2:04 p.m.
Hey all,
I've noticed that when dkms is installed, it's not grabbing the right
kernel-devel package as a dependency. Instead, it grabs
kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not sure
why. If I install kernel-devel first, everything is fine. Otherwise,
it grabs the wrong kernel-devel and DKMS modules don't build.
I filed a bug for it on RHBZ too:
https://bugzilla.redhat.com/show_bug.cgi?id=1228897
Anyone have any idea why this is happening and a way to work around it?
--
真実はいつも一つ！/ Always, there's only one truth!
2:41 p.m.
On 09.06.2015 21:04, Neal Gompa wrote:
I've noticed that when dkms is installed, it's not grabbing
the right
kernel-devel package as a dependency.
Because that's not possible with ordinary dependencies (might be
possible with soft dependencies [Suggests, Enhances etc]) unless we
change something in the kernel packaging (see below).
Instead, it grabs
kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not sure
why.
Because all kernel*devel package provide kernel-devel iirc.
Anyone have any idea why this is happening and a way to work around
it?
Create something like a meta-package "kernel-devel-all" that depends on
all available kernel-devel packages (kernel-devel, kernel-PAE-devel,
kernel-debug-devel, ...) for the arch in question; then add "Requires:
kernel-devel-all" to the akmods and dkms packages. That's messy and
creates overhead for users, but that's afaics the only way it will work
for everyone; otherwise you'll always run into situations where a
kernel-devel package for one kernel variant gets installed while you are
running different variant. Example: You get kernel-devel via some
dependency in akmods or dkms; but you are running kernel-PAE on your
i686 machine, so building modules with akmods or dkms will fail, as
that's requires kernel-PAE-devel.
HTH; CU, knurd
3:01 p.m.
This has also been a problem for several releases with the akmods from
rpmfusion. I think the more "correct" solution (read to the end) would be
to somehow prioritize the kernel-devel package (possibly multiple) that
matches the installed kernel(s). kernel-debug-devel is never the "correct"
choice when only standard kernels are installed, but it gets installed
because of alphabetical order; this ordering is one way to get something
useful installed instead. This of course is pie-in-the-sky and doesn't
match with how dependencies can currently be declared and resolved, AFAIK.
-Dan
On Tue, Jun 9, 2015 at 3:41 PM, Thorsten Leemhuis <fedora(a)leemhuis.info>
wrote:
On 09.06.2015 21:04, Neal Gompa wrote:
> I've noticed that when dkms is installed, it's not grabbing the right
> kernel-devel package as a dependency.
Because that's not possible with ordinary dependencies (might be
possible with soft dependencies [Suggests, Enhances etc]) unless we
change something in the kernel packaging (see below).
> Instead, it grabs
> kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not sure
> why.
Because all kernel*devel package provide kernel-devel iirc.
> Anyone have any idea why this is happening and a way to work around it?
Create something like a meta-package "kernel-devel-all" that depends on
all available kernel-devel packages (kernel-devel, kernel-PAE-devel,
kernel-debug-devel, ...) for the arch in question; then add "Requires:
kernel-devel-all" to the akmods and dkms packages. That's messy and
creates overhead for users, but that's afaics the only way it will work
for everyone; otherwise you'll always run into situations where a
kernel-devel package for one kernel variant gets installed while you are
running different variant. Example: You get kernel-devel via some
dependency in akmods or dkms; but you are running kernel-PAE on your
i686 machine, so building modules with akmods or dkms will fail, as
that's requires kernel-PAE-devel.
HTH; CU, knurd
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
3:24 p.m.
Am 09.06.2015 um 22:01 schrieb Dan Book:
This has also been a problem for several releases with the akmods
from
rpmfusion. I think the more "correct" solution (read to the end) would
be to somehow prioritize the kernel-devel package (possibly multiple)
that matches the installed kernel(s). kernel-debug-devel is never the
"correct" choice when only standard kernels are installed, but it gets
installed because of alphabetical order; this ordering is one way to get
something useful installed instead. This of course is pie-in-the-sky and
doesn't match with how dependencies can currently be declared and
resolved, AFAIK
honestly that is a *non* problem at all
* if you need to compile modules you need kernel-devel
* if you *once* had it installed yum7dnf will keep it up-to-date
3:38 p.m.
That's great, for people who already know what to do. The problem is for
people who try to install dkms or akmods and then it doesn't work. There
are plenty of people who use dkms or akmods modules but don't compile
anything themselves (and those people are less likely to understand the
kernel-devel issue).
