> how true. i thought i'd just glom 2.3.42 and i'd not > have to reinstall the 3.1 pcmcia stuff - what optimism. :) If you are not so interested in helping debug the PCMCIA code in the kernel, you can try the latest beta PCMCIA package. I made a few changes so that it should build correctly on a 2.3.42 or later kernel with PCMCIA completely turned off in the kernel. I've only tested with a couple cards but it seems to all work correctly. My plan is to gradually migrate the standalone package to use the 2.3.* interfaces, and hopefully gradually converge with the 2.3.* kernel PCMCIA stuff, and shake out the bugs in the process. I think the changes Linus made in the 2.3.* tree were too much all at once, and the 2.3.* tree is so broken that debugging is almost impossible. > undoubtedly. one of the biggest problems i noticed was > simply that of confusion. This is an unfortunate feature of how kernel configuration is done; PCMCIA driver options are not grouped together, just as PCI drivers are not grouped together. This is partly why for the standalone PCMCIA package, I just bag it and require that you build everything whether you need it or not: building 20 or so driver modules only takes a couple minutes, doesn't have any run-time downside, and avoids a lot of configuration headaches. > ok. is there a convention to signify which is stable, or is > there simply an accepted last-known-stable version? There is a convention that when I bump the first or middle number (like 2.8.23 to 3.0.0, or 3.0.14 to 3.1.0), that's a bigger change, and you should expect ".m" versions to be less stable than ".n" where m is less than n. That isn't always true, but that's my goal. I post experimental snapshots to ftp://sourceforge.org/pcmcia/NEW and these are identified by date of release. -- Dave |