From pijnacker at dse.nl Fri Mar 30 04:18:27 2007 From: pijnacker at dse.nl (Ronald Pijnacker) Date: Fri, 30 Mar 2007 10:18:27 +0200 Subject: [Mkrf-users] Support for building on Windows In-Reply-To: References: Message-ID: <20070330081827.GI31850@best.ms.philips.com> Hi all, To get ruby-opengl built on windows, some changes are necessary to mkrf. Before starting on this, I have some questions: Is is allright to limit support to the Microsoft tool-chain (cl/link) for now? Since this is the only tool-chain I have available now, this would be my preferred choice. For linking windows dlls, one should make explicit which symbols should be exported. This can be done in code, but also by passing command line parameter '/export:' multiple time, or by passing a file containing these symbols using paramter '/def:' . Should all of these be supported, or do we want to force the users into writing the code such that these parameters are not necessary? One final question: where do I sent the patches? To the list, or to the maintainer? Groeten, Ronald. From hoanga at alum.rpi.edu Fri Mar 30 04:50:06 2007 From: hoanga at alum.rpi.edu (Al Hoang) Date: Fri, 30 Mar 2007 17:50:06 +0900 Subject: [Mkrf-users] Support for building on Windows In-Reply-To: <20070330081827.GI31850@best.ms.philips.com> References: <20070330081827.GI31850@best.ms.philips.com> Message-ID: <20070330085006.GD11009@samsara.bebear.net> On Fri, Mar 30, 2007 at 10:18:27AM +0200, Ronald Pijnacker wrote: > Hi all, > > To get ruby-opengl built on windows, some changes are necessary to mkrf. > Before starting on this, I have some questions: > > Is is allright to limit support to the Microsoft tool-chain (cl/link) > for now? Since this is the only tool-chain I have available now, this > would be my preferred choice. This is fine. I'd prefer it work first with the Microsoft tool chain right now as I imagine that more Windows developers will be using the MS tool chain than other toolchains. > > For linking windows dlls, one should make explicit which symbols should > be exported. This can be done in code, but also by passing command line > parameter '/export:' multiple time, or by passing a file containing > these symbols using paramter '/def:' . > Should all of these be supported, or do we want to force the users into > writing the code such that these parameters are not necessary? I'm not familiar with the pros and cons of each method. Are there tradeoffs for forcing developers to write their code to hide the details of sending command line parameters? My gut feeling is that support for these parameters would be good as there might be a bit of code people might want to port to Windows where changing the source code might not be that trivial but it still might be possible to muck with the build process in order to get the code to compile properly. > > One final question: where do I sent the patches? To the list, or to the > maintainer? To the list is better, but to the maintainer is fine as well. That happens to be me.