-Dan
On Tue, Jun 9, 2015 at 4:24 PM, Reindl Harald <h.reindl(a)thelounge.net>
wrote:
Am 09.06.2015 um 22:01 schrieb Dan Book:
> This has also been a problem for several releases with the akmods from
> rpmfusion. I think the more "correct" solution (read to the end) would
> be to somehow prioritize the kernel-devel package (possibly multiple)
> that matches the installed kernel(s). kernel-debug-devel is never the
> "correct" choice when only standard kernels are installed, but it gets
> installed because of alphabetical order; this ordering is one way to get
> something useful installed instead. This of course is pie-in-the-sky and
> doesn't match with how dependencies can currently be declared and
> resolved, AFAIK
>
honestly that is a *non* problem at all
* if you need to compile modules you need kernel-devel
* if you *once* had it installed yum7dnf will keep it up-to-date
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
3:43 p.m.
Am 09.06.2015 um 22:38 schrieb Dan Book:
That's great, for people who already know what to do. The problem
is for
people who try to install dkms or akmods and then it doesn't work. There
are plenty of people who use dkms or akmods modules but don't compile
anything themselves (and those people are less likely to understand the
kernel-devel issue).
which issue?
you get the most recent kernel-devel and that's it
if you don't reboot to the most recent kernel installed on your system
before biuld dkmos/akmod modules - bad luck - but how is that a Fedora
problem at all - both are helpers in case of wrong hardware decisions
and so you need in the worst case to ivent 10 minutes of our time you
saved by bouy random broken hardware
that *is not* a Fedora problem at all - period
On Tue, Jun 9, 2015 at 4:24 PM, Reindl Harald
<h.reindl(a)thelounge.net
<mailto:h.reindl@thelounge.net>> wrote:
Am 09.06.2015 um 22:01 schrieb Dan Book:
This has also been a problem for several releases with the
akmods from
rpmfusion. I think the more "correct" solution (read to the end)
would
be to somehow prioritize the kernel-devel package (possibly
multiple)
that matches the installed kernel(s). kernel-debug-devel is
never the
"correct" choice when only standard kernels are installed, but
it gets
installed because of alphabetical order; this ordering is one
way to get
something useful installed instead. This of course is
pie-in-the-sky and
doesn't match with how dependencies can currently be declared and
resolved, AFAIK
honestly that is a *non* problem at all
* if you need to compile modules you need kernel-devel
* if you *once* had it installed yum7dnf will keep it up-to-date
3:50 p.m.
On Tue, Jun 9, 2015 at 4:43 PM, Reindl Harald <h.reindl(a)thelounge.net> wrote:
Am 09.06.2015 um 22:38 schrieb Dan Book:
>
> That's great, for people who already know what to do. The problem is for
> people who try to install dkms or akmods and then it doesn't work. There
> are plenty of people who use dkms or akmods modules but don't compile
> anything themselves (and those people are less likely to understand the
> kernel-devel issue).
which issue?
you get the most recent kernel-devel and that's it
if you don't reboot to the most recent kernel installed on your system
before biuld dkmos/akmod modules - bad luck - but how is that a Fedora
problem at all - both are helpers in case of wrong hardware decisions and so
you need in the worst case to ivent 10 minutes of our time you saved by bouy
random broken hardware
that *is not* a Fedora problem at all - period
You're still assuming they installed kernel-devel manually. Dan is
saying people using akmods or DKMS don't do that, which is true. In
short, you're wrong that it isn't a Fedora problem but as Thorsten
said there is no great solution.
Please stop arguing now.
josh
1:12 a.m.
Dan Book wrote on 09.06.2015 22:01:
This has also been a problem for several releases with the akmods
from
rpmfusion.
There was always a problem; there where just less people running into
it, because yum/dnf installed "kernel-devel" most of the time, which
matched the kernel that was running on most systems; but quite a few
x86-32 users ran into problems afaics, as kernel-PAE get's installed
there sometimes and hence they needed kernel-PAE-devel.
Mosts of the docs and howtos on the net don't mention that, which leads
to confused and frustrating users; those in the end might be one of
multiple problems why they switch to another distribution.
I think the more "correct" solution (read to the end)
would
be to somehow prioritize the kernel-devel package (possibly multiple)
that matches the installed kernel(s).
That's not a solution, that's solving the problem for some users and
ignoring others (those that use kernel-PAE for example).
[...]
CU
thl
On Tue, Jun 9, 2015 at 3:41 PM, Thorsten Leemhuis
<fedora(a)leemhuis.info
<mailto:fedora@leemhuis.info>> wrote:
On 09.06.2015 21:04, Neal Gompa wrote:
> I've noticed that when dkms is installed, it's not grabbing the right
> kernel-devel package as a dependency.
Because that's not possible with ordinary dependencies (might be
possible with soft dependencies [Suggests, Enhances etc]) unless we
change something in the kernel packaging (see below).
> Instead, it grabs
> kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not sure
> why.
Because all kernel*devel package provide kernel-devel iirc.
> Anyone have any idea why this is happening and a way to work around it?
