2:42 p.m.
Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
---
scripts/fast-build.sh | 13 +++++++++++++
1 file changed, 13 insertions(+)
create mode 100755 scripts/fast-build.sh
diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
new file mode 100755
index 0000000..19eaa4d
--- /dev/null
+++ b/scripts/fast-build.sh
@@ -0,0 +1,13 @@
+#! /bin/sh
+# Description:
+# rpmbuild combo to build the given architecture with
+# no debugging information, perf and tools.
+#
+# Sample usage:
+# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
+
+if [ -z "$1" ] || [ -z "$2" ]; then
+ echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
+fi
+
+rpmbuild --target $1 --without debuginfo --without perf --without tools --rebuild $2
--
2.7.4
4:25 p.m.
On Fri, Jun 10, 2016 at 12:42:59PM -0700, Miguel Flores Silverio wrote:
Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
---
scripts/fast-build.sh | 13 +++++++++++++
1 file changed, 13 insertions(+)
create mode 100755 scripts/fast-build.sh
You're busy eliminating pointless scripts, and then adding another one in
their place right here. :)
Just use the baseonly target in the spec, possibly with a patch to have
that also disable tools building to get exactly what you were after here.
Or add a target similar to baseonly for fastbuild.
--
Jarod Wilson
jarod(a)redhat.com
8:15 p.m.
On Fri, Jun 10, 2016 at 5:25 PM, Jarod Wilson <jarod(a)redhat.com> wrote:
On Fri, Jun 10, 2016 at 12:42:59PM -0700, Miguel Flores Silverio
wrote:
> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
> ---
> scripts/fast-build.sh | 13 +++++++++++++
> 1 file changed, 13 insertions(+)
> create mode 100755 scripts/fast-build.sh
You're busy eliminating pointless scripts, and then adding another one in
their place right here. :)
Actually, this isn't pointless at all.
Just use the baseonly target in the spec, possibly with a patch to
have
that also disable tools building to get exactly what you were after here.
Or add a target similar to baseonly for fastbuild.
That could be done, but then people have to know all the various
rpmbuild switches. It's a pain. From a novice user experience, it's
very difficult to figure out. You have to know that target exists,
which means reading the spec, which means understanding the spec,
which is easy for you and I but is impossible as a _starting_ point.
The purpose of the script is to give someone that just wants to build
a kernel without all the crap a very easy thing to run. The intended
user isn't us. It's so that we don't have to explain how to do it
over and over (or point people to a confusing wiki page) to people
that don't live in the kernel spec.
josh
8:45 a.m.
On Fri, Jun 10, 2016 at 09:15:52PM -0400, Josh Boyer wrote:
On Fri, Jun 10, 2016 at 5:25 PM, Jarod Wilson
<jarod(a)redhat.com> wrote:
> On Fri, Jun 10, 2016 at 12:42:59PM -0700, Miguel Flores Silverio wrote:
>> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
>> ---
>> scripts/fast-build.sh | 13 +++++++++++++
>> 1 file changed, 13 insertions(+)
>> create mode 100755 scripts/fast-build.sh
>
> You're busy eliminating pointless scripts, and then adding another one in
> their place right here. :)
Actually, this isn't pointless at all.
> Just use the baseonly target in the spec, possibly with a patch to have
> that also disable tools building to get exactly what you were after here.
> Or add a target similar to baseonly for fastbuild.
That could be done, but then people have to know all the various
rpmbuild switches. It's a pain. From a novice user experience, it's
very difficult to figure out. You have to know that target exists,
which means reading the spec, which means understanding the spec,
which is easy for you and I but is impossible as a _starting_ point.
The purpose of the script is to give someone that just wants to build
a kernel without all the crap a very easy thing to run. The intended
user isn't us. It's so that we don't have to explain how to do it
over and over (or point people to a confusing wiki page) to people
that don't live in the kernel spec.
