[sane-devel] problems compiling on cygwin and a java question

Giuseppe Sacco giuseppe@eppesuigoccas.homedns.org
Sun, 22 Aug 2004 07:47:12 +0200

Henning Meier-Geinitz wrote:
> Hi,
>>1. if we just use the vanilla 'configure' without specifing any 
>>arguments then we get an error when compiling backend/coolscan.c, at 
>>line 2593. The error is: sa_handler field doesn't exists.
> Strange. sa_handler is also used by other backends. Maybe we need
> another definition for SIGACTION in sanei/sanei_backend.h?

I checked even with the latest gcc 3.3.3. I cannot tell what happened, 
but the 1.0.14 is compiled correctly while the CVS snapshot isn't.

>>2. if you use a different compiler, i.e., 'gcc-2' instead of 'gcc', the 
>>compilation process stops on backend/gt68xx.c, line 242, since SHM_R is 
>>not definded. How may we tell 'configure' to do not compile a specific 
>>backend? Should we specify all other backends we need using the BACKENDS 
>>gt68xx_shm_channel.c: In function `shm_channel_new':
>>gt68xx_shm_channel.c:242: `SHM_R' undeclared (first use in this function)
>>gt68xx_shm_channel.c:242: (Each undeclared identifier is reported only once
>>gt68xx_shm_channel.c:242: for each function it appears in.)
>>gt68xx_shm_channel.c:242: `SHM_W' undeclared (first use in this function
> I'll add a workaround for that one to CVS. If you like to use that
> backend, you can also edit gt68xx.c and #undef USE_FORK.

For this I found a different fix: Instead of including ipc.h ans shm.h I 
did included cygipc/ipc.h and cygipc/shm.h. I did this change in all 
backends that require SHM_R.

Moreover I found another problem when using the libusb-win32: please 
have a look at the the bug tracking in Alioth (when it will be up again).

Finally there is a problem when trying to compile the japi directory: a 
file Sane.h is missing and the makefile looks for libsane.so in 
../backend/libs instead of ../backend/.libs . (I did not reported the 
bug since Alioth is unreachable).