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

Re: Bug#147077: Please fix make-kpkg to build packages with proper conflicts



>>"Brian" == Brian Mays <brian@debian.org> writes:


 Brian> Some people are successfully using systems with both the built
 Brian> in kernel pcmcia modules and the stand-alone pcmcia modules
 Brian> installed.  (I have done it myself.  It works in all but the
 Brian> most unusual cases.)  Since you are advocating supporting
 Brian> obsolete (possibly broken) configurations, such as the one you
 Brian> cite above, do you advocate allowing both sets of modules to
 Brian> be installed at the same time?

	I would defer to your experience here; I am not personally
 using the implementation cited above, but one person who is wrote me
 in private email. I would not be against stating the setup that needs
 the symlink when using the builtin kernel pcmcia modules is obsolete.

	However, since we support people downloading the kernel
 sources from kernel.org directly, whether the symlinks are good or
 not is moot; upstream still has these links, and we must deal with
 them. 

	I am more concerned, though, about having modules with the
 same name in different part of the lib/modules/$(uname -r) hierarchy,
 perhaps because I do not know what the ordering of the search paths
 are. Personally, I would not recommend installing both sets of
 modules besides each other; I would say that for most novices, this
 is likely to be something that was unintended.

	Given that, I would not prohibit contemporaneous installation
 of both sets of pcmcia modules, but I would strongly advocate a
 diagnostic being issued, and perhaps even ask the person installing
 if this is what the intend.

 >> h) I would much rather not go down the path of looking into .config
 >> file to check for particular modules, make separate image files
 >> for these modules on the fly, and if these modules have been
 >> compiled in do other things, special casing each module as
 >> needed. 

 Brian> Then what is your opinion about the separate
 Brian> kernel-pcmcia-modules-2.4.18-xxx packages that Herbert has
 Brian> been producing?  He has already implemented one of these
 Brian> "special casing" scenarios for splitting out the pcmcia
 Brian> modules from the kernel-image-2.4.18 packages.

	I am not sure if my opinion has much meaning. I certainly am
 not a fan of the gazillion kernel image packages for the same
 architecture, complete with the corresponding header, doc, and source
 binary packages; I have never been convinced of the ROI (if I may
 stretch the term) or compiling for k6/k7/i386/i586/i686/i686smp
 separately justifies the utility for the users. But, as author of
 kernel-package, I may be biased about the simplicity of and rationale
 for everyone compiling their own, tailor made, kernel.

	I am similarly not a fan of splitting out the modules from
 the kernel image packages, but it all boils down to whether one
 should expect users to do kernel compiles.

 Brian> I think that users will find it surprising that the official
 Brian> Debian kernel packages are structured somewhat differently (in
 Brian> that some modules are split into their own package) than the
 Brian> kernel-image packages that they produce themselves.  Don't you
 Brian> agree?

	I do think that this is not a good scenario, but this seems to
 be a fait accompli.


 >> b) The preinst could check for the symlinks before install, and ask
 >> the users to remove them, or abort install, or whatever. This
 >> educates the users, and allows the human to take corrective action
 >> as needed. 

 Brian> How can the preinst remove symlinks that belong to (i.e., are
 Brian> contained as part of) another package?  That is clearly a
 Brian> violation of policy, UNLESS these "compatibility symlinks" are
 Brian> not contained in the kernel-image (or kernel-pcmcia-modules)
 Brian> package, but are created by that package's postinst script.

	No, I said "Ask the users to remove the symlinks". So, go
 ahead, remove the symlinks, come back, and hit enter while I wait and
 whistle at the flowers. The sysadmin can remove any symlink he/she
 wants, it is their machine.

	If not, you can just abort at that point; the human can decide
 to use the kernel pcmcia modules, recompile a kenrel and pcmcia
 modules (built in or stand alone), remove the symlinks and retry, or
 whatever. The human gets to decide, we just provide the tools that
 they can use.

	manoj
-- 
 Sometimes you get an almost irresistible urge to go on living.
Manoj Srivastava   <srivasta@debian.org>  <http://www.debian.org/%7Esrivasta/>
1024R/C7261095 print CB D9 F4 12 68 07 E4 05  CC 2D 27 12 1D F5 E8 6E
1024D/BF24424C print 4966 F272 D093 B493 410B  924B 21BA DABB BF24 424C


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



Reply to: