FW: [Bioperl-l] need help making biographics tests

Lincoln Stein lstein@cshl.org
Sun, 21 Jul 2002 09:44:24 -0700


Hi,

Good suggestion, but it's not an issue of byte order, date encodings,
character encodings, or anything else that is going to be helped by
converting to a text format!  There are subtle differences in color
values due among different versions of libgd and libpng, presumably
due to the 24->8 bit conversion that libd performs.

I'm currently moving the GD library to use libgd 2.0, which generates
24-bit image colors.  This will make the situation slightly worse,
because there will be some machines that use libgd 1.X and others that
use 2.X.

Lincoln

Hilmar Lapp writes:
 > Hmm. This is not going to be very extensible to neither additional 
 > platforms nor additional tests, and maintenance won't be fun 
 > either.  Assuming that the problem is related to byte order, date 
 > encodings, character encodings, and similar things, what about 
 > converting before the comparison step all generated images to a text 
 > format, e.g. PGM or PNM, which would be unaffected by any of the 
 > aforementioned problems, and would compare the pixel values only 
 > (PGM doesn't contain much more). So long as the images aren't 
 > largish ...
 > 
 > Just a suggestion.
 > 
 > 	-hilmar
 > 
 > On Saturday, July 20, 2002, at 06:49  AM, Lincoln Stein wrote:
 > 
 > > Hi All,
 > >
 > > There is an annoying problem with the bio::graphics module.  Depending
 > > on the combination of various versions of libgd, libfreetype and
 > > libpng that happen to be installed on the target system, bio::graphics
 > > will generate visually identically but binary different picture files.
 > > As a result, the first three regression tests fail when run on a
 > > different machine from which they were generated.
 > >
 > > To fix this, I need to collect images generated on various
 > > architectures.  Could you help me out by doing the following?
 > >
 > >   1) cd to the "live" bioperl
 > >   2) run the command perl t/BioGraphics.t --write
 > >   3) find the generated files t/data/biographics/t1.png, t2.png and
 > >      t3.png and send them to me as attachments?
 > >
 > > I will construct the set of unique variants and rewrite the regression
 > > tests to return true if any of them match.
 > >
 > > Thanks much!
 > >
 > > Lincoln
 > >
 > >
 > > --
 > > ========================================================================
 > > Lincoln D. Stein                           Cold Spring Harbor 
 > > Laboratory
 > > lstein@cshl.org			                  Cold Spring Harbor, NY
 > > Positions available at my lab: see http://stein.cshl.org/#hire
 > > ========================================================================
 > > _______________________________________________
 > > Bioperl-l mailing list
 > > Bioperl-l@bioperl.org
 > > http://bioperl.org/mailman/listinfo/bioperl-l
 > >
 > 	-hilmar
 > --
 > -------------------------------------------------------------
 > Hilmar Lapp                            email: lapp at gnf.org
 > GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
 > -------------------------------------------------------------
 > 
 > 
 > 
 > 

-- 
========================================================================
Lincoln D. Stein                           Cold Spring Harbor Laboratory
lstein@cshl.org			                  Cold Spring Harbor, NY
Positions available at my lab: see http://stein.cshl.org/#hire
========================================================================