Fx3 changed the behaviour to render automatic layout tables when Bug 402567 was fixed, to be consistent with W3C suggestions (see URL) and with IE and Safari behaviours. Now min-width of "table-layout:auto" cells is equal to the width of their contents. This behaviour anyway causes problems with 'overflow' property, that can't be any more applied to not-fixed tables and their contents. My proposal is to not set the min-width equal to the content width if overflow property is set to a value other than 'visible'. This shouldn't cause a regression for Bug 402567 and it doesn't break W3C specs. See also David Baron proposal in Bug 433621 PS: this bug is an 'evolution' of my idea in Bug 409736 Comment #16 and 18.

(In reply to comment #0) > My proposal is to not set the min-width equal to the content width if overflow > property is set to a value other than 'visible'. This shouldn't cause a > regression for Bug 402567 and it doesn't break W3C specs. If I'm understanding correctly, this behavior is precisely what Bug 402567 changed. Your proposal sounds basically like "undo the patch for Bug 402567." See this chunk of bug 402567 comment 23, with the attachment of the first patch: > Created an attachment (id=289541) [details] > fix v1 > > This patch makes nsHTMLScrollFrame::GetMinWidth() return the min-width of the > scrolled frame. (matching IE / Konqueror, as described in my prev. comment.) (Note: nsHTMLScrollFrame = the wrapper frame that's generated for any element with overflow != visible) If that's not what you're proposing, can you be a bit more specific?

Emh. I must say I didn't understood that the min-width behaviour was changed only for overflow != visible. I thought it was changed independently by overflow value. I think I'll return to the idea in Bug 409736 Comment #16 when I have a more clear head. You can WONTFIX it and sorry for the mess.

(In reply to comment #2) > Emh. I must say I didn't understood that the min-width behaviour was changed > only for overflow != visible. I thought it was changed independently by > overflow value. Yup -- this was changed specifically for scrollframes, i.e. for frames with overflow != visible. Sorry if that was unclear in earlier discussions. > I think I'll return to the idea in Bug 409736 Comment #16 when I have a more > clear head. You can WONTFIX it and sorry for the mess. K. WONTFIXing.