From:	Maurice Lombardi <Maurice DOT Lombardi AT ujf-grenoble DOT fr>
Newsgroups:	comp.os.msdos.djgpp
Subject:	Problem with last patch of grx23
Date:	Wed, 21 Jun 2000 22:52:40 +0200
Organization:	Universite Joseph Fourier - Grenoble 1
Lines:	62
Message-ID:	<39512B18.CE67EC4C@ujf-grenoble.fr>
NNTP-Posting-Host:	knautie.ujf-grenoble.fr
Mime-Version:	1.0
X-Trace:	amazone.ujf-grenoble.fr 961621833 42831 193.54.234.27 (21 Jun 2000 21:10:33 GMT)
X-Complaints-To:	abus AT ujf-grenoble DOT fr
NNTP-Posting-Date:	21 Jun 2000 21:10:33 GMT
X-Mailer:	Mozilla 4.7 [fr] (Win98; U)
X-Accept-Language:	fr,it,en
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

http://www.gnu.de/software/GRX/grx23.update.000520.gz

1 - it is not gzipped
2 - four hunks failed (with respect to the original grx23.zip, no update)
    in highlow.h grdriver.h makebase.dj and makefile.bcc
    two (in makebase.dj and makefile.bcc) contain 
    "return return linefeed" instead of "return linefeed" as end of lines
    dtou solves that
    the other are cosmetic changes (some extra spaces to align consecutive lines) which seem
different between your version of "original" grx23 and the one on simtelnet. 
    The best is probably to make fresh new diffs rather than to try to modify the present diff file.

3 - more seriously, compiling tests on grx23 fails at the first file (arctest.c) for not finding
GrLine. The problem seems that the src/bgi/line.c coming from bcc2grx (which is merged now) shadows
in the merged library libgrx20.a the src/draw/line.c which comes from the original grx23 (and which
contains the function GrLine() ). This was not a problem with the old setup where compiling with
-lbcc2 -lgrx20 (for bcc2grx) or simply -lgrx20 for pure grx solved the problem. 
In fact looking to the output of the "ar -rv" command in the make command I see that indeed
bgi/line.o is replaced (flagged r) instead of to be added (a). And the same is true for polygon.o
fillpatt.o and fillpoly.o according to the same listing.
Since the problem is with the filenames, not with the functions which they contain (the systematic
prefix Gr in libgrx avoids that), I have renamed these four files (in src/bgi) as lineb.c polygonb.c
fillpatb.c and fillpolb.c and modified accordingly src/stdobjs.mak. All compiles without problem
now.
All tests run perfectly except for one in bccbgi: the sierpinski test (i.e. the first floodfill
test), while the snake test (the second floodfill) works. In looking to floodfill() contained in
src/bgi/fldfill.c I see that you have changed the content of that function with respect to old
BCC2GRX to call directly grx functions. If I restore the old content by replacing at the beginning
of the floodfill() function:
#if (GRX_VERSION_API-0)>=0x0225
by
#if 0
everything works perfectly. I have not tried to understand further.

4 - A directory grx23/chr containing the .chr files (which are presently in src/bgi) should be
included in order that the character tests work (or replace the BGI_PATH in the tests programs, but
creating a chr directory analogous to the font directory seems better).

5 - I have tried the pascal tests. They work except for two problems which were already presents in
previous versions of grx and bgi2grx.
    - pascal/bgi/demos/color.pas fails due to an invalid typecast between real and word. Replacing
word() by a function realtoword() which rounds the real corrects it.
    - the polygon tests (DrawPoly and FillPoly) in pascal/bgi/demos/demo.pas fail due to a problem
when transferring "open arrays" parameters between pascal and C for recent alphas of gpc. gpc pushes
in that case an extra parameter on the stack so that "var Polypoints: array of PointType" in pascal
is no more equivalent to "PointType *Polypoints" in C, because C pushes only the address of
Polypoints. The trick is to use in the pascal unit an untyped variable "var Polypoints" which pushes
also only the address of Polypoints on the stack. With this correction the demo program works
perfectly. It is necessary to change that way pascal/bgi/bgi2grx.pas and also pascal/grx20.pas (I
have no test program at hand to check in that case however).
    
Hope this helps.

-- 
       Maurice Lombardi
Laboratoire de  Spectrometrie Physique,
Universite Joseph Fourier de Grenoble, BP87
38402 Saint Martin d'Heres Cedex     FRANCE
Tel: 33 (0)4 76 51 47 51
Fax: 33 (0)4 76 51 45 44
mailto:Maurice DOT Lombardi AT ujf-grenoble DOT fr

- Raw text -