Re: Problem with arcs in gpic

[Michael Cain]

Following up on my own report after poking through things this  
morning.  In the src/preproc/pic directory, in file object.cpp, in the  
routine declared as "linear_object *object_spec::make_arc", lines 1833  
and 1834 have the statement

    while (radius < d)
      radius *= 2.0;

In my example, in some cases the test (radius < d) was satisfied, but  
the difference in the two values was on the order of 1e-10.  Since  
radius was 0.1, doubling the radius leads to the fairly gross error  
that I was encountering.  For the time being, my workaround is to  
replace that statement with

    if (radius < d)
      radius = d;

which fixed my problem.  The original code appears to have been  
intended to deal with the situation when radius is orders of magnitude  
less than d.  My replacement code simply assigns radius the smallest  
value that avoids problems with the floating point code/libraries  
further on in the routine.

On Aug 26, 2009, at 9:41 AM, Michael Cain wrote:

> I'm having the following problem with drawing a series of arcs in  
> gpic.  I'm running version 1.20.1, compiled from sources on Mac OS X  
> version 10.5.8.  The following code draws a sequence of connected  
> arcs, alternating in size, from top to bottom:
>
> .PS
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
>         arc  cw rad 0.075 to Here+(0,-0.15)
>         arc ccw rad 0.025 to Here+(0,-0.05)
> .PE
>
> Part way through the sequence, the radius for the arcs gets messed  
> up.  This happens in the output from gpic alone, shown here.  I  
> tried the same experiment on an Ubuntu Linux box, also with gpic  
> version 1.20.1, and got the same result.
>
> .lf 1 foo.mm
> .lf 1
> .PS 1.600i 0.100i
> .\" -0.025 -1.6 0.075 0
> .\" 0.000i 1.600i 0.100i 0.000i
> .nr 00 \n(.u
> .nf
> .nr 0x 1
> \h'0.100i'
> .sp -1
> \D't -1.000p'\h'1.000p'
> .sp -1
> \h'0.025i'\v'0.150i'\D'a 0.000i -0.075i -0.000i -0.075i'
> .sp -1
> \h'0.025i'\v'0.150i'\D'a 0.000i 0.025i -0.000i 0.025i'
> .sp -1
> \h'0.025i'\v'0.350i'\D'a 0.000i -0.075i -0.000i -0.075i'
> .sp -1
> \h'0.025i'\v'0.350i'\D'a 0.000i 0.025i -0.000i 0.025i'
> .sp -1
> \h'0.025i'\v'0.550i'\D'a 0.000i -0.075i -0.000i -0.075i'
> .sp -1
> \h'0.025i'\v'0.550i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> \h'0.025i'\v'0.750i'\D'a 0.000i -0.075i -0.000i -0.075i'
> .sp -1
> \h'0.025i'\v'0.750i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> \h'0.025i'\v'0.950i'\D'a 0.000i -0.075i -0.000i -0.075i'
> .sp -1
> \h'0.025i'\v'0.950i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> \h'0.025i'\v'1.150i'\D'a -0.130i -0.075i 0.130i -0.075i'
> .sp -1
> \h'0.025i'\v'1.150i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> \h'0.025i'\v'1.350i'\D'a -0.130i -0.075i 0.130i -0.075i'
> .sp -1
> \h'0.025i'\v'1.350i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> \h'0.025i'\v'1.550i'\D'a -0.130i -0.075i 0.130i -0.075i'
> .sp -1
> \h'0.025i'\v'1.550i'\D'a 0.043i 0.025i -0.043i 0.025i'
> .sp -1
> .sp 1.600i+1
> .if \n(00 .fi
> .br
> .nr 0x 0
> .lf 18
> .PE
> .lf 19
>
>
>
> _______________________________________________
> bug-groff mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-groff