[gs-devel] RE: [gs-code-review] RE: Radial shading regressions

[Jeong Hee Kim]

It is just reminding post.
Now can I commit the code ?

Jeong

> -----Original Message-----
> From: gs-code-review-admin at ghostscript.com 
> [mailto:gs-code-review-admin at ghostscript.com]On Behalf Of Jeong Hee Kim
> Sent: Thursday, May 09, 2002 1:53 PM
> To: Jack Moffitt
> Cc: gs-code-review at ghostscript.com
> Subject: [gs-code-review] RE: Radial shading regressions
> 
> 
> Jack,
> 
> I have found that there were problems when a radius of circle is 0.
> I fixed it and confirmed that there are no difference between the output
> from GS7.04 and the fixed one with 442-01.ps and 483-01.ps.
> But was there any regression problem with 245-01 ?
> I think there was no problem with 245-01.
> 
> Following is the diff.
> 
> Thank you.
> Jeong
> 
> Index: gxshade1.c
> ===================================================================
> RCS file: /cvs/ghostscript/gs/src/gxshade1.c,v
> retrieving revision 1.14
> diff -r1.14 gxshade1.c
> 748,749c748,771
> <  * For differnt radii, compute the coords for /Extend
> <  * r0 must be greater than r1 
> ---
> >  * For differnt radii, compute the coords for /Extend option.
> >  * r0 MUST be greater than r1.
> >  *
> >  * The extension is an area which is bounded by the two exterior common
> >  * tangent of the given circles except the area between the circles.
> >  *
> >  * Therefore we can make the extension with the contact points between
> >  * the tangent lines and circles, and the intersection point of
> >  * the lines (Note that r0 is greater than r1, therefore the 
> exterior common 
> >  * tangent for the two circles always intersect at one point.
> >  * (The case when both radii are same is handled by 
> 'R_compute_extension_bar')
> >  * 
> >  * A brief algorithm is following.
> >  *
> >  * Let C0, C1 be the given circle with r0, r1 as radii.
> >  * There exist two contact points for each circles and
> >  * say them p0, p1 for C0 and q0, q1 for C1.
> >  *
> >  * First we compute the intersection point of both tangent 
> lines (isecx, isecy).
> >  * Then we get the angle between a tangent line and the line 
> which penentrates
> >  * the centers of circles.
> >  *
> >  * Then we can compute 4 contact points between two tangent 
> lines and two circles,
> >  * and 2 points outside the cliping area on the tangent lines.
> 750a773
> > 
> 757,759c780,783
> <     floatp dis;
> <     floatp cir_dis;
> <     floatp k;
> ---
> > 	floatp dist_c0_isec;
> >     floatp dist_c1_isec;
> > 	floatp dist_p0_isec;
> >     floatp dist_q0_isec;
> 761d784
> <     floatp l;
> 764,770c787,788
> <     coord[0][0] = isecx = (x1-x0)*r0 / (r0-r1) + x0;
> <     coord[0][1] = isecy = (y1-y0)*r0 / (r0-r1) + y0;
> < 
> <     dis = hypot(x1-isecx, y1-isecy);
> <     k = sqrt(dis*dis-r1*r1);
> <     cost = k / dis;
> <     sint = r1 / dis;
> ---
> >     isecx = (x1-x0)*r0 / (r0-r1) + x0;
> >     isecy = (y1-y0)*r0 / (r0-r1) + y0;
> 772,773c790,795
> <     cir_dis = hypot(x1-x0, y1-y0);
> <     l = sqrt(cir_dis*cir_dis - (r0-r1)*(r0-r1));
> ---
> > 	dist_c0_isec = hypot(x0-isecx, y0-isecy);
> >     dist_c1_isec = hypot(x1-isecx, y1-isecy);
> > 	dist_p0_isec = sqrt(dist_c0_isec*dist_c0_isec - r0*r0);
> > 	dist_q0_isec = sqrt(dist_c1_isec*dist_c1_isec - r1*r1);    
> >     cost = dist_p0_isec / dist_c0_isec;
> >     sint = r0 / dist_c0_isec;
> 775,776c797,798
> <     dx0 = ((x1-isecx)*cost - (y1-isecy)*sint) / dis;
> <     dy0 = ((x1-isecx)*sint + (y1-isecy)*cost) / dis;
> ---
> >     dx0 = ((x0-isecx)*cost - (y0-isecy)*sint) / dist_c0_isec;
> >     dy0 = ((x0-isecx)*sint + (y0-isecy)*cost) / dist_c0_isec;
> 778,779c800,801
> <     dx1 = ((x1-isecx)*cost - (y1-isecy)*sint) / dis;
> <     dy1 = ((x1-isecx)*sint + (y1-isecy)*cost) / dis;
> ---
> >     dx1 = ((x0-isecx)*cost - (y0-isecy)*sint) / dist_c0_isec;
> >     dy1 = ((x0-isecx)*sint + (y0-isecy)*cost) / dist_c0_isec;
> 781,784c803,808
> <     coord[1][0] = isecx + dx0 * k;
> <     coord[1][1] = isecy + dy0 * k;
> <     coord[2][0] = isecx + dx1 * k;
> <     coord[2][1] = isecy + dy1 * k;
> ---
> > 	coord[0][0] = isecx;
> > 	coord[0][1] = isecy;
> >     coord[1][0] = isecx + dx0 * dist_q0_isec;
> >     coord[1][1] = isecy + dy0 * dist_q0_isec;
> >     coord[2][0] = isecx + dx1 * dist_q0_isec;
> >     coord[2][1] = isecy + dy1 * dist_q0_isec;
> 786,787c810,811
> <     coord[3][0] = isecx + dx0 * (l+k);
> <     coord[3][1] = isecy + dy0 * (l+k);
> ---
> >     coord[3][0] = isecx + dx0 * dist_p0_isec;
> >     coord[3][1] = isecy + dy0 * dist_p0_isec;
> 790,791c814,815
> <     coord[5][0] = isecx + dx1 * (l+k);
> <     coord[5][1] = isecy + dy1 * (l+k);
> ---
> >     coord[5][0] = isecx + dx1 * dist_p0_isec;
> >     coord[5][1] = isecy + dy1 * dist_p0_isec;
> 
> 
> 
> > -----Original Message-----
> > From: Jack Moffitt [mailto:jack at artifex.com]
> > Sent: Thursday, May 09, 2002 2:56 AM
> > To: Jeong Hee Kim
> > Cc: gs-devel at ghostscript.com
> > Subject: Radial shading regressions
> > 
> > 
> > Jeong,
> > 
> > Our nightly regression found 3 files that failed after your radial
> > shading patch.  We'd appreciate it if you could investigate these errors
> > and see if there was a bug in your patch.
> > 
> > The files were 245-01.ps, 442-01.ps, and 483-01.ps, all available in
> > on casper in /home/regression/comparefiles.
> > 
> > Thanks,
> > jack.
> 
> _______________________________________________
> gs-code-review mailing list
> gs-code-review at ghostscript.com
> http://www.ghostscript.com/mailman/listinfo/gs-code-review
>