page offset lost after page break

[Steve Kleene]

                           Groff Bug Report

Send the completed form to address@hidden

GROFF VERSION: 
GNU groff version 1.18.1

MACHINE:
ICT Intel Pentium 4 3.0 GHz processor (Essential series)

OS:
Linux drx 2.6.18-5-686 #1 SMP Mon Dec 24 16:41:07 UTC 2007 i686 GNU/Linux

COMPILER:
Debian Etch packages
  groff 1.18.1.1-12 GNU
 groff-base 1.18.1.1-12

INPUT FILES:

============================begin input file=================================
.ds CH

.LP
.nh
.po 0.5i
.ll 7.5i
.ps 11
.vs 0.175i
.de PP
.sp 0.08i
.ti 0.3i
..
.PP
Specific Aim\ 1 tests the hypothesis that each of the two channel types (CNG
and Cl channels) is concentrated in a single small portion of the ciliary
length.
In the case of the CNG channels, we have already confirmed that channel
expression is confined to a narrow band by modeling electrophysiological
recordings [23,24].
.bp
.PP
As shown in the Preliminary Studies (Figs. 3 and 4), our preliminary model
represents the experimental result well with a range of tested Ca
buffers.
However, the modeled slope of the current-vs-time relation is somewhat less
steep than the experimental slope.

.ce
.TS
box;
lfI |cfI |cfI
l|c |c.
Constant        Value   References
=
total [Ca]      0.5-2.3 mM      set by choice of solutions
_
total [buffer]  0.2-2.0 mM      set by choice of solutions
.TE
=============================end input file==================================

COMMAND LINE:
groff -mgs -t >z
(I have gs.tmac as a link to s.tmac)

DESCRIPTION OF INCORRECT BEHAVIOUR:
The input requests a page offset of 0.5i, but after the page break it gets
reset to 1i.

The offset remains 0.5i if I delete ll. 2-3 (blank line + ".LP").  In that
case, though, the table fails at the first line (".TS") with this error:

  <standard input>:27: fatal error: input stack limit exceeded (probable 
infinite loop)

If I then also delete the four-line definition of .PP and the two calls to
it, I get the desired page offsets and table, although the request to center
the table is ignored.


SUGGESTED FIX [optional]:
I could get the needed PS if I deleted lines 2-3, all definitions and calls
to .PP, and then pushed the table to the center with a .po determined by
trial and error.  .ce, .ti, and .in failed.

Without the table or gtbl, I could fix the offset problem just by deleting
ll 2-3.  But that messed up the table.  I can send a similar sample problem
that uses geqn instead of gtbl.