Newsgroups: news.software.b Path: utzoo!henry From: henry@utzoo.uucp (Henry Spencer) Subject: Re: [meta]Configure again (was Re: Peter's gripes) Message-ID: <1990Feb5.173059.8653@utzoo.uucp> Organization: U of Toronto Zoology References: <519@smcnet.UUCP> <1990Jan30.041541.23578@elroy.jpl.nasa.gov> <572@sci34hub.UUCP> <1990Feb4.222953.19169@utzoo.uucp> <1990Feb5.053532.2286@iwarp.intel.com> Date: Mon, 5 Feb 90 17:30:59 GMT In article <1990Feb5.053532.2286@iwarp.intel.com> merlyn@iwarp.intel.com (Randal Schwartz) writes: >| >Maybe some code that would run test compiles to find out if the >| >functions are there or not..... with that, build could answer some >| >of the questions itself. >| It's really hard to do this 100% portably... > >Well, why not just steal all the stuff from Larry's Configure and >Metaconfigure? ... There were three reasons why we didn't do this, and don't plan to: 1. Our experience with Larry's stuff -- admittedly not with the latest versions -- indicated that it was easily confused by non-standard systems. There are too many variables. For example, programs like nm(1) -- exactly what you want for figuring out such things -- have very system-dependent output formats. [Insert 3000-word flame about System V cretins who don't understand that nm(1) output is useful to programs, not just to humans, and hence should have kept its simple standardized original format.] 2. Automatic answers based on the current system are almost guaranteed to be wrong if you're trying to configure software that will run on some other machine. 3. Out of 70+ questions in build, about fifteen could perhaps be answered automatically with some hope of correctness. Not worth it. -- SVR4: every feature you ever | Henry Spencer at U of Toronto Zoology wanted, and plenty you didn't.| uunet!attcan!utzoo!henry henry@zoo.toronto.edu