Re: Plea for minimal tagbits

Common combinations of options should simply work on all platforms. I've done a lot of work to make sure that union-type+mule works on Solaris, for example. Each such porting exercise cleans up the code base and makes XEmacs more portable. If we have a known combination of options that fail on some platform, it is very likely that it is either a basic cross-platform bug in XEmacs that needs to be fixed, or some platform-specific thing that can be tweaked in a few lines of code. Although we can't test every combination of options, some combinations are guaranteed to work everywhere: - configure - configure --with-mule The only excuse for things to be put into PROBLEMS is a vendor-supplied bug (such as HP's sed), not an XEmacs-supplied one. Although --gung-ho will not be the default (except perhaps for some platforms), it will be useful as a bug exposer. Developers demand more core dumps! OG> On Thu, Apr 30, 1998 at 08:35:06AM +0200, Oscar Figueiredo wrote: OG> That's *exactly* the reason why I want one and only one option being OG> chosen. I did not know about this problem, even if I have the OG> resources (i.e. solaris 2.5.1 and the Sun compiler) to solve OG> it. People just compile without minimal tagbits and forget it. The OG> opposite goes for, say, cygwin and the lack of valid DATA_SEG_BITS OG> definition. OG> There is the same kind of problem with union type and alpha. Union OG> type is required on alpha. Nobody knows why, mostly because everybody OG> forgot about it and every people compiling on alpha is adding OG> --use-union-type without thinking about it. Now that I know about it, OG> I'm going to track it (there is something inflooping in the gc code). OG> This may or may not unravel other problems. OG> Such habits are understandable, but bad when we're approaching a OG> release. I wouldn't like to see "Use minimal tagbits on cygwin", OG> "don't use them on solaris with sun's cc" and "use union type on OG> alpha" becoming answers to FAQs in comp.emacs.xemacs. But it is very OG> easy to miss them if the failure report happens to come in at the OG> wrong time. Right. Let's fix them. OG> So please people, if you have some experience of "if I don't do this OG> and that it does not work on my system" and the information is not yet OG> in the PROBLEMS file, say so. Loudly. Every other snapshot. Or it will OG> come back and bite us for the release. NOW is the time to make sure that all combinations of options work. OG> And please Steve, at least when code freeze is enforced, settle for a OG> common set of options and enforce them in configure.in. I, for OG> instance, have access to 7 different systems[1], 13 compilers total. I OG> won't try 16[2] different options combinations on each. Martin