Create something like a meta-package "kernel-devel-all" that depends on
all available kernel-devel packages (kernel-devel, kernel-PAE-devel,
kernel-debug-devel, ...) for the arch in question; then add "Requires:
kernel-devel-all" to the akmods and dkms packages. That's messy and
creates overhead for users, but that's afaics the only way it will work
for everyone; otherwise you'll always run into situations where a
kernel-devel package for one kernel variant gets installed while you are
running different variant. Example: You get kernel-devel via some
dependency in akmods or dkms; but you are running kernel-PAE on your
i686 machine, so building modules with akmods or dkms will fail, as
that's requires kernel-PAE-devel.
HTH; CU, knurd
--
devel mailing list
devel(a)lists.fedoraproject.org <mailto:devel@lists.fedoraproject.org>
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
7:44 a.m.
On Wed, Jun 10, 2015 at 2:12 AM, Thorsten Leemhuis <fedora(a)leemhuis.info>
wrote:
Dan Book wrote on 09.06.2015 22:01:
> This has also been a problem for several releases with the akmods from
> rpmfusion.
There was always a problem; there where just less people running into
it, because yum/dnf installed "kernel-devel" most of the time, which
matched the kernel that was running on most systems; but quite a few
x86-32 users ran into problems afaics, as kernel-PAE get's installed
there sometimes and hence they needed kernel-PAE-devel.
Mosts of the docs and howtos on the net don't mention that, which leads
to confused and frustrating users; those in the end might be one of
multiple problems why they switch to another distribution.
> I think the more "correct" solution (read to the end) would
> be to somehow prioritize the kernel-devel package (possibly multiple)
> that matches the installed kernel(s).
That's not a solution, that's solving the problem for some users and
ignoring others (those that use kernel-PAE for example).
"The kernel-devel package that matches the installed kernel(s)" would
include kernel-PAE-devel matching kernel-PAE, in this fantasy land I
invented.
> [...]
CU
thl
-Dan
> On Tue, Jun 9, 2015 at 3:41 PM, Thorsten Leemhuis <fedora(a)leemhuis.info
> <mailto:fedora@leemhuis.info>> wrote:
>
> On 09.06.2015 21:04, Neal Gompa wrote:
> > I've noticed that when dkms is installed, it's not grabbing the
right
> > kernel-devel package as a dependency.
>
> Because that's not possible with ordinary dependencies (might be
> possible with soft dependencies [Suggests, Enhances etc]) unless we
> change something in the kernel packaging (see below).
>
> > Instead, it grabs
> > kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not
sure
> > why.
>
> Because all kernel*devel package provide kernel-devel iirc.
>
> > Anyone have any idea why this is happening and a way to work
around it?
>
> Create something like a meta-package "kernel-devel-all" that depends
on
> all available kernel-devel packages (kernel-devel, kernel-PAE-devel,
> kernel-debug-devel, ...) for the arch in question; then add
"Requires:
> kernel-devel-all" to the akmods and dkms packages. That's messy and
> creates overhead for users, but that's afaics the only way it will
work
> for everyone; otherwise you'll always run into situations where a
> kernel-devel package for one kernel variant gets installed while you
are
> running different variant. Example: You get kernel-devel via some
> dependency in akmods or dkms; but you are running kernel-PAE on your
> i686 machine, so building modules with akmods or dkms will fail, as
> that's requires kernel-PAE-devel.
>
> HTH; CU, knurd
> --
> devel mailing list
> devel(a)lists.fedoraproject.org <mailto:devel@lists.fedoraproject.org>
> https://admin.fedoraproject.org/mailman/listinfo/devel
> Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
>
>
>
>
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
6:24 a.m.
On Wed, Jun 10, 2015 at 8:44 AM, Dan Book <grinnz(a)gmail.com> wrote:
On Wed, Jun 10, 2015 at 2:12 AM, Thorsten Leemhuis <fedora(a)leemhuis.info>
wrote:
> Dan Book wrote on 09.06.2015 22:01:
> > This has also been a problem for several releases with the akmods from
> > rpmfusion.
>
> There was always a problem; there where just less people running into
> it, because yum/dnf installed "kernel-devel" most of the time, which
> matched the kernel that was running on most systems; but quite a few
> x86-32 users ran into problems afaics, as kernel-PAE get's installed
> there sometimes and hence they needed kernel-PAE-devel.
>
> Mosts of the docs and howtos on the net don't mention that, which leads
> to confused and frustrating users; those in the end might be one of
> multiple problems why they switch to another distribution.
>
> > I think the more "correct" solution (read to the end) would
> > be to somehow prioritize the kernel-devel package (possibly multiple)
> > that matches the installed kernel(s).
>
> That's not a solution, that's solving the problem for some users and
> ignoring others (those that use kernel-PAE for example).
>
"The kernel-devel package that matches the installed kernel(s)" would
include kernel-PAE-devel matching kernel-PAE, in this fantasy land I
invented.
