urlichs@smurf.sub.org (Matthias Urlichs) (08/24/90)
In news.software.b, article <3ND5KOC@xds13.ferranti.com>, peter@ficc.ferranti.com (Peter da Silva) writes: < In article <1990Aug21.161506.21784@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes: < > Somehow I doubt this; B News didn't exactly hold your hand every step of < > the way either. < < It did a better job than C news. Really. C news is a far superior product < once you get it installed, but it's a pain to install and there are all < sorts of administration gotchas. [...] It starts when you try to compile the sources as Joe User, but install them under "news". This process is straightforward, _but_ first you take the last part of doit.bin and tack it in front of doit.news.... < > >Huh? Where is there a UNIX system where "chown $NEWSMASTER $NEWSCTL/active" < > >won't work? ... < > Systems where chown is in some strange place, like /etc, that isn't in the < > default search path. < Of course, the "build" script does ask where chown is, so you do know the < strange place when you install the scripts... Just watch someone install first perl and then C News. People tend to look at me somewhat incredulously at these times. Larrys config stuff finds out all about the machine you don't want to know anyway, and C news asks you. Granted that configure doesn't work everywhere, but in the majority of standard cases it's great. But I seem to remember that we had this discussion a few months ago... -- Matthias Urlichs -- urlichs@smurf.sub.org -- urlichs@smurf.ira.uka.de Humboldtstrasse 7 - 7500 Karlsruhe 1 - FRG -- +49+721+621127(Voice)/621227(PEP)
henry@zoo.toronto.edu (Henry Spencer) (08/24/90)
In article <5fq0f2.zm6@smurf.sub.org> urlichs@smurf.sub.org (Matthias Urlichs) writes: >Larrys config stuff finds out all about the machine you don't want to know >anyway, and C news asks you... We find it very interesting that we get mail occasionally saying "thank god you didn't use Configure!". All that automatic guesswork is not infrequently wrong, and is *invariably* wrong if you are trying to configure for a machine other than the one you are running on... which is increasingly common in these network-oriented days. You might want to make a list of the things "build" asks you, and figure out which ones could be guessed automatically. Not very many. -- Committees do harm merely by existing. | Henry Spencer at U of Toronto Zoology -Freeman Dyson | henry@zoo.toronto.edu utzoo!henry
davecb@yunexus.YorkU.CA (David Collier-Brown) (08/26/90)
urlichs@smurf.sub.org (Matthias Urlichs) writes: | Just watch someone install first perl and then C News. | People tend to look at me somewhat incredulously at these times. | Larrys config stuff finds out all about the machine you don't want to know | anyway, and C news asks you. Granted that configure doesn't work everywhere, | but in the majority of standard cases it's great. Alas, the common case isn't the general case. **However**, if the common case is common enough, mabe its worthwhile to add config to the "contributed software" part of the distribution, assuming circumstances permit. --dave (of course, it might then become infinitely large...) c-b -- David Collier-Brown, | davecb@Nexus.YorkU.CA, ...!yunexus!davecb or 72 Abitibi Ave., | {toronto area...}lethe!dave Willowdale, Ontario, | "And the next 8 man-months came up like CANADA. 416-223-8968 | thunder across the bay" --david kipling
richard@pegasus.com (Richard Foulk) (08/26/90)
>>Larrys config stuff finds out all about the machine you don't want to know >>anyway, and C news asks you... > >We find it very interesting that we get mail occasionally saying "thank god >you didn't use Configure!". All that automatic guesswork is not infrequently >wrong, and is *invariably* wrong if you are trying to configure for a machine >other than the one you are running on... which is increasingly common in >these network-oriented days. As with anything, experiences and tastes vary. The question is how many people would prefer not to have something like Configure -- I suspect it is a very small minority. And the number of people cross-compiling Cnews has got to be miniscule. Configure handles configuring for a different machine (but with the same OS and CPU types) quite well. >You might want to make a list of the things "build" asks you, and figure >out which ones could be guessed automatically. Not very many. At least half. -- Richard Foulk richard@pegasus.com
karl@MorningStar.Com (Karl Fox) (08/27/90)
Since cnews now remembers the answers to previous runs of build's questions, it would be a relatively simple job to create a separate shell script called something like "build-defaults" that examined the system and created an initial set of default answers. Looking in the C library for strchr, symlink, etc. and looking for chown and chgrp are pretty simple, and they don't have to be 100% accurate, since they are only *defaults*. Plus, you wouldn't have to run it if you were politically opposed to Configure. -- "I hear you guys deal with such dreck | Karl Fox, Morning Star Technologies as SNA and X.25." -Ed Vielmetti | karl@MorningStar.Com
vixie@decwrl.dec.com (Paul A Vixie) (08/27/90)
[Guy Middleton] >> This is exactly why I don't like Configure. I have ten different >> architecture [sic] here, and a canned set of Build-answers for each one. >> I can maintain everything from a single machine, instead of running >> Configure ten times every time I need to recompile. I do this by saving aside the .sh and .h files that "Configure" builds. Granted, Configure would be a lot better if it supported multiple architectures -- but then, so would Make. There have been a lot of attempted solutions (VPATH comes to mind), none of which solves the problem as well as handcrafted and handmaintained symlinks do. Don't blame Configure for the world's ills. -- Paul Vixie DEC Western Research Lab <vixie@wrl.dec.com> Palo Alto, California ...!decwrl!vixie
gamiddle@maytag.waterloo.edu (Guy Middleton) (08/27/90)
In article <5fq0f2.zm6@smurf.sub.org> urlichs@smurf.sub.org (Matthias Urlichs) writes: > Larrys config stuff finds out all about the machine you don't want to know > anyway, and C news asks you. Granted that configure doesn't work everywhere, > but in the majority of standard cases it's great. This is exactly why I don't like Configure. I have ten different architecture here, and a canned set of Build-answers for each one. I can maintain everything from a single machine, instead of running Configure ten times every time I need to recompile.
