In r221670 we changed the way we get the metrics to avoid rounding that was causing a negative line gap to be computed. Since the font metrics value is indeed a float value, we also removed the rounding when setting the line height. However, this caused some test failures because now we report non integer line heights, see for example: https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r221719%20(3018)/fast/css/font-calculated-value-pretty-diff.html https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r221719%20(3018)/fast/ruby/ruby-line-height-pretty-diff.html I think it would be better to bring back the rounding for consistency with all other ports, but also with other browsers. This will require another rebaseline, but a lot smaller :-)
Created attachment 320103 [details] Patch
Committed r221724: <http://trac.webkit.org/changeset/221724>
Tom, did this reintroduce the bug with Evolution? Is it OK?
(In reply to Michael Catanzaro from comment #3) > Tom, did this reintroduce the bug with Evolution? Is it OK? I don't know about Evolution (I don't work on it, nor use it anymore), but the testcase that made for bug 149703 was running fine when I tested it yesterday. It was running fine with and without this change. I will probably add it as a layout test, so we are sure, that it works in the future.