359510 - [reflow branch] Chats table too wide if exist contact with long name

Status: VERIFIED FIXED

(Core :: Layout: Tables, defect)

Description

[Oleg Romashin (:romaxa)]

Open gmail.com with chats support.

Add any contact with long name.

Table with contact list has width ~ >= size of long name.

See screenshoot

Comment 1

[Oleg Romashin (:romaxa)]

Attachment: Screenshot of with bad rendered table.

Sorry, but simple test creating, cause some problems...

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

So the way this worked in pre-reflow-branch Mozilla was by making containers with 'overflow' not 'visible' have a min-width of 0, regardless of their contents.  This would be a simple patch to nsLayoutUtils::IntrinsicForContainer (check aFrame->GetType against nsLayoutAtoms::scrollFrame in addition to checking for IsFrameOfType(eReplaced)).

However, the way it works in WinIE is completely different, and I think preferable -- percentage widths inside a table cell with an explicit width are based on the explicit width of the cell rather than the actual width of the cell, when the size of the cell is increased by other columns or by that column.

That said, given that we precompute table widths now and don't shrink them down at the end, I'm not sure that will actually get us the desired result...

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(In reply to comment #2)
> So the way this worked in pre-reflow-branch Mozilla was by making containers
> with 'overflow' not 'visible' have a min-width of 0, regardless of their
> contents.  This would be a simple patch to nsLayoutUtils::IntrinsicForContainer
> (check aFrame->GetType against nsLayoutAtoms::scrollFrame in addition to
> checking for IsFrameOfType(eReplaced)).

Note that pre-reflow-branch Mozilla did this for all scroll frames, not just those with percentage widths, which is what I was proposing here...

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Actually, on second thought, I really don't like the WinIE width behavior, and nobody else does that.

Old Mozilla and Opera always give things with 'overflow' non-'visible' a min width of 0.  I'm considering either doing that unconditionally or only when they have percentage widths.  I suppose I'm better off keeping compatibility with what we've been doing rather than changing it, so I'm leaning towards doing it unconditionally.

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

OK, fixed on reflow branch by giving scroll frames a min width of 0, unconditionally.

Comment 6

[Oleg Romashin (:romaxa)]

Verified