[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#23143: make-kpkg fails on Alpha



Hi,
>>"Christopher" == Christopher C Chimelis <chris@beezer.med.miami.edu> writes:

 Christopher> On 4 Jun 1998, Manoj Srivastava wrote:

 Christopher> I can describe a possible solution, if you're willing at
 Christopher> this point to put time into it.  I can also put time
 Christopher> into testing whatever changes you make prior to making
 Christopher> an "official" modification to the package (and
 Christopher> hopefully, by that time, I can hold off RL enough to
 Christopher> patch some myself).

	I have been graciously offered access to an alpha box; I'll
 see what I can do about modifying the package so that people can
 actually test the modifications (I think testing a new kernel without
 access to the console is tempting fate)


 Christopher> FYI, MILO is much different than LILO and would require
 Christopher> some user intervention at reboot time to enable a new
 Christopher> kernel unless it was named the same as the old one with
 Christopher> the exact same location.  MILO is more like a bootloader
 Christopher> that also loads PALcode than a boot-block-based boot
 Christopher> manager.

	I see. Well, we can use something that is done for boot
 loaders on other architectures; assuming that MILO does not choke on
 symbolic links. On other architectues, kernel-image makes /vmlinuz a
 symlink to the latest kernel image, and /vmlinuz.old is a link to the
 previous kernel image. After the symlinks are changed, LILO/SILO are
 rerun.

 Christopher> FYI, Alpha kernels are generated as gzipped files, but
 Christopher> are named vmlinux.gz for some odd reason.  They are
 Christopher> located in arch/alpha/boot (same as other arch's
 Christopher> basically).  MILO doesn't care about the file name as
 Christopher> long as it can find the file itself (it still has
 Christopher> problems following links, unfortunately).

	If MILO does not care about file names, but it has a problem
 with symbolic links, I can arrange for hard links to be created, I
 think. Hmm. abstractions, abstractions.

 Christopher> Here's an albeit brief solution for make-kpkg:

 Christopher> 1. Find a way to check the build architecture (dpkg
 Christopher>    --print-architecture works, but may be inefficient to
 Christopher>    call.  It does return just "alpha", though, so if
 Christopher>    that helps...if you need more info, let me know and
 Christopher>    I'll find a command to get it for you)

	Well, since this is just done once during a kernel compile, I
 don't think the inefficiency shall be noticeable ;-). make-kpkg
 already calls this, I just wasn't sure what it rteturns on an
 alpha. This is good. I can now separate stuff for the alpha based on
 this. 

	I am not using anything as sophisticated as a lookup table in
 the rules file, I'll just special case the alpha as requiring
 "boot". This is easy. 

 Christopher> There are other problems with building a kernel on the
 Christopher> Alphas, but they relate more to the existing
 Christopher> kernel-source packages and not with make-kpkg

	This is good.


 Christopher> I'll retest symlinks with MILO hopefully this week (I
 Christopher> need to resolve that issue for myself as well, so I
 Christopher> would like to know for sure if improvements have been
 Christopher> made since last year about this).  Either way, links may
 Christopher> or may not be important when it comes to Alphas since
 Christopher> MILO will let you boot different images without needing
 Christopher> the same filename.  I could point you towards some MILO
 Christopher> docs if you'd like, but be warned that they're sketchy
 Christopher> at times.  Also, without console access to an Alpha, you
 Christopher> would never be able to see MILO in action to see what I
 Christopher> am describing, unfortunately.

	Well, let me know. Would be nice if symlinks were resolved,
 would save me some coding. In the meanwhile, I can make sure the
 correct commands are generated for the alpha, and that we move the
 correct file into /boot; it shall be called /boot/vmlinuz-VERSION for
 consistency.

	That would mean that the package would work for people, the
 only issue remaining whether MILO can be fed a consistent file name
 so people do not have to manually remember to change stuff. 

	Once we have stuff tested, we can see about getting it into
 Hamm (we may just need to have different versions in hamm for the
 alpha vs other architectures if it proves hard to get the new
 versions into the freeze, but that is something we can tackle when we
 get this working).

	manoj
 ps: I am not subscribed to the alpha list, so please CC me any
 responses you wish me to read ;-)
-- 
 ...difference of opinion is advantageious in religion.  The several
 sects perform the office of a common censor morum over each other.
 Is uniformity attainable?  Millions of innocent men, women, and
 children, since the introduction of Christianity, have been burnt,
 tortured, fined, imprisoned; yet we have not advanced one inch
 towards uniformity. Thomas Jefferson, "Notes on Virginia"
Manoj Srivastava  <srivasta@acm.org> <http://www.datasync.com/%7Esrivasta/>
Key C7261095 fingerprint = CB D9 F4 12 68 07 E4 05  CC 2D 27 12 1D F5 E8 6E


--
To UNSUBSCRIBE, email to debian-alpha-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org


Reply to: