[Groff] Alas, I've hit a troff wall

[Eric S. Raymond]

I've hit a wall on the block column allocation thing.  I'm pretty sure
my C++ for counting block-containing columns is correct and that the general
theory of the emitted troff is sound (Werner thought so too
when I described it).  I've checked in logic that will enable anyone
to try out the new rule under the global flag 'experimental'.

But it doesn't work.  It leads to block columns widths that are too large.
The problem almost has to be in the troff generated by one of these 
two lines:

table.cpp:691 -- the block column width computation itself
 
table.cpp:2142 -- computation of available line width once non-block 
                  columns have been laid out.

My suspicions now focus on line 2142, but staring at it hasn't been 
getting me anywhere.  I'm getting stale and frustrated; I need to go work
on something else for a while.

I've enclosed the test load I use. To see the behavor under the existing
default, remove the 'experimental' flags from the table headers.

-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>

TEST
Description: Text document