[sane-devel] Patches for SANE 1.0.9 be/fe compiled on IRIX

[Henning Meier-Geinitz]

Hi,

On Sun, Dec 29, 2002 at 01:54:41AM +0100, Andrea Suatoni wrote:
> I have packaged the SANE 1.0.9 backends / frontends for IRIX, using the SGI
> MIPSpro C compiler. The packages should upgrade the 1.0.8 versions currently
> available on http://freeware.sgi.com/

Thanks.

> In order to compile properly the two packages,

Was there anything that didn't compile at all? I don't ahve access to
Irix any longer, so the last version tested was 1.0.8.

I have some questions about some of these patches:

> +++ sane-backends-1.0.9-patched/backend/Makefile.in	Thu Dec 12 21:55:32 2002
> @@ -46,6 +46,7 @@
>  CFLAGS = @CFLAGS@
>  LDFLAGS = @LDFLAGS@
>  BACKENDLIBS = @LIBS@ @DL_LIB@
> +GPHOTO2_LIBS = @GPHOTO2_LIBS@
>  DEFS = @DEFS@
>  
>  LIBTOOL = ../libtool
> @@ -135,6 +136,11 @@
>  
>  
>  .PHONY: all clean depend dist distclean install uninstall
> +
> +libsane-gphoto2.la: gphoto2.lo gphoto2-s.lo $(EXTRA) $(LIBOBJS)
> +	@$(LIBTOOL) $(MLINK) $(CC) -export-dynamic -o $@ $($*_LIBS) \
> +	$(LDFLAGS) $(GPHOTO2_LIBS) $(BACKENDLIBS) $^ -rpath $(libsanedir) \
> +	-version-info $(V_MAJOR):$(V_REV):$(V_MINOR)
>  
>  libsane-%.la: %.lo %-s.lo $(EXTRA) $(LIBOBJS)
>  	@$(LIBTOOL) $(MLINK) $(CC) -export-dynamic -o $@ $($*_LIBS) \

What's the reason for the gphoto2 changes and the separate rule for
building the library?

> diff -ruN sane-backends-1.0.9/backend/as6e.c sane-backends-1.0.9-patched/backend/as6e.c
> --- sane-backends-1.0.9/backend/as6e.c	Tue Dec  5 20:10:20 2000
> +++ sane-backends-1.0.9-patched/backend/as6e.c	Thu Dec 12 21:56:54 2002
> @@ -604,8 +604,7 @@
>  	  return (SANE_STATUS_GOOD);
>  	}			/*else */
>      }
> -  else
> -    return (SANE_STATUS_IO_ERROR);
> +  return (SANE_STATUS_IO_ERROR);
>  }

Is this to fix a warning or is this a real change in code?

> +++ sane-backends-1.0.9-patched/backend/canon-sane.c	Thu Dec 12 21:58:46 2002
> @@ -1798,7 +1798,7 @@
>    SANE_Status status;
>    SANE_Byte *out, *red, *green, *blue;
>    SANE_Int ncopy;
> -  size_t nread, i, pixel_per_line;
> +  size_t nread = 0, i, pixel_per_line;
>  
>    DBG (21, ">> read_fb620\n");
>  

That's because of the DBG that prints nread before initializing it?
The correct fix is to move the DBG after the setting nread, I guess.

> +++ sane-backends-1.0.9-patched/include/sane/config.h.in	Thu Dec 12 23:36:53 2002
> @@ -393,3 +393,13 @@
>  
>  /* Define to `unsigned long' if <sys/types.h> does not define. */
>  #undef u_long
> +
> +/* Define a the function prototypes if these function are missing. */
> +#ifndef HAVE_STRSEP
> +char *strsep(char **stringp, const char *delim);
> +#endif
> +
> +#ifndef HAVE_STRNDUP
> +#include <sys/types.h>
> +char *strndup(const char * s, size_t n);
> +#endif

This will be removed automatically with the next run of autoheader. If
it makes sense to add these declarations, I can put the code for
generation in configure.in.


