Hi, I converted fcdsl from kmod to kernel-mod. I also made som other changes, for example I removed the /etc/capi.conf file as we should not won the same file with capi4k-utils and tell the user in pkg_postinst to change it instead. I removed ppp from DEPEND because its not needed for fcdsl to compile. I tested this on my system but I have no fcdsl, so unforunately I cant test the detection code.
Created attachment 44311 [details] fcdsl-2.6.20.7.ebuild
Created attachment 44312 [details, diff] fcdsl.diff I think we should add these changes as a -r1 to give the users a chance to try them out.
unfortunatelly, I couldn't test this either. I'll test as ebuild and submit as ~x86 in about 12 hours, letting the ~x86 people test it for us.
submitted to the tree, closing this. Stefan, tx a lot for your help.
Well the new ebuild is better than the first one, but the install of single builded detected module is broken, now. Fixing it right now. Patch is comming.
Created attachment 44390 [details, diff] Patch for detect card This patch solving the problem that the firmware of the detected card isn't installed. Furthermore changing "unset ARCH" to "set_arch_to_kernel" and make to "kernel-mod_src_compile" Seems now working.
Well, something... Without ppp a user couldn't make an internet connection. Therefore it should be in RDEPEND?
tuxxer: what do you think, should we own /etc/capi.conf together with capi4k-utils, or should we tell the user to change the file himself?
Also, why did you change the kernel stuff? Its not better the way you are doing it, its more code and we plan to soon change the kernel-mod.eclass and in general the kernel eclass layout, where this maybe could break then.
The fcdsl also has a capi-isdn-interface, right? So a user may only want to fax, therefore I took ppp out of the dependancies. And it is not really needed for the package to work, is it?
*** Bug 71969 has been marked as a duplicate of this bug. ***
patch partially applied because Stefan and I believe it is best to install all modules, so we removed card detection from src_compile. tuxxer, please test the new -r1.
To 1.) "...what do you think, should we own /etc/capi.conf together with capi4k-utils, or should we tell the user to change the file himself?" => I think it is the better way to serve it with capi4k-utils To 2.) "...Also, why did you change the kernel stuff? Its not better the way you are doing it, its more code and we plan to soon change the kernel-mod.eclass and in general the kernel eclass layout, where this maybe could break then." => Sorry, I'm neither a gentoo developer nor firm with eclass internas of kernel-mod. I'm only try to share my drivers - which I need - with all community members. I included it, because much ebuilds do it that way. To 3.) "...The fcdsl also has a capi-isdn-interface, right? So a user may only want to fax, therefore I took ppp out of the dependancies. And it is not really needed for the package to work, is it?" => Well, buying an ISDN and ADSL card for fax only, in my opinion, make no sense. The main target is to get an online connection and therefore you need ppp. It makes it not easier for an user when he could fax but can't go into I-Net and have no clue about installing ppp for that. To 4.) "...patch partially applied because Stefan and I believe it is best to install all modules, so we removed card detection from src_compile" => I think this is one way to do it, but I don't think all users want to have all firmwares (5 cards) on their HD when they owns only one. Then you could erase the whole detection stuff, too. A detection section, when detection is done, make no sense, in my opinion.
tuxxer, detection is used in config part of the ebuild which is the normal way to do it. you should read messages displayed in pkg_postinst more carefully. about instalation of all modules, I don't hear anyone complaining that his/her kernel comes with tons of irrelevant (from their point of view) modules. a driver should support a wide range of devices. pls test the new ebuild for allowing me to close this bug.
Well, I'm a little confused... There are many ebuild which change it behavior for removing not needed or old parts (module-init-tools => IUSE="no-old-linux", glibc => IUSE="userlocales") And there a driver packages where you can choose the needed drivers (alsa-driver => ALSA_CARDS="...", libgphoto2 etc.) All these changes were made by user wishes, because these users didn't wanted that unneeded parts would be installed. Furthermore I don't want to install all five drivers on my HD, because I have only one ADSL card, not five. I think I remove that package from my PC and split out the drivers for my card into a separate ebuild for my own. That is my personal opinion and isn't personal. Please can someone other please test that ebuild? Thanks.
of course, we could add something like a user-configurable var $FCDSL_CARDS, just we would not use detection to install only the detected card's module.
Well, this seems to be a adequate compromise. Did we the detection then anymore?
yes, we need for the part of the crowd who won't use this var. detection is used in pkg_config for editing /etc/capi.conf and tc/modules.d/fcdsl. I will modify the -r1 ebuild tonight.
now -r1 has support for env. variable FCDSL_CARDS. you could put a series of module names in it which will force ebuild to install only these modules. please test it and report the results.
Well, anything works smoothly, now. Without FCDSL_CARDS all drivers are build, when I'm using FCDSL_CARDS="fcdsl" only the driver for my AVM FRITZ!Card DSL is builded. Well done!
closing this... thanks for your support.
