You are setting all auto margins to zero on relatively positioned elements and then setting margin-right back to auto to deal with the overconstraning that results. See the above URL.

Troy - I would think this bug should probably be yours, since you've been dealing with this stuff. You've already fixed most of the problems on this page. I think my analysis above is still correct, except it now only applies to one test (class=seven). Or something like that... I just fixed up the above URL to deal with the CSS2 errata.

Giving the bug back to Kipp. nsHTMLReflowState is (now) calculating the margin-left correctly, but because the block/inline code doesn't use the calculated margins in the nsHTMLReflowState struct we're still getting the wrong value

Setting all current Open/Normal to M4.

The test works correctly now.

chrisd as QA contact

Verified fixed.

Test seven on the above URL ( http://www.fas.harvard.edu/~dbaron/csstest/sec100303 ) regressed between the builds of 1999-03-26 and 1999-03-29. The state of this bug seems the same as it was when I made the 1999-01-25 comment above.

Fixed. The code to force the equality in section 10.3.3 of the spec was only applied when *extra* space was available...

Using 5/17 build, verified fixed.