>
> > [...]
>
> CU
> thl
>
-Dan
>
> > On Tue, Jun 9, 2015 at 3:41 PM, Thorsten Leemhuis <fedora(a)leemhuis.info
> > <mailto:fedora@leemhuis.info>> wrote:
> >
> > On 09.06.2015 21:04, Neal Gompa wrote:
> > > I've noticed that when dkms is installed, it's not grabbing
the
> right
> > > kernel-devel package as a dependency.
> >
> > Because that's not possible with ordinary dependencies (might be
> > possible with soft dependencies [Suggests, Enhances etc]) unless we
> > change something in the kernel packaging (see below).
> >
> > > Instead, it grabs
> > > kernel-debug-devel. This occurs on Fedora 21 and 22, and I'm not
> sure
> > > why.
> >
> > Because all kernel*devel package provide kernel-devel iirc.
> >
> > > Anyone have any idea why this is happening and a way to work
> around it?
> >
> > Create something like a meta-package "kernel-devel-all" that
> depends on
> > all available kernel-devel packages (kernel-devel, kernel-PAE-devel,
> > kernel-debug-devel, ...) for the arch in question; then add
> "Requires:
> > kernel-devel-all" to the akmods and dkms packages. That's messy
and
> > creates overhead for users, but that's afaics the only way it will
> work
> > for everyone; otherwise you'll always run into situations where a
> > kernel-devel package for one kernel variant gets installed while
> you are
> > running different variant. Example: You get kernel-devel via some
> > dependency in akmods or dkms; but you are running kernel-PAE on your
> > i686 machine, so building modules with akmods or dkms will fail, as
> > that's requires kernel-PAE-devel.
> >
> > HTH; CU, knurd
> > --
> > devel mailing list
> > devel(a)lists.fedoraproject.org <mailto:devel@lists.fedoraproject.org
> >
> > https://admin.fedoraproject.org/mailman/listinfo/devel
> > Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
> >
> >
> >
> >
> --
> devel mailing list
> devel(a)lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/devel
> Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
>
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
What about some kind of virtual provides defined in repos/rpm/somewhere
that would automatically grab the kernel-devel package associated with the
*exact* kernel that is running at the time yum/dnf is installing a program
that depends on it? That would allow for things like DKMS to function
properly, since they'll have what they need to build kernel modules. Going
forward, kernel upgrades will also drag in the appropriate kernel-devel
packages to match, keeping things sane.
--
真実はいつも一つ！/ Always, there's only one truth!
6:55 a.m.
On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa <ngompa13(a)gmail.com> wrote:
What about some kind of virtual provides defined in
repos/rpm/somewhere that
would automatically grab the kernel-devel package associated with the exact
kernel that is running at the time yum/dnf is installing a program that
depends on it? That would allow for things like DKMS to function properly,
since they'll have what they need to build kernel modules. Going forward,
kernel upgrades will also drag in the appropriate kernel-devel packages to
match, keeping things sane.
That would help DKMS at the cost of breaking rpmfusion, koji, or any
other scenario where you want to build a kernel module for a kernel
that isn't running. The kernel-devel package is meant to be flexible
enough so that you can install it without having anything close to the
same kernel version actually running. Think "all our builders run
RHEL7, but we want to build modules for Fedora."
It's worth pointing out that the Fedora DKMS package could implement
the logic you suggest in DKMS itself. Then it would either fail if
the relevant kernel-devel isn't installed (or kernel-PAE-devel, etc).
However, that still doesn't really make it an automated working
solution and still requires manual intervention to actually get it
installed.
As I said, there are no great solutions here.
josh
7:19 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
Josh Boyer wrote on 12.06.2015 13:55:
On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa
<ngompa13(a)gmail.com> wrote:
[...]
As I said, there are no great solutions here.
A "works most of the time"-solution would be: Install kernel-devel by
default. But I'm not seriously suggesting that, because I fully agree:
It's not a great solution.
Did anyone(¹) look at soft dependencies in rpm? Can they make our
tools install the kernel-devel packages in the variants that match the
kernel variants installed? I suspect they are made to solve problems
like this, but I'm not sure; and I don't know how far soft dependencies
are supported in out current stack of packaging tools.
CU
knurd
(¹) no, I'm not looking at you Josh
7:40 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
----- Original Message -----
From: "Thorsten Leemhuis" <fedora(a)leemhuis.info>
To: "Development discussions related to Fedora"
<devel(a)lists.fedoraproject.org>
Sent: Friday, June 12, 2015 2:19:10 PM
Subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft-
Re: DKMS is not installing
the right kernel-devel package)
Josh Boyer wrote on 12.06.2015 13:55:
> On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa <ngompa13(a)gmail.com> wrote:
> [...]
> As I said, there are no great solutions here.
