The reftests for bug 379349 all seem to fail at certain logical resolutions. If I set layout.css.dpi to 96 or 120, they pass, but if I set it to 110 they fail. (In particular, at 110dpi, both the tests and the references seem broken, but the references change differently from the tests: in both cases the borders of the columns show column widths as though there were only two columns, but in the reference the contents of the columns are laid out as though the widths were correct (whereas in the tests the contents match the borders).)
You can put this into reftest-analyzer.xhtml (etc.) to see the failures. I'm actually not sure what DPI this was at; it's on a VNC session to my desktop (where xdpyinfo says 100dpi), post-bug 384090.
I'm not sure if this is a bug in the code or the tests, but I sort of suspect code based on the odd behavior.
Seems to be a rounding problem. The column and container sizes here are all in points rather than pixels. Or at least that's one of the problems.
While we should fix the rounding problems --- and have reftests for this bug --- the column tests should use px rather than pt so that they don't test more than they need to.
No comments here in a long time. IS this bug still valid?
You need to log in before you can comment on or make changes to this bug.