Fair enough, though are people really needing to build custom kernels that
often? (It's quite likely I'm out of touch with reality). :)
--
Jarod Wilson
jarod(a)redhat.com
8:51 a.m.
On Mon, Jun 13, 2016 at 9:45 AM, Jarod Wilson <jarod(a)redhat.com> wrote:
On Fri, Jun 10, 2016 at 09:15:52PM -0400, Josh Boyer wrote:
> On Fri, Jun 10, 2016 at 5:25 PM, Jarod Wilson <jarod(a)redhat.com> wrote:
> > On Fri, Jun 10, 2016 at 12:42:59PM -0700, Miguel Flores Silverio wrote:
> >> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
> >> ---
> >> scripts/fast-build.sh | 13 +++++++++++++
> >> 1 file changed, 13 insertions(+)
> >> create mode 100755 scripts/fast-build.sh
> >
> > You're busy eliminating pointless scripts, and then adding another one in
> > their place right here. :)
>
> Actually, this isn't pointless at all.
>
> > Just use the baseonly target in the spec, possibly with a patch to have
> > that also disable tools building to get exactly what you were after here.
> > Or add a target similar to baseonly for fastbuild.
>
> That could be done, but then people have to know all the various
> rpmbuild switches. It's a pain. From a novice user experience, it's
> very difficult to figure out. You have to know that target exists,
> which means reading the spec, which means understanding the spec,
> which is easy for you and I but is impossible as a _starting_ point.
>
> The purpose of the script is to give someone that just wants to build
> a kernel without all the crap a very easy thing to run. The intended
> user isn't us. It's so that we don't have to explain how to do it
> over and over (or point people to a confusing wiki page) to people
> that don't live in the kernel spec.
Fair enough, though are people really needing to build custom kernels that
often? (It's quite likely I'm out of touch with reality). :)
Need and desire are two different things. It's unknown if there is a
major need (I would think actually fewer minor/specific needs), but
many people are interested in doing so as a "dip the toe in the water"
exercise.
josh
10:01 a.m.
>> >> Signed-off-by: Miguel Flores Silverio
<floresmigu3l(a)gmail.com>
>> >> ---
>> >> scripts/fast-build.sh | 13 +++++++++++++
>> >> 1 file changed, 13 insertions(+)
>> >> create mode 100755 scripts/fast-build.sh
>> >
>> > You're busy eliminating pointless scripts, and then adding another one
in
>> > their place right here. :)
>>
>> Actually, this isn't pointless at all.
>>
>> > Just use the baseonly target in the spec, possibly with a patch to have
>> > that also disable tools building to get exactly what you were after here.
>> > Or add a target similar to baseonly for fastbuild.
>>
>> That could be done, but then people have to know all the various
>> rpmbuild switches. It's a pain. From a novice user experience, it's
>> very difficult to figure out. You have to know that target exists,
>> which means reading the spec, which means understanding the spec,
>> which is easy for you and I but is impossible as a _starting_ point.
>>
>> The purpose of the script is to give someone that just wants to build
>> a kernel without all the crap a very easy thing to run. The intended
>> user isn't us. It's so that we don't have to explain how to do it
>> over and over (or point people to a confusing wiki page) to people
>> that don't live in the kernel spec.
>
> Fair enough, though are people really needing to build custom kernels that
> often? (It's quite likely I'm out of touch with reality). :)
Need and desire are two different things. It's unknown if there is a
major need (I would think actually fewer minor/specific needs), but
many people are interested in doing so as a "dip the toe in the water"
exercise.
I know with my non x86 use cases I'm a well and truly excluded from
this usecase (although I have numerous aliases that basically achieve
the same thing).
But I'm finding I'm having to do even custom x86_64 kernel builds a
lot more recently in particular for newer Intel laptops to get decent
battery life (mjg's SATA patches) and I know others doing similar for
skylake laptops. I'm not sure how best to solve those sorts of issues
given it's mostly a problem with upstream but the script certainly has
a useful usecase for those without a general understanding of the
Fedora kernel spec feature options.