A "works most of the time"-solution would be: Install kernel-devel by
default. But I'm not seriously suggesting that, because I fully agree:
It's not a great solution.
Did anyone(¹) look at soft dependencies in rpm? Can they make our
tools install the kernel-devel packages in the variants that match the
kernel variants installed? I suspect they are made to solve problems
like this, but I'm not sure; and I don't know how far soft dependencies
are supported in out current stack of packaging tools.
CU
knurd
(¹) no, I'm not looking at you Josh
AFAIK, it is discussed these days whether weak dependencies can be used to express package
preferences: https://fedoraproject.org/wiki/PackagingDrafts/WeakDependencies IIRC, there
were some concerns but I don't remember what was the conclusion.
--
Radek Holý
Associate Software Engineer
Software Management Team
Red Hat Czech
7:48 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On Fri, Jun 12, 2015 at 8:40 AM, Radek Holy <rholy(a)redhat.com> wrote:
----- Original Message -----
> From: "Thorsten Leemhuis" <fedora(a)leemhuis.info>
> To: "Development discussions related to Fedora" <
devel(a)lists.fedoraproject.org>
> Sent: Friday, June 12, 2015 2:19:10 PM
> Subject: Can soft dependencies help to get the proper kernel-devel
packages? (Was: Soft- Re: DKMS is not installing
> the right kernel-devel package)
>
> Josh Boyer wrote on 12.06.2015 13:55:
> > On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa <ngompa13(a)gmail.com>
wrote:
> > [...]
> > As I said, there are no great solutions here.
>
> A "works most of the time"-solution would be: Install kernel-devel by
> default. But I'm not seriously suggesting that, because I fully agree:
> It's not a great solution.
>
> Did anyone(¹) look at soft dependencies in rpm? Can they make our
> tools install the kernel-devel packages in the variants that match the
> kernel variants installed? I suspect they are made to solve problems
> like this, but I'm not sure; and I don't know how far soft dependencies
> are supported in out current stack of packaging tools.
>
> CU
> knurd
>
> (¹) no, I'm not looking at you Josh
AFAIK, it is discussed these days whether weak dependencies can be used to
express package preferences:
https://fedoraproject.org/wiki/PackagingDrafts/WeakDependencies IIRC,
there were some concerns but I don't remember what was the conclusion.
--
Radek Holý
Associate Software Engineer
Software Management Team
Red Hat Czech
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Soft/weak dependencies are allowed, according to FESCo
<https://fedorahosted.org/fesco/ticket/1203#comment:6>;. The main problem
would be how to structure it to trigger appropriately for this case. Is it
even possible to generate a soft dependency at install time? Otherwise,
how do we ensure the "right" one is picked?
--
真実はいつも一つ！/ Always, there's only one truth!
8:16 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
----- Original Message -----
From: "Neal Gompa" <ngompa13(a)gmail.com>
To: "Development discussions related to Fedora"
<devel(a)lists.fedoraproject.org>
Sent: Friday, June 12, 2015 2:48:19 PM
Subject: Re: Can soft dependencies help to get the proper kernel-devel
packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel
package)
On Fri, Jun 12, 2015 at 8:40 AM, Radek Holy < rholy(a)redhat.com
> wrote:
> ----- Original Message -----
> > From: "Thorsten Leemhuis" < fedora(a)leemhuis.info >
> > To: "Development discussions related to Fedora" <
> > devel(a)lists.fedoraproject.org >
> > Sent: Friday, June 12, 2015 2:19:10 PM
> > Subject: Can soft dependencies help to get the proper kernel-devel
> > packages? (Was: Soft- Re: DKMS is not installing
> > the right kernel-devel package)
> >
> > Josh Boyer wrote on 12.06.2015 13:55:
> > > On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa < ngompa13(a)gmail.com >
> > > wrote:
> > > [...]
> > > As I said, there are no great solutions here.
> >
> > A "works most of the time"-solution would be: Install kernel-devel
by
> > default. But I'm not seriously suggesting that, because I fully agree:
> > It's not a great solution.
> >
> > Did anyone(¹) look at soft dependencies in rpm? Can they make our
> > tools install the kernel-devel packages in the variants that match the
> > kernel variants installed? I suspect they are made to solve problems
> > like this, but I'm not sure; and I don't know how far soft
dependencies
> > are supported in out current stack of packaging tools.
> >
> > CU
> > knurd
> >
> > (¹) no, I'm not looking at you Josh
> AFAIK, it is discussed these days whether weak dependencies can
be used to
> express package preferences:
> https://fedoraproject.org/wiki/PackagingDrafts/WeakDependencies IIRC, there
> were some concerns but I don't remember what was the conclusion.
> --
> Radek Holý
> Associate Software Engineer
> Software Management Team
> Red Hat Czech
> --
> devel mailing list
> devel(a)lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/devel
> Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Soft/weak dependencies are allowed, according to FESCo . The main
problem
would be how to structure it to trigger appropriately for this case. Is it
even possible to generate a soft dependency at install time? Otherwise, how
do we ensure the "right" one is picked?
