Why was this moved to fink-users? I am moving it back now.
At 9:06 Uhr +0100 19.01.2002, Martin Costabel wrote: >Alexander Strange wrote: >> >> Hey, this is a perfect example of why we should have this new .info >> format :) > >I rather think your answer an example for some misunderstanding that I >can see in this whole excited discussion of new info file formats and >package options: > >When Jeff says that he doesn't have the time to maintain all the >possible options, then a new format would NOT help him. The problem is >not that the "old" format requires you to have several info files, one >for each option, and the "new" format would allow you to create all the >options from just one info file. The problem is that you (the >maintainer) have to maintain ALL the offered options. You have to >configure them all and patch them all and compile them all and test them >all, and this is completely independent of the format of the info file. If you look at existing package variants, there are few cases that actually require different patches. Most consist of adding/removing lines to the configureparams & the depends line. But you are right of course, the various combinations of a package have to be tested, which certainly increases the work load (for example, it multiples the compile time, since each variant would need to be built). However, if an author doesn't have the time for this, well, he can do several simple things: 1) Just don't offer the variants. After all, this is meant to be an added possibility, not a requirement 2) Only test the "core variants", marke the other variants/combos as untested, and let others do the testing To apply this to the example in question, pine vs. pine-ssl: the patch is identical. The differences are only in the PatchScript and the CompileScript. AFAICT, it would be trivial to turn this into a single package with two varianst, one for ssl, one for ldap. This new package could have 4 instead of 2 different results produced from. Now the maintainer (Jeff in this case, but this holds true for anybody, me included) has double work in terms of compilation time in the least... so what we could do is to add a list of "tested" combos, and only these are allowed... but if somebody has the time to test the others, this list could then be extended. > >I think in comparison of the work involved with porting and maintaining >a package, the question whether you have one or several info files is a >minor detail. Creating several info files that differ only by a few >lines is a trivial thing that can be done instantly with any text >editor. Well, with any text editor, as long as they remain simple text >files, anyway. I disagree. Keeping the files in sync is not a trivial work, and indeed, authors failed to always do it cleanly in the past. Also, when the variants where created, it was very common to forget things, like adding Conflicts/Replaces/Provides fields. So in any case there is some saving involved, and it's not completly minor. The main reason there is no xchat-nox and xchat-nognome package is BTW exactly the fact that it is super-tedious to keep 6(!) .info files in sync. I wouldn't mind the compile time, actually, my machine can do that over night. To sum it: a new package file format is certainls *not* a magic bullet. But it *does* solve/simplify various issues, and maybe more importantly, creates new possibilities. Max -- ----------------------------------------------- Max Horn Software Developer email: <mailto:[EMAIL PROTECTED]> phone: (+49) 6151-494890 _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
