grid-abspos-items-015.html is a little too wide, so each of its lines have a linewrap, which makes it too tall for the reftest snapshot to capture
Categories
(Core :: Layout: Grid, task)
Tracking
()
Tracking | Status | |
---|---|---|
firefox102 | --- | fixed |
People
(Reporter: dholbert, Assigned: dholbert)
Details
Attachments
(2 files)
In looking at a recent Try failure log [1] for reftest grid-abspos-items-015.html, I noticed that it's organized into lines of floated grid elements -- but it's just a little bit too wide to fit in the reftest window, so the last element on each line ends up wrapping to a new line.
So each "line" ends up being 2 lines tall, which makes the whole test twice as tall as it intends to be. And the bottom portion doesn't get tested as a result, since it's offscreen.
It looks like the test content is about 830px wide on my system, and I think maybe 800px is the size of the reftest viewport. So we could rewrite the test to be a bit smaller, or we could split it into two subtests if that's simpler.
[1] This will expire at some point, but here's the Try log I was looking at:
https://treeherder.mozilla.org/logviewer?job_id=377605909&repo=try&lineNumber=6286
Assignee | ||
Comment 1•2 years ago
|
||
Assignee | ||
Updated•2 years ago
|
Assignee | ||
Comment 2•2 years ago
|
||
Before this change, each "line" has 6 floats, each of which has a border-box
that's 138px wide. So, each line has a total width of 828px.
Unfortunately this is too wide for the 800x1000 reftest viewport (as described
in reftest.jsm), so each line wraps a single float. This looks awkward, but
worse, it means the test as a whole ends up being twice as tall as we intend
for it to be, and some of the content gets pushed off the bottom of the
viewport and doesn't get compared in the reftest harness.
After this change, each float's border-box is 8px skinnier (130px wide), which
puts each line at 780px, which is skinny enough to fit without wrapping.
I'm also changing various hardcoded coordinates in the reference case by 8px as
well, to keep the testcase passing.
I've also made a 1px change to the reference case's 'bottom' declaration in the
very last line, to fix a preexisting bug that gets uncovered by this fix. This
value has been off by 1px, but it wasn't causing a teset failure because it was
outside of the viewport. Now it's inside the viewport, so we can see that it
was wrong and we can fix it here.)
Updated•2 years ago
|
Pushed by dholbert@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/8db850bac657 Reduce sizes in grid-abspos-items-015.html (and -016 and their reference case) to fit in reftest viewport. r=dshin
Description
•