> +++ sane-frontends-1.0.9-patched/src/Makefile.in	Fri Dec 13 00:39:37 2002
> @@ -83,7 +83,7 @@
>  	    $(DESTDIR)$(bindir)/$${program}; \
>  	done
>  	$(INSTALL_DATA) $(srcdir)/sane-style.rc \
> -          $(DESTDIR)$(datadir)/sane-style.rc
> +          $(DESTDIR)$(sanedatadir)/sane-style.rc
>  
>  uninstall:
>  	@for program in $(BINPROGS); do \
> @@ -90,7 +90,7 @@
>  	  echo removing $(bindir)/$${program}...; \
>  	  rm -f $(bindir)/$${program}; \
>  	done
> -	rm -f $(datadir)/sane-style.rc
> +	rm -f $(sanedatadir)/sane-style.rc
>  
>  xscanimage: $(XSCANIMAGE_OBJS) $(LIBSANEI) $(LIBLIB)
>  	$(LINK) $(XSCANIMAGE_OBJS) $(LIBSANEI) \

That one is cool. Nobody has ever found out about this one. Will be
applied to CVS.

> diff -ruN sane-frontends-1.0.9/src/xcam.c sane-frontends-1.0.9-patched/src/xcam.c
> --- sane-frontends-1.0.9/src/xcam.c	Sat Jun  9 14:52:05 2001
> +++ sane-frontends-1.0.9-patched/src/xcam.c	Fri Dec 13 00:41:42 2002
> @@ -56,8 +56,6 @@
>    }
>  Canvas;
>  
> -Preferences preferences;
> -
>  static const char *prog_name;
>  static const SANE_Device **device;
>  static GSGDialog *dialog;
> @@ -672,13 +670,29 @@
>  									\
>      case 24:								\
>        /* Is this correctly handling all byte order cases? */		\
> -      if ((endian) == GDK_LSB_FIRST)					\
> +      if ((endian) != GDK_LSB_FIRST)					\
>  	{								\
> -	  buf[0] = (b) >> 8; buf[1] = (g) >> 8; buf[2] = (r) >> 8;	\
> +	  if (bpp == 4)							\
> +	    {								\
> +	      buf[1] = (b) >> 8; buf[2] = (g) >> 8; buf[3] = (r) >> 8;	\
> +	      buf[0] = 0;						\
> +	    }								\
> +	    else							\
> +	    {								\
> +	      buf[0] = (b) >> 8; buf[1] = (g) >> 8; buf[2] = (r) >> 8;	\
> +	    }								\
>  	}								\
>        else								\
>  	{								\
> -	  buf[0] = (r) >> 8; buf[1] = (g) >> 8; buf[2] = (b) >> 8;	\
> +	  if (bpp == 4)							\
> +	    {								\
> +	      buf[1] = (r) >> 8; buf[2] = (g) >> 8; buf[3] = (b) >> 8;	\
> +	      buf[0] = 0;						\
> +	    }								\
> +	    else							\
> +	    {								\
> +	      buf[0] = (r) >> 8; buf[1] = (g) >> 8; buf[2] = (b) >> 8;	\
> +	    }								\
>  	}								\
>        buf += (bpp);							\
>        break;								\
> @@ -687,7 +701,7 @@
>        /* Is this correctly handling all byte order cases?  It assumes	\
>  	 the byte order of the host is the same as that of the		\
>  	 pixmap. */							\
> -      rgb = (((r) >> 8) << 16) | (((g) >> 8) << 8) | ((b) >> 8);	\
> +      rgb = (((b) >> 8) << 16) | (((g) >> 8) << 8) | ((r) >> 8);	\
>        ((guint32 *)buf)[0] = rgb;					\
>        buf += (bpp);							\
>        break;								\

That one doesn't work on Linux/X11/i386. The colors are wrong. The old
code was correct (at least for 24 bit, 4 bytes per pixel, little
endian). Well, at least it worked here :-)

I'm fascinated that someone actually cares about xcam :-)

Bye,
  Henning