--
真実はいつも一つ！/ Always, there's only one truth!
Well, they may be allowed but it's not clear yet whether they can be used for any
other purpose then for pointing to packages that enables optional features of the package
in question.
--
Radek Holý
Associate Software Engineer
Software Management Team
Red Hat Czech
8:18 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On 06/12/2015 02:48 PM, Neal Gompa wrote:
Soft/weak dependencies are allowed, according to FESCo
It
doesn't matter much what FESCO says of believes in this case.
ATM, neither is the technical infrastructure is in place (it is
evolving) nor are impact/consequences clear not are the prototypical
use-cases of soft/weak-deps clear.
@Thorsten: I do not think, weak/soft-deps are of use in your case.
Unfortunatley, I do not have solution, either.
Ralf
8:23 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On Fri, Jun 12, 2015 at 9:18 AM, Ralf Corsepius <rc040203(a)freenet.de> wrote:
On 06/12/2015 02:48 PM, Neal Gompa wrote:
> Soft/weak dependencies are allowed, according to FESCo
It doesn't matter much what FESCO says of believes in this case.
ATM, neither is the technical infrastructure is in place (it is evolving)
nor are impact/consequences clear not are the prototypical use-cases of
soft/weak-deps clear.
@Thorsten: I do not think, weak/soft-deps are of use in your case.
Unfortunatley, I do not have solution, either.
Ralf
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
What do you mean by the technical infrastructure for it is lacking?
--
真実はいつも一つ！/ Always, there's only one truth!
8:41 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On 06/12/2015 03:23 PM, Neal Gompa wrote:
On Fri, Jun 12, 2015 at 9:18 AM, Ralf Corsepius
<rc040203(a)freenet.de> wrote:
> On 06/12/2015 02:48 PM, Neal Gompa wrote:
>
>> Soft/weak dependencies are allowed, according to FESCo
>
> It doesn't matter much what FESCO says of believes in this case.
>
> ATM, neither is the technical infrastructure is in place (it is evolving)
> nor are impact/consequences clear not are the prototypical use-cases of
> soft/weak-deps clear.
>
> @Thorsten: I do not think, weak/soft-deps are of use in your case.
> Unfortunatley, I do not have solution, either.
What do you mean by the technical infrastructure for it is lacking?
E.g. /usr/bin/rpm still lacks the --what* command-line options
corresponding to soft/weak deps.
This issue was mentioned the rpm maintainer in an FPC meeting (IIRC, 3
weeks ago), when we were discussing soft/weak-deps but nothing seems to
have happened since.
Therefore I just (I could have done so earlier, but have been on
vacation) filed an RFE against rpm:
https://bugzilla.redhat.com/show_bug.cgi?id=1231247
Ralf
9:10 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On Fri, Jun 12, 2015 at 9:41 AM, Ralf Corsepius <rc040203(a)freenet.de> wrote:
On 06/12/2015 03:23 PM, Neal Gompa wrote:
>
> On Fri, Jun 12, 2015 at 9:18 AM, Ralf Corsepius <rc040203(a)freenet.de>
> wrote:
>>
>> On 06/12/2015 02:48 PM, Neal Gompa wrote:
>>
>>> Soft/weak dependencies are allowed, according to FESCo
>>
>>
>> It doesn't matter much what FESCO says of believes in this case.
>>
>> ATM, neither is the technical infrastructure is in place (it is evolving)
>> nor are impact/consequences clear not are the prototypical use-cases of
>> soft/weak-deps clear.
>>
>> @Thorsten: I do not think, weak/soft-deps are of use in your case.
>> Unfortunatley, I do not have solution, either.
> What do you mean by the technical infrastructure for it is lacking?
E.g. /usr/bin/rpm still lacks the --what* command-line options corresponding
to soft/weak deps.
This issue was mentioned the rpm maintainer in an FPC meeting (IIRC, 3 weeks
ago), when we were discussing soft/weak-deps but nothing seems to have
happened since.
Therefore I just (I could have done so earlier, but have been on vacation)
filed an RFE against rpm:
https://bugzilla.redhat.com/show_bug.cgi?id=1231247
Ralf
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Doesn't libsolv/hawkey handle this within itself, though?
--
真実はいつも一つ！/ Always, there's only one truth!
8:15 a.m.
New subject: Can soft dependencies help to get the proper kernel-devel packages? (Was: Soft- Re: DKMS is not installing the right kernel-devel package)
On 06/12/2015 02:40 PM, Radek Holy wrote:
AFAIK, it is discussed these days whether weak dependencies can be
used to express package preferences:
https://fedoraproject.org/wiki/PackagingDrafts/WeakDependencies IIRC,
there were some concerns but I don't remember what was the
conclusion.
