Newsgroups: comp.emacs Path: utzoo!utgpu!news-server.csri.toronto.edu!torsqnt!geac!maccs!antel!mike From: mike@antel.uucp (Michael Borza) Subject: Re: uemacs 3.10m/3.11B porting questions Reply-To: mike@antel.uucp (Michael Borza) Organization: Antel Optronics Inc., Burlington, Ontario, Canada Date: Tue, 2 Apr 91 14:23:33 GMT Message-ID: <1991Apr2.142333.7871@antel.uucp> References: <1991Mar28.083424.7514@odin.corp.sgi.com> <1991Mar29.092148.6534@irisa.fr> In article <1991Mar29.092148.6534@irisa.fr> decouty@irisa.fr (Bertrand Decouty) writes: >- in eproto.h, mlwrite(va_dcl) must be mlwrite(va_alist) (if you use >varargs) >============================== >If anybody have other patches... >Bertrand Decouty well, now that you mention it,.... Compiling uemacs3.11beta with Turbo C++ 2.0, I found the following change to `mlwrite' in `display.c' was required to get it to build: change: CDECL NEAR mlwrite(fmt) char *fmt; {... register va_list ap; to: CDECL NEAR mlwrite(char *fmt,....) { .... va_list ap; If not, TC++ 2.0 fails later in the compile when an attempt to take the address of `ap' is made in TC++'s varargs implementation. I also have other minor changes to makefiles and the like, which I'll submit to Dan once I have changes to build using Zortech C++ as well. While I'm posting, I'd like an opinion: I propose to explicitly declare routines which are to return no value to return VOID. This is done inconsistently throughout the code, resulting in unnecessary warnings for a (largish) number of routines. I found this especially annoying using a heavily "ANSIfied" compiler like TC++, but that's probably just years of pcc's indifference talking. In any event, this change should be transparent, since VOID is suitably defined in ESTRUCT.H for each target system. Comments? cheers, mike borza. -- Michael Borza Antel Optronics Inc. (416)335-5507 3325B Mainway, Burlington, Ont., Canada L7M 1A6 work: mike@antel.on.ca or uunet!utai!utgpu!maccs!antel!mike home: mike@boopsy.uucp or uunet!utai!utgpu!maccs!boopsy!mike