Peter
2:39 p.m.
On 06/13/2016 06:45 AM, Jarod Wilson wrote:
On Fri, Jun 10, 2016 at 09:15:52PM -0400, Josh Boyer wrote:
> On Fri, Jun 10, 2016 at 5:25 PM, Jarod Wilson <jarod(a)redhat.com> wrote:
>> On Fri, Jun 10, 2016 at 12:42:59PM -0700, Miguel Flores Silverio wrote:
>>> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
>>> ---
>>> scripts/fast-build.sh | 13 +++++++++++++
>>> 1 file changed, 13 insertions(+)
>>> create mode 100755 scripts/fast-build.sh
>>
>> You're busy eliminating pointless scripts, and then adding another one in
>> their place right here. :)
>
> Actually, this isn't pointless at all.
>
>> Just use the baseonly target in the spec, possibly with a patch to have
>> that also disable tools building to get exactly what you were after here.
>> Or add a target similar to baseonly for fastbuild.
>
> That could be done, but then people have to know all the various
> rpmbuild switches. It's a pain. From a novice user experience, it's
> very difficult to figure out. You have to know that target exists,
> which means reading the spec, which means understanding the spec,
> which is easy for you and I but is impossible as a _starting_ point.
>
> The purpose of the script is to give someone that just wants to build
> a kernel without all the crap a very easy thing to run. The intended
> user isn't us. It's so that we don't have to explain how to do it
> over and over (or point people to a confusing wiki page) to people
> that don't live in the kernel spec.
Fair enough, though are people really needing to build custom kernels that
often? (It's quite likely I'm out of touch with reality). :)
It happens more often than you think. A pretty common case is for a test
patch. Usually I end up giving a scratch build to have a bug reporter
verify that a patch works. It's less work for me (and possibly faster)
to have a reporter just make a build themselves.
Thanks,
Laura
1:32 p.m.
On 06/10/2016 03:42 PM, Miguel Flores Silverio wrote:
Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
---
scripts/fast-build.sh | 13 +++++++++++++
1 file changed, 13 insertions(+)
create mode 100755 scripts/fast-build.sh
diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
new file mode 100755
index 0000000..19eaa4d
--- /dev/null
+++ b/scripts/fast-build.sh
@@ -0,0 +1,13 @@
+#! /bin/sh
+# Description:
+# rpmbuild combo to build the given architecture with
+# no debugging information, perf and tools.
+#
+# Sample usage:
+# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
+
+if [ -z "$1" ] || [ -z "$2" ]; then
+ echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
+fi
+
+rpmbuild --target $1 --without debuginfo --without perf --without tools --rebuild $2
--
Is the --target really necessary? Why not just do $(arch)?
FWIW jboyer ... I'm leaning towards Jarod's comment on whether or not this is
necessary. Why not put in a comment in the spec on this? OTOH ... it's a small
script so why I am complaining? :)
P.
2.7.4
_______________________________________________
kernel mailing list
kernel(a)lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/kernel@lists.fedoraproject.org
1:41 p.m.
On Mon, Jun 13, 2016 at 2:32 PM, Prarit Bhargava <prarit(a)redhat.com> wrote:
On 06/10/2016 03:42 PM, Miguel Flores Silverio wrote:
> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
> ---
> scripts/fast-build.sh | 13 +++++++++++++
> 1 file changed, 13 insertions(+)
> create mode 100755 scripts/fast-build.sh
>
> diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
> new file mode 100755
> index 0000000..19eaa4d
> --- /dev/null
> +++ b/scripts/fast-build.sh
> @@ -0,0 +1,13 @@
> +#! /bin/sh
> +# Description:
> +# rpmbuild combo to build the given architecture with
> +# no debugging information, perf and tools.
