Re: kernel headers
Hi,
>>"Avery" == Avery Pennarun <apenwarr@worldvisions.ca> writes:
Avery> On 17 Jan 1998, Manoj Srivastava wrote:
Avery> The FSSTND, by design, says nothing about local modifications.
Avery> The accepted place for source code to go is /usr/src; I believe
Avery> that we should understand and accept that this practice is not
Avery> going to change in the near future.
Au contraire; in me decade long experience as a UNIX sysadmin
the accepted place for source to go is in /usr/local/src (most GNU
software builds in /usr/local by default). I do not recall seeing
/usr/src on Ultrix, Digital Unix, HP-UX, AIX, or on old SUN
machines.
Since it has never been standard unless on home brewed Linux
Boxes (slackware or SLS [any one still remember SLS?]), I think it is
quite easy to teach people to stick to /usr/local/src.
Avery> Sigh. The above statement was a generalization meant to apply
Avery> if it turns out that I was breaking Debian policy. It is still
Avery> my opinion that installing a local kernel source tree in
Avery> /usr/src/linux is _not_ breaking Debian policy.
I beg to disagree.
Avery> However, my point in the quote above was that _if_ I were
Avery> breaking Debian policy by extracting the kernel into
Avery> /usr/src/linux, Debian should quietly recover. As things are,
Avery> it loudly mangles things.
No, it complains loudly, as it should. Someone walked into
it's bailiwick, and it has a right to complain. Remove the offending
directory, and the install shall proceed gracefully.
Avery> I maintain that it would be _EASY_ for Debian to quietly
Avery> recover from this situation: "Just because it's not officially
Avery> guaranteed to work" (kernel-headers thinks it should own
Avery> /usr/src/linux) "doesn't mean it's not allowed to work."
Avery> (kernel-headers should accept that it can't own /usr/src/linux,
Avery> because that doesn't hurt anything).
I guess kernel-headers is just as stubborn as you seem to be ;-)
Avery> To put it bluntly, we are testing for a stupid invariant.
Avery> There is only one reason that kernel-headers does not work
Avery> perfectly when /usr/src/linux already exists: kernel-headers
Avery> checks to make sure /usr/src/linux does not already exist.
Avery> That's equivalent to writing the following in the postinst for
Avery> pppd:
Avery> if (cos(3.1415926535) != -1) return ERROR; else continue
Avery> installation
Avery> True, cos(PI) should always be -1 (barring floating point
Avery> inaccuracy), but whether this is true has absolutely no bearing
Avery> on whether pppd will work after it's installed.
On this we again disagree. I think it is relevant. Kernel
headers has always manipulated /usr/src/linux, and any machine where
this is not possible is non standard, Lord only knows what else may
be messed up. I do not think that quietly ignoring errors is correct;
I prefer to complain loudly.
Avery> If the user complies with FSSTND _and_ has a /usr/src/linux
Avery> containing valid header files (the only two parts of Debian
Avery> policy which you have quoted me), then the user complies with
Avery> Debian policy. If the user complies with Debian policy and has
Avery> package installation failures, then something is seriously
Avery> wrong with the policy or the packages.
The user does not need comply with the FSSTND for the
directories under the Vendors control; Debian does a good job. The
user should stay out of vendor space.
Avery> I am asking for the simplest of simple changes to the
Avery> kernel-headers / kernel-source install script: do not return a
Avery> failure code to dpkg if the symlink to /usr/src/linux fails.
Avery> That one change eliminates the whole problem we're arguing
Avery> about, while causing no other problems.
I shall consider this, but I am not at the moment hopeful that
I shall accede to this request (in fact, I don't think I should, on
principle, and the FSSTND). I do apologize for my anal retentiveness.
manoj
--
"I believe in a God which doesn't need heavy financing." Fletch
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
Reply to: