On Sunday 05 January 2003 05:15 pm, Alexandre Julliard wrote:
Eric Pouech eric.pouech@wanadoo.fr writes:
what about an environment variable ? we already use something like this for the wineserver (WINESERVER), the wine console (WINECONSOLE)... why not rpcss (WINERPCSS) ?
I don't think we want a separate environment variable for every single Wine program. Simply setting the correct WINEDLLPATH should be enough (and runtest should be doing that already AFAICS).
I think this is also not cross platform portable. I probably need to generate a matrix of _s.c, _c.c, and .h files for powermac/linux, sparc, ia64, er.. what was the list again? Ugh. That approach is not going to scale very gracefully... widl, once it is capable, is the easy-to-identify solution, but presumably the full infrastructure that would allow RPC stub generation would be quite an undertaking... all those nasty modes, /Oi.* business, architecture switches, and, perhaps worst of all, format-string generation... probably a barrel of snakes in there alone.
I guess, the real answer, is go brush up on my bison skills... until then, I guess a matrix of files is the best I can do.
I plan to implement that using the library aproach mentioned by Eric, instead of the one-big-exe approach, also mentioned by Eric, to achieve this. It's more verbose, but, it supports the needs of a multiplatform test, and is the more elegant solution w.r.t. widl in this intermediate stage: widl can very nearly generate the stringinput.h I would need for libraries, whereas with this /rename trick, I had to use MIDL's stringinput.h.
I think you shouldn't comitt my K01 until I fix some of that stuff; it will probably break a lot of trees to throw it into CVS in this condition. Hopefully, I can figure out the basic make incantations, and if not, well, you will hear from me.