Andreas Hanke wrote: > Hi, > > Rajko M schrieb: >> That is reason to ask for gcc. I know how to handle this, but new to >> Linux will get lost, and we will have to handle a lot of help requests >> from them. > > PLEASE. > > The web is full of > > - forum posts > - wiki entries > - other documentation resources > > which clearly say: > > "In order to compile a kernel module, the packages gcc, make and > kernel-source are needed." > > The user will "get lost" only if he's not interested in getting the job > done anyway.
I remember my start with Linux. I had strong interest to succeed, but many times it looked to me like a lost case. 1) Documentation in many fields obsolete, insufficient or missing, 2) Books, not many and obsolete even fresh out of the print, often artificially oversized, as number of the pages makes them look serious and sell better. 3) Altavista (at the time top search engine) not very helpful, as one has to know right words. Synonyms will lead in wrong direction, and I knew only terms used in other OS. 4) Linux is different enough, that more you know about another option, more problems you have to overcome. So, it is not always user interest (lack of) to make people feel lost. >> Another option is weak dependency, as Christian mentioned, but I haven't >> seen recently packages that will be installed to satisfy dependencies >> and have check box enabled, which would be sign that I can remove them >> from selection. > > That's because "weak" dependencies are not weak at all. > > There are just two cases: > > - The package that another package depends on is installed by default. > This is the case with "Requires" and "Recommends". "Requires" == > "Recommends" here. This is probably because how it is treated in package management software ie. "Recommends" will be marked for installation without any additional question or special marks that you can skip it loosing some functionality. > - The package that another package depends on is not installed by > default. This is the case with "Suggests". "Suggests" is very similar to > not having a dependency at all. > > There is no intermediate step between these and no way to prevent a > "Recommended" package from being installed other than installing it and > uninstalling it again or setting the package to "taboo" - which requires > the user to know in advance that the dependency is meant to be "weak". Taboo option is something that many users learn long time after they start to use Software management. The same is with option to remove software after installation, besides that it includes 2 operations that should not happen at all. > I can explain why I don't like the idea of kernel-source depending on > gcc: It breaks the concept of being able to replace the system compiler > that has just now been introduced, because it will force the system > compiler to be installed even if the user already has another compiler. > > On the other hand, you might argue that kernel modules must be built > with the system compiler anyway... And that's correct. So maybe the > proposal makes sense, but I'm unsure. I'm a bit puzzled. If different compiler gives correct code for another applications, what makes it unable to compile kernel modules? Missing some special kernel module API, or sort of it. Sorry, but non-programmers need few words more. If this, about system compiler, is hard fact, than there is nothing to be unsure about. I don't like few hundreds MB more, just to be able to run video adapter in full featured mode, but if I have to, then there is no options. I either have kernel sources and gcc installed, and used, or I have no 3D, and for instance Google Earth runs skipping the frames (nvidia with nv driver) or not at all (old ati board with ati driver). -- Regards, Rajko. Visit http://en.opensuse.org/MiniSUSE --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
