Bug 6905
Summary: | drivers/md/dm.h should have to be in include | ||
---|---|---|---|
Product: | Drivers | Reporter: | Levente Farkas (lfarkas) |
Component: | Other | Assignee: | drivers_other |
Status: | REJECTED INVALID | ||
Severity: | normal | CC: | protasnb |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.17 | Subsystem: | |
Regression: | --- | Bisected commit-id: |
Description
Levente Farkas
2006-07-26 00:36:28 UTC
1) this sounds like an out-of-tree issue, which this BZ is not used for. 2) this sounds like an issue for the distros. kernel.org is not responsible for any headers packages (yet, there is some effort to that end in -mm, IIRC). Rejecting. at the same time redhat/fedora kernel package maintaner (dave jones) reject this request since he said it's a upstream bug: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197645 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197494 they assume kernel headers under the include/ directory... it's ridiculous to expect distros to move files around the tree. That way just leads to mass confusion and "this builds on distro X but not distro Y" situations. It also doesn't scale. This is just but one example, imagine if every subsystem had the same problem ? In a time when we're trying to get distro kernels to match upstream kernels closer, recommending deviation is absurd. If that file contains valid things for out-of-tree stuff to compile against, it belongs in include/. FWIW, the only file in truecrypt using dm.h is Linux/Kernel/Dm-target.c which wants: #define DM_NAME "device-mapper" #define DMERR(f, arg...) printk(KERN_ERR DM_NAME ": " DM_MSG_PREFIX ": " f "\n", ## arg) struct dm_dev (and dereferences up to ->bdev->bd_dev) What is the verdict on this problem? and what is exactly wrong with moving/adding the header file to include, does it have some negative impact? Thanks. |