>Just to chime in here with some past experience.. Is there any way it >would be acceptable to use gcc (or even the Intel compiler) as the >mandatory Windows C compiler? That would save everyone alot of >ongoing hassle. MS does not maintain the C compiler portion of VC++ >and it is very old standards wise, half your changes in this patch are >due to it not supporting C99.
I installed the Intel compiler (version 11.0.066) and tried using that within the WDK build environment to build just sminfo. The good news is that sminfo did build within the WDK environment and run. The bad news is that every change to sminfo.c that was posted was still needed by the Intel compiler, plus it required a couple of other changes as well. :( I didn't spend any time looking into the compile issues, so I don't know if changing the build environment would eliminate some of the changes. I also did not try using gcc on Windows. (Btw, I think we can fix the const issue.) I would like to avoid the other changes, but it's not looking like it will happen. >So, really what you are proposing is to abandon all modern C >constructs in the offical source tree :| Some of this is acutally >harmful run-time wise (like removing const on the static variables) >and harmful maintenance wise (removing C99 named initalizers) What I'm really proposing is that the IB management utilities package support both Linux and Windows. The alternative is to have independent packages with separate source code bases. And unless there's a way to eliminate the changes, they'll be there. I just don't know where there is yet. Btw, Arlin can provide more details on the other required changes. We only have a few of the diags ported at this time (i.e. the easiest ones to port). - Sean _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