moraes@cs.toronto.edu (Mark Moraes) (08/27/90)
richard@pegasus.com (Richard Foulk) writes: >As with anything, experiences and tastes vary. The question is how many >people would prefer not to have something like Configure -- I suspect >it is a very small minority. The question is whether Henry and Geoff want to use something like Configure. Configure assumes you haven't touched your system since it was powered on. It's really hard to convince Configure to stop outguessing you. For example, some of our SysV machines have been convinced to see the error of some of their ways. Configure gets really upset on such machines... Some of us don't like "build" much either, but at least "build" doesn't make any explicit assumptions -- it has the courtesy of asking you what it should do. That's A Good Thing, even if the affectionate nickname for this process is "The Interrogation". Time to move on to the discussion about why Henry and Geoff shouldn't use imake... :-) Mark
tneff@bfmny0.BFM.COM (Tom Neff) (08/27/90)
In article <90Aug26.184739edt.417@smoke.cs.toronto.edu> moraes@cs.toronto.edu (Mark Moraes) writes: >Configure assumes you haven't touched your system since it was powered >on. It's really hard to convince Configure to stop outguessing you. No it's not - let it finish and edit the .sh files to suit. Configure invites you to do it. If Configure has a logic problem on some architecture, mail the patch to Larry. -- Cogito ergo spud. I think, O OO O Tom Neff therefore I yam. -- anon O OO OO tneff@bfmny0.BFM.COM
peter@ficc.ferranti.com (Peter da Silva) (08/27/90)
In article <1990Aug26.195308.11696@maytag.waterloo.edu> gamiddle@maytag.waterloo.edu (Guy Middleton) writes: > This is exactly why I don't like Configure. I have ten different architecture > here, and a canned set of Build-answers for each one. I can maintain > everything from a single machine, instead of running Configure ten times every > time I need to recompile. But you can save your Configure answers, too! Better than Build, even, since you have to run Build to rebuild those doit files... I tried making a non- interactive version of build, but it's got too much non-saved state. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
henry@zoo.toronto.edu (Henry Spencer) (08/27/90)
In article <1990Aug26.085636.19024@pegasus.com> richard@pegasus.com (Richard Foulk) writes: >>You might want to make a list of the things "build" asks you, and figure >>out which ones could be guessed automatically. Not very many. > >At least half. I make it 24 out of 78, even being generous and assuming doing some automatic experiments which could possibly be dangerous. And some of those would require confirmation anyway, and the guessing code would fail on some machines. Simply asking a single question for all the str/mem functions -- a change I am considering -- would cut out fully half of those guessable answers. -- Committees do harm merely by existing. | Henry Spencer at U of Toronto Zoology -Freeman Dyson | henry@zoo.toronto.edu utzoo!henry
cudep@warwick.ac.uk (Ian Dickinson) (08/30/90)
In article <1990Aug24.165844.3563@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >We find it very interesting that we get mail occasionally saying "thank god >you didn't use Configure!". All that automatic guesswork is not infrequently >wrong, and is *invariably* wrong if you are trying to configure for a machine >other than the one you are running on... which is increasingly common in >these network-oriented days. All very well and good - but why isn't the actual build seperated from installation a bit more? This is the one real nasty thing about building Cnews. Not everyone likes building software as a (semi)privilidged user (ie. news) and I'd like to build it as myself, and THEN install as root and news. This is not as easy as it should be (IMHO). As far as Configure goes - I've got used to its mistakes, but it's not that easy for a beginner who doesn't know his machine too well. Cheers, -- \/ato. Ian Dickinson. GNU's not got BSE. Cut Cerebus some slack! vato@cu.warwick.ac.uk Plinth. vato@tardis.cs.ed.ac.uk Sabeq. gdd046@cck.cov.ac.uk "Nuke me tender, nuke me good!"
mjr@decuac.DEC.COM (Marcus J. Ranum) (08/31/90)
In article <1990Aug30.164310.8647@warwick.ac.uk>, cudep@warwick.ac.uk (Ian Dickinson) writes: > Not everyone likes building software as a (semi)privilidged user (ie. news) > and I'd like to build it as myself, and THEN install as root and news. I've taken to doing the configuring and building as myself, and when I get around to installing it, I'll just chown the whole subdir tree to news and work from there. mjr. -- Eagles may soar, free and proud, but weasles never get sucked into jet engines.
louie@sayshell.umd.edu (Louis A. Mamakos) (08/31/90)
Not to pick on anyone in particular, but what the $%*&! do you want for free? It is certainly easy to suggest that C news be (or the installation scripts) be modified/"fixed" to do things the "right" way, but those that "do" decide "how." Distributing software which attempts to run on a wide variety of platforms is a difficult task; each of us has their own idea of what the "best" way is. louie P.S. Count your blessings that C news doesn't try to convert unsigned longs to/from doubles! My NTP daemon had to be shipped with a compiler test for that case since the majority of platforms got it wrong! It's a cold, hard world out there..