From: Alex R. M. <ale...@mo...> - 2004-11-18 12:18:07
|
Brian Gerkey wrote: > On Tue, 16 Nov 2004, Brian Gerkey wrote: > >> On Tue, 16 Nov 2004, Alex R. Mosteo wrote: >> >>> playerclient.h contains some specs for functions returning int which >>> later in the implementation are declared as returning int32_t. The >>> gcc3.3 in cygwin chokes on it, while my 3.4 in linux doesn't say >>> anything. I've changed the int to int32_t. Should it be the other way >>> around? >>> >>> camera.h redefines int32_t to a different type (unsigned int instead >>> of long unsigned int). I've matched this redefinition. Would it be >>> simply better to remove it completely? >> >> >> Yep, those should be fixed. One of the problems with developing >> primarily >> in Linux is that it's so forgiving. > > > I've fixed both of these. In the C++ client, rather than change the > method declarations to return int32_t as Alex's patch did, I've changed > the implementations to return int. In general, the user shouldn't have > to use specifically sized types, unless it's strictly necessary. > > In cmucam2's camera.h, I took out the redefines of the standard types, and > #included "playerconfig.h" instead, which defines them in a portable way. > >>> cannonvcc4.cc contains a call to cfmakeraw function in termios.h >>> which simply doesn't exist in the termios.h supplied with cygwin. >>> I've commented this line. Anyone knows of some workaround? I guess >>> disabling this device would suffice too if you don't need it. > > > I've added glibc's implementation of cfmakeraw(3). It will be compiled > into replace/libreplace.a if it's not found during configure. I've > changed all the drivers that call cfmakeraw(3) to #include "replace.h". > Then they can call cfmakeraw() unconditionally, without checking > HAVE_CFMAKERAW. Thanks, Brian, you're really quick. I'm trying to obliterate cygwin from my HD to restart from scratch. I'll post any relevant news. Kind regards, Alex. |