GOTO Masanori <gotom@debian.or.jp> writes:
> At Mon, 03 Nov 2003 13:13:11 +0900,
> GOTO Masanori wrote:
>>
>> At Sun, 02 Nov 2003 11:58:16 -0500,
>> Andr�Rold�wrote:
>> > I got the following compiling lilo (This seems to be a gcc problem also):
>> >
>> > geometry.c: En la función `lvm_bmap':
>> > geometry.c:438: error: subscripted value is neither array nor pointer
>> > geometry.c:450: error: subscripted value is neither array nor pointer
>> >
>> > The relevant lines are:
>> >
>> > unsigned short iop;
>> > static int lvmfd;
>> > struct lv_bmap *lbm;
>> > ...
>> > line 438: if (ioctl(lvmfd, LVM_GET_IOP_VERSION, &iop) < 0)
>> >
>> > line 450: if (ioctl(lvmfd, LV_BMAP, lbm) < 0) {
>>
>> It described at include/linux/lvm.h in 2.4, but there is no such file
>> in 2.6. LVM is completely replaced to DM. In another word, lilo
>> can't compile under 2.6 kernel, and it's not ready for LVM. This is
>> bad.
>>
>> At least software like lilo should fix for 2.6 kernel. It should
>> handle both 2.4 and 2.6. This is lilo bug. I would like to reassign
>> to lilo.
>>
>> However, this bug is caused by linux-kernel-headers. One idea is to
>> provide linux-kernel-headers-2.4? Or lilo should include kernel
>> headers? I don't think the latter is bad because each package
>> maintains kernel headers. It can absorb the kernel version difference
>> for each package, but each time we need to track kernel and check kernel
>> headers.
>>
>> We need to discuss about this kind of area. Any suggestions?
>
> I discussed with Daniel, and he suggested to use kernel-headers-2.4
> package, or copy header constants to the file. The latter can be
> easily fixed for this case.
I think that putting those constants into the file is not the best
solution. See below.
>
>
> Back to the idea to use kernel-headers-2.4 package.
> linux-kernel-headers package is now 2.6 based, and it governs under
> /usr/include/{asm,linux,asm-generic}. This is not changed. So lilo
> can be fixed using kernel-headers-2.4 package.
>
> However the problem is for example kernel-headers-2.4.21-4 package has
> "/usr/src/kernel-headers-2.4.21-4" dir. This means that lilo package
> can't use kernel-headers-2.4 package easily, and lilo package links to
> this kenrel-header package strongly. kernel-headers package is
> modified sometimes frequently, so many packages link to different
> kernel-headers package. The one idea is that kernel-headers-2.4.*
> packages provide /usr/src/kernel-headers-2.4 symbolic links. We need
> to think about the case: multiple dirs are existed (like 2.4.21-4,
> 2.4.22-1, ...), though. But this idea leaves some
> linux-kernel-headers roles, and I don't think using kernel-headers is
> always good.
I've already sent an e-mail to John Coffman (LILO upstream) and asked him
about the current status of LILO having in mind that (very) soon the 2.6
kernel series will be the stable branch of linux.
I think that making a symbolic link from /usr/src/kernel-headers-2.4* to
/usr/src/kernel-headers-2.4 is a good idea but this have to be debated.
In the meantime, I'll try to fix this by adding those constants to the
needed LILO header file and I will wait for the upstream answer about this
topic.
Thanks.
>
> Any suggestions are welcome.
>
> Regards,
> -- gotom
>
>
> --
> To UNSUBSCRIBE, email to debian-glibc-request@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
>
--
Andres Roldan
Fluidsignal Group <aroldan@fluidsignal.com>
The Debian Project <aroldan@debian.org>
GPG Key-ID: 0xB29396EB
Home Page http://people.fluidsignal.com/~aroldan
Attachment:
pgphyyZ1T6om7.pgp
Description: PGP signature