it’s clear this is a pretty big project that Rob Buis did! But we can indeed take advantage of those patches and redo this for WebKit; it will be 10 or so patches to do it all. I am about to land a first patch.
Created attachment 246251 [details] Patch
Committed r179810: <http://trac.webkit.org/changeset/179810>
This may have caused a perf regression on the HTML5 full-render test: - https://perf.webkit.org/#mode=charts&chartList=%5B%5B%22mac-mavericks%22%2C%22Parser%2Fhtml5-full-render%3ATime%22%5D%5D (Mavericks / 3.6% regression) - https://perf.webkit.org/#mode=charts&chartList=%5B%5B%22efl%22%2C%22Parser%2Fhtml5-full-render%3ATime%22%5D%5D (EFL / 2.3% regression) I am not 100% sure it is caused by this change but the range on the EFL bot is small enough that I'd say it is very likely.
(In reply to comment #7) > This may have caused a perf regression on the HTML5 full-render test: > - > https://perf.webkit.org/#mode=charts&chartList=%5B%5B%22mac- > mavericks%22%2C%22Parser%2Fhtml5-full-render%3ATime%22%5D%5D (Mavericks / > 3.6% regression) > - > https://perf.webkit.org/ > #mode=charts&chartList=%5B%5B%22efl%22%2C%22Parser%2Fhtml5-full- > render%3ATime%22%5D%5D (EFL / 2.3% regression) > > I am not 100% sure it is caused by this change but the range on the EFL bot > is small enough that I'd say it is very likely. Looking at the change though, I really don't see what could have caused a regression. I think Ryosuke is planning to confirm the regression locally (may be a flake on the bot or another commit).
Are there any SVG <use> elements in the full-render test?
*** Bug 121050 has been marked as a duplicate of this bug. ***