On Sun, Aug 29, 2004 at 03:55:17PM +0200, Christoph Hellwig wrote: > I just talked to Jens Axboe about all cdrom failures we see, and he > asked the right questions. Except on powerpc, hppa and the alpha > 'default' kernel we have CONFIG_IDE_TASKFILE_IO turned on, which is an > alternate, experimental and much less tested codepath through the ide > subsystem. We should turn if off for the next kernel-image uploads.
We should also commonize some the config file generation, since when generating a new config file for a new kernel, it is not always evident to know which are the stuff to enable and which one, and it then results in situations as the above, where some chose one direction, and others chose the other. Ideally we would have some tool that processes all the Kconfig options, and extract those that are not arch dependent, and put them in a common file, and the arch packages will use only the real arch dependent entries in their package, with maybe a mechanism to override those entries in the common file that needs overriding for a given arch/subarch. In the meantime, a reference config file arch packagers can base upon to generate the per arch config files would already be a good idea. Friendly, Sven Luther