The OR syntax I mentioned in my previous mail in the other thread could
help here as well. It could be expressed with something like:
Requires: kernel-devel = 4.0.2-1.fc22 | kernel-devel-virtualprovide =
4.0.2-1.fc22
so it would pick kernel-devel by default as it's the right choice for
most cases, but if the user already has kernel-PAE-devel installed, then
it would satisfy the kernel-devel-virtualprovide virtual provide and
just use that.
---
Kalev
3:28 a.m.
On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa
<ngompa13(a)gmail.com> wrote:
> What about some kind of virtual provides defined in repos/rpm/somewhere that
> would automatically grab the kernel-devel package associated with the exact
> kernel that is running at the time yum/dnf is installing a program that
> depends on it? That would allow for things like DKMS to function properly,
> since they'll have what they need to build kernel modules. Going forward,
> kernel upgrades will also drag in the appropriate kernel-devel packages to
> match, keeping things sane.
That would help DKMS at the cost of breaking rpmfusion, koji, or any
other scenario where you want to build a kernel module for a kernel
that isn't running. The kernel-devel package is meant to be flexible
enough so that you can install it without having anything close to the
same kernel version actually running.
I don’t think this would break the model: there would be no changes to packaging
kernel-devel, you would still be able to install kernel-devel-$whateverversion, but a
magic (dnf install ^kernel-devel-current) or “Requires: ^kernel-devel-current” would refer
to a specific one.
It's worth pointing out that the Fedora DKMS package could
implement
the logic you suggest in DKMS itself. Then it would either fail if
the relevant kernel-devel isn't installed (or kernel-PAE-devel, etc).
However, that still doesn't really make it an automated working
solution and still requires manual intervention to actually get it
installed.
It is actually already fairly easy to usefully (but not reliably) automate this, copying
an ugly hack from Chrome packaging, something like
%post dkms
the_details=…
echo 'dnf install kernel-devel-$the_details' | at now+1minute
As I said, there are no great solutions here.
Yeah, the above is another example of that ☺
Mirek
6:44 a.m.
On Mon, Jun 15, 2015 at 4:28 AM, Miloslav Trmač <mitr(a)redhat.com> wrote:
> On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa
<ngompa13(a)gmail.com> wrote:
> > What about some kind of virtual provides defined in repos/rpm/somewhere that
> > would automatically grab the kernel-devel package associated with the exact
> > kernel that is running at the time yum/dnf is installing a program that
> > depends on it? That would allow for things like DKMS to function properly,
> > since they'll have what they need to build kernel modules. Going forward,
> > kernel upgrades will also drag in the appropriate kernel-devel packages to
> > match, keeping things sane.
>
> That would help DKMS at the cost of breaking rpmfusion, koji, or any
> other scenario where you want to build a kernel module for a kernel
> that isn't running. The kernel-devel package is meant to be flexible
> enough so that you can install it without having anything close to the
> same kernel version actually running.
I don’t think this would break the model: there would be no changes to packaging
kernel-devel, you would still be able to install kernel-devel-$whateverversion, but a
magic (dnf install ^kernel-devel-current) or “Requires: ^kernel-devel-current” would refer
to a specific one.
> It's worth pointing out that the Fedora DKMS package could implement
> the logic you suggest in DKMS itself. Then it would either fail if
> the relevant kernel-devel isn't installed (or kernel-PAE-devel, etc).
> However, that still doesn't really make it an automated working
> solution and still requires manual intervention to actually get it
> installed.
It is actually already fairly easy to usefully (but not reliably) automate this, copying
an ugly hack from Chrome packaging, something like
> %post dkms
> the_details=…
> echo 'dnf install kernel-devel-$the_details' | at now+1minute
> As I said, there are no great solutions here.
Yeah, the above is another example of that ☺
Mirek
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
So why not implement something like kernel-devel-current? DKMS and
those who depend on the running kernel's kernel-devel can use it,
while the rest don't have to...
--
真実はいつも一つ！/ Always, there's only one truth!
7:01 a.m.
On Mon, Jun 15, 2015 at 4:28 AM, Miloslav Trmač <mitr(a)redhat.com> wrote:
> On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa
<ngompa13(a)gmail.com> wrote:
> > What about some kind of virtual provides defined in repos/rpm/somewhere that
> > would automatically grab the kernel-devel package associated with the exact
> > kernel that is running at the time yum/dnf is installing a program that
> > depends on it? That would allow for things like DKMS to function properly,
> > since they'll have what they need to build kernel modules. Going forward,
> > kernel upgrades will also drag in the appropriate kernel-devel packages to
> > match, keeping things sane.
>
> That would help DKMS at the cost of breaking rpmfusion, koji, or any
> other scenario where you want to build a kernel module for a kernel
> that isn't running. The kernel-devel package is meant to be flexible
> enough so that you can install it without having anything close to the
> same kernel version actually running.
