mailing-list for TeXmacs Users

[David Allouche <address@hidden>]

I just isolated a very annoying typesetter bug which kept screwing up a 
document I am working on. Now, I would have posted it to the TeXmacs 
Resources WikiWeb, but the antalia server seems to be down at the moment (I 
get a 502, Proxy Error).

I attached all the documents I talk of here in one archive.

The bug
-------

To reproduce the bug you need to have a document with at least three floats. 
I used floats appended at the end of each paragraph. In floatup1.tm, I have 
put enough text in each paragraph so that each float is placed at its 'here' 
location. If the amount of text in a paragraph is too small, the floats may 
be grouped togheter. For the first example I want each float to be separated 
from the other.

All the floats after the second one are forced on the top of the page, even 
if they are allowed to be placed anywhere except at 'top'.

Moreover, such floated-up floats, if they are *not* the last float in the 
flow, will cause the space under the line they are placed at (in the document 
tree, not on the typeset boxes) to be reduced to zero. If you remove the 
'Finally some table', it causes the vertical space after the 'Some text 
again' paragraph to reappear.

A workaround
------------

I must admit that having floats stuck at the top of the page is annoying but 
not fatal. But having vertical space removed is NOT acceptable. Luckily, you 
can work around it by forbidding placement at 'here' for floated-up floats.

More info
---------

The floatup2.tm and floatup3.tm documents display interesting behaviour which 
could help track down the origin of the problem.

It looks like the bug is triggered by some side effect taking place at the 
moment the second float is flushed.

The floatup2.tm document shows this pretty clearly. As given, the first and 
second floats are flushed because of the presence of the third float. As a 
proof, try to remove the third float, and the first two floats will be 
flushed one paragraph farther (because the 'some table again' will be a third 
float, and cause the flush).

Any float after the second one will get stuck at the top of the page, even if 
it was flushed at the same time. That is also visible in the floatup2.tm 
document.

The floatup3.tm document uses multicolum to show how this problem may 
interact with the page breaker. Here, 'professional' page breaking is used.

Personal Conlusion
------------------

Is there anyone here, besides Joris, who understands how TeXmacs' typesetter 
works? I'm getting tired of typing typesetter bug reports, they are almost 
always hard to isolate, and long to describe. I think the typesetter code 
would definitely gain from being more widely eyeball-proofed.

Attachment: floatup.tar.bz2
Description: BZip2 compressed data