There's an incorrectly placed PR_MAX in the code for pref width distribution of colspanning cells. The old code didn't really make much sense. I found this while writing testcases related to bug 363144 and bug 363874.

I'm not sure why I thought the old logic here was correct. Since the span-pref and span-min widths on the columns are accumulated as *additions* to the min and pref (since it's easier to divide them up among the columns that way), I should have the PR_MAX outside the additions. This is compatible with WinIE7 -- not sure why I thought the other way was -- or if I'd considered it.

Note that the above patch is a diff against a tree to which the patch for bug 363329 has been applied. That's why it's not a cvs diff.

Comment on attachment 249801 [details] [diff] [review] patch + +#ifdef DEBUG_dbaron_off + printf("table %p col %d span: min=%d pref=%d spec=%d pct=%f\n", + mTableFrame, col, colFrame->GetMinCoord(), + colFrame->GetPrefCoord(), colFrame->GetHasSpecifiedCoord(), + colFrame->GetPrefPercent()); +#endif Please move this code to the place where you deleted it first: http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&root=/cvsroot&subdir=mozilla/layout/tables&command=DIFF&root=/cvsroot&file=nsTableColFrame.cpp&rev1=3.90&rev2=3.91#7 and remove the DEBUG_dbaron_off there. I just forgot in my previous patch the following lines } + printf("\nm:%d c:%d p:%d sm:%d sc:%d sp:%d f:%d", GetMinCoord(), GetPrefCoord(), + GetPrefPercent(), GetSpanMinCoord(), GetSpanPrefCoord(), + GetSpanPrefPercent(), GetFinalWidth()); printf("\n%s**END COL DUMP** ", indent); with that change r+

Comment on attachment 249801 [details] [diff] [review] patch Makes sense. sr=bzbarsky. Sorry for the lag... :(

I really want this simple debugging code -- the other stuff spews out tons of information I don't need and is much harder to use.

(Or, at least, it's unnecessarily verbose and spread over too many lines.)

Fix checked in to trunk, 2007-01-12 17:08 PST (-0800, America/Los_Angeles).