I don’t think this would break the model: there would be no changes to packaging
kernel-devel, you would still be able to install kernel-devel-$whateverversion, but a
magic (dnf install ^kernel-devel-current) or “Requires: ^kernel-devel-current” would refer
to a specific one.
I'm not understanding what you're suggesting here.
josh
7:16 a.m.
On Mon, Jun 15, 2015 at 8:01 AM, Josh Boyer <jwboyer(a)fedoraproject.org> wrote:
On Mon, Jun 15, 2015 at 4:28 AM, Miloslav Trmač
<mitr(a)redhat.com> wrote:
>> On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa <ngompa13(a)gmail.com> wrote:
>> > What about some kind of virtual provides defined in repos/rpm/somewhere
that
>> > would automatically grab the kernel-devel package associated with the exact
>> > kernel that is running at the time yum/dnf is installing a program that
>> > depends on it? That would allow for things like DKMS to function properly,
>> > since they'll have what they need to build kernel modules. Going
forward,
>> > kernel upgrades will also drag in the appropriate kernel-devel packages to
>> > match, keeping things sane.
>>
>> That would help DKMS at the cost of breaking rpmfusion, koji, or any
>> other scenario where you want to build a kernel module for a kernel
>> that isn't running. The kernel-devel package is meant to be flexible
>> enough so that you can install it without having anything close to the
>> same kernel version actually running.
>
> I don’t think this would break the model: there would be no changes to packaging
kernel-devel, you would still be able to install kernel-devel-$whateverversion, but a
magic (dnf install ^kernel-devel-current) or “Requires: ^kernel-devel-current” would refer
to a specific one.
I'm not understanding what you're suggesting here.
josh
--
devel mailing list
devel(a)lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Essentially, a virtual provides called "kernel-devel-current" would
always resolve to grab the kernel-devel package that corresponds with
the running kernel, be it kernel-PAE, kernel-debug, or kernel (or for
things like the Raspberry Pi 2, kernel-rpi2 or whatever it would be
named). Other things that are safe with "kernel-devel" alone can still
use that, preserving existing behavior when doing so.
--
真実はいつも一つ！/ Always, there's only one truth!
7:25 a.m.
On Mon, Jun 15, 2015 at 8:16 AM, Neal Gompa <ngompa13(a)gmail.com> wrote:
On Mon, Jun 15, 2015 at 8:01 AM, Josh Boyer
<jwboyer(a)fedoraproject.org> wrote:
> On Mon, Jun 15, 2015 at 4:28 AM, Miloslav Trmač <mitr(a)redhat.com> wrote:
>>> On Fri, Jun 12, 2015 at 7:24 AM, Neal Gompa <ngompa13(a)gmail.com>
wrote:
>>> > What about some kind of virtual provides defined in repos/rpm/somewhere
that
>>> > would automatically grab the kernel-devel package associated with the
exact
>>> > kernel that is running at the time yum/dnf is installing a program that
>>> > depends on it? That would allow for things like DKMS to function
properly,
>>> > since they'll have what they need to build kernel modules. Going
forward,
>>> > kernel upgrades will also drag in the appropriate kernel-devel packages
to
>>> > match, keeping things sane.
>>>
>>> That would help DKMS at the cost of breaking rpmfusion, koji, or any
>>> other scenario where you want to build a kernel module for a kernel
>>> that isn't running. The kernel-devel package is meant to be flexible
>>> enough so that you can install it without having anything close to the
>>> same kernel version actually running.
>>
>> I don’t think this would break the model: there would be no changes to packaging
kernel-devel, you would still be able to install kernel-devel-$whateverversion, but a
magic (dnf install ^kernel-devel-current) or “Requires: ^kernel-devel-current” would refer
to a specific one.
>
> I'm not understanding what you're suggesting here.
Essentially, a virtual provides called
"kernel-devel-current" would
always resolve to grab the kernel-devel package that corresponds with
the running kernel, be it kernel-PAE, kernel-debug, or kernel (or for
things like the Raspberry Pi 2, kernel-rpi2 or whatever it would be
named). Other things that are safe with "kernel-devel" alone can still
use that, preserving existing behavior when doing so.
Yes, I understand the concept. What I don't understand is how one
would actually implement it. As far as I can tell, it would have to
be done outside of the kernel package itself. Perhaps as a DNF plugin
or something. If so, then great but thus far the DNF developers have
been trying very hard to not treat the kernel and its subpackages as
special by any means.
josh
3259
days inactive
3265
days old
24 comments
9 participants
participants (9)
-
Dan Book -
Josh Boyer -
Kalev Lember -
Miloslav Trmač -
Neal Gompa -
Radek Holy -
Ralf Corsepius -
Reindl Harald -
Thorsten Leemhuis