398157 - "ASSERTION: didn't subtract all that we added" with <select>, <div style="margin: 0 100%;">

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: testcase

This testcase triggers a bunch of assertions, including one I haven't seen elsewhere recently:

###!!! ASSERTION: didn't subtract all that we added: 'space == 0 && ((l2t == FLEX_PCT_LARGE) ? (-0.001f < basis.f && basis.f < 0.001f) : (basis.c == 0 || basis.c == nscoord_MAX))', file /Users/jruderman/trunk/mozilla/layout/tables/BasicTableLayoutStrategy.cpp, line 1013

Note that this isn't the same assertion as in bug 398043.  That one is in BasicTableLayoutStrategy::ComputeColumnIntrinsicWidths; this one is in BasicTableLayoutStrategy::ComputeColumnWidths.

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: fix

Trivial fix.

Issue arises when "width" is nscoord_MAX in ComputeColumnWidths.  We'd been subtracting directly off of "width", and we need to use NSCoordSaturatingSubtract instead.

Comment 2

[Daniel Holbert [:dholbert]]

Fix checked in.

Checking in BasicTableLayoutStrategy.cpp;
/cvsroot/mozilla/layout/tables/BasicTableLayoutStrategy.cpp,v  <--  BasicTableLayoutStrategy.cpp
new revision: 3.260; previous revision: 3.259
done

Comment 3

[Marcia Knous [:marcia]]

This looks okay on Win XP using Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a9pre) Gecko/2007100204 Minefield/3.0a9pre. However, using the same build on mac paralyzes my entire system.  On one Intel test machine I had to shut down the machine to get out of the cycle. juanb had the same thing happen to him.

Comment 4

[Daniel Holbert [:dholbert]]

Don't think the hang is caused by this bug's fix... stephend tried this Mac build:

  Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a9pre) Gecko/2007100120 Minefield/3.0a9pre

(after the fix was checked in) and it didn't hang.

Comment 5

[Carsten Book [:Tomcat]]

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a9pre) Gecko/2007100204 Minefield/3.0a9pre -> paralyzes my entire system on mac.

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a9pre) Gecko/2007100104 Minefield/3.0a9pre -> is working fine

so this patch caused a regression on mac.

Comment 6

[Carsten Book [:Tomcat]]

did some more testings and this seems to be no regression, the hang is also reproducible in 2007100104 and Gran Paradiso Alpha 8 Builds when you use this testcase and click on the "bar" in this testcase.

Filed Bug 398336 for this problem.

Comment 7

[Jesse Ruderman]

Crashtest checked in.