Nephero wrote:
> Looking at the diff, I have a couple of questions:
>
> 1. "configure" has already been run, and lex/yacc source files have be
> processed to produce the intermediate C files. Is this due to Debian
> policy, or is it an artifact of the way that the package was produced?
This was an artifact, Federico already pointed this problem, and tomorrow
I'll see what can be done.
OK. If you don't intend to include these files in the source package,
you should run "make maintainer-clean" to get rid of them. A quick
summary of cleanup make targets, from least to most aggressive:
mostlyclean: removes object files, and a few others.
clean: also removes the libraries, and the staging directories
(basically a local copy of everything that gets installed) containing
the executables.
distclean: also removes the files which are generated by the configure
script.
maintainer-clean: also removes the C files which are generated by
lex/yacc.
> 2. What was the purpose of this patch:
>
> --- grass-5.0.2.orig/src/display/devices/lib/Text2.c
> +++ grass-5.0.2/src/display/devices/lib/Text2.c
> @@ -170,9 +170,11 @@
> double text_size_x,double text_size_y,double text_rotation,
> char *string)
> {
> - double sinrot = sin (D2R (text_rotation)) ;
> - double cosrot = cos (D2R (text_rotation)) ;
> -
> +double sinrot,cosrot;
> + //double sinrot = sin (D2R (text_rotation)) ;
> + //double cosrot = cos (D2R (text_rotation)) ;
> + sinrot = sin (D2R (text_rotation)) ;
> + cosrot = cos (D2R (text_rotation)) ;
> am_inside = 0 ;
> curx = basex = (double)x ;
> cury = basey = (double)y ;
>
> Is this due to a gcc bug?
The png driver didn't compile correctly, (but the package was created and
usable anyway) so I made this little fix and now it seems fine (compiles and
runs fine). I think the catch was the definition/assignment in the same
line. I basically just splitted them in different lines. The gcc I used is
the one bundled with Debian Woody. It doesn't seem to me a gcc bug. Maybe
Federico or someone else can better answer this question than me. Please
remember the patch applies to 5.0.2, Silke already pointed out that it's no
use for the CVS version, that had already fixed this problem indipendently.
There is a known problem with the PNG driver in 5.0.2, caused by the
omission of $(MATHLIB) from the Gmakefile. Your patch includes a fix
to the Gmakefile.
The above patch will have no effect if you have a working compiler
(except that it will cause compilation to fail if the compiler doesn't
support C++ double-slash comments).
Perhaps someone tried the above, then realised that the problem was
actually in the Gmakefile, but left the changes in? Please try it
without the above patch; if it fails, I would like to see the error
message.
--
Glynn Clements <glynn.clements@virgin.net>