> +#
> +# Sample usage:
> +# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
> +
> +if [ -z "$1" ] || [ -z "$2" ]; then
> + echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
> +fi
> +
> +rpmbuild --target $1 --without debuginfo --without perf --without tools --rebuild
$2
> --
Is the --target really necessary? Why not just do $(arch)?
Not following that question.
FWIW jboyer ... I'm leaning towards Jarod's comment on
whether or not this is
necessary. Why not put in a comment in the spec on this?
Because the spec is a nightmare of text and confusion for anyone that
hasn't spent a week staring at it. If we want to get really picky,
it's one of the worst spec files in Fedora aside from gcc. There are
many reasons why it is so, but new users don't know those reasons and
frankly don't care. If someone wants to tackle that beast, I'm all
for massive improvements that leave all the pieces functional.
OTOH ... it's a small
script so why I am complaining? :)
I'm ignoring your complaining.
josh
2:35 p.m.
On 06/13/2016 11:41 AM, Josh Boyer wrote:
On Mon, Jun 13, 2016 at 2:32 PM, Prarit Bhargava
<prarit(a)redhat.com> wrote:
> On 06/10/2016 03:42 PM, Miguel Flores Silverio wrote:
>> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
>> ---
>> scripts/fast-build.sh | 13 +++++++++++++
>> 1 file changed, 13 insertions(+)
>> create mode 100755 scripts/fast-build.sh
>>
>> diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
>> new file mode 100755
>> index 0000000..19eaa4d
>> --- /dev/null
>> +++ b/scripts/fast-build.sh
>> @@ -0,0 +1,13 @@
>> +#! /bin/sh
>> +# Description:
>> +# rpmbuild combo to build the given architecture with
>> +# no debugging information, perf and tools.
>> +#
>> +# Sample usage:
>> +# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
>> +
>> +if [ -z "$1" ] || [ -z "$2" ]; then
>> + echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
>> +fi
>> +
>> +rpmbuild --target $1 --without debuginfo --without perf --without tools
--rebuild $2
>> --
>
> Is the --target really necessary? Why not just do $(arch)?
Not following that question.
I think the question is do we ever want to build for an arch that doesn't
match the current system. As long as we have 32-bit available the target
is useful so I think it should stay.
> FWIW jboyer ... I'm leaning towards Jarod's comment on
whether or not this is
> necessary. Why not put in a comment in the spec on this?
Because the spec is a nightmare of text and confusion for anyone that
hasn't spent a week staring at it. If we want to get really picky,
it's one of the worst spec files in Fedora aside from gcc. There are
many reasons why it is so, but new users don't know those reasons and
frankly don't care. If someone wants to tackle that beast, I'm all
for massive improvements that leave all the pieces functional.
Yes. Having a quick build shortcut reduces the burden on users building
their own kernel.
Thanks,
Laura
3:34 p.m.
On 06/13/2016 03:35 PM, Laura Abbott wrote:
On 06/13/2016 11:41 AM, Josh Boyer wrote:
> On Mon, Jun 13, 2016 at 2:32 PM, Prarit Bhargava <prarit(a)redhat.com> wrote:
>> On 06/10/2016 03:42 PM, Miguel Flores Silverio wrote:
>>> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
>>> ---
>>> scripts/fast-build.sh | 13 +++++++++++++
>>> 1 file changed, 13 insertions(+)
>>> create mode 100755 scripts/fast-build.sh
>>>
>>> diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
>>> new file mode 100755
>>> index 0000000..19eaa4d
>>> --- /dev/null
>>> +++ b/scripts/fast-build.sh
>>> @@ -0,0 +1,13 @@
>>> +#! /bin/sh
>>> +# Description:
>>> +# rpmbuild combo to build the given architecture with
>>> +# no debugging information, perf and tools.
>>> +#
>>> +# Sample usage:
>>> +# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
>>> +
>>> +if [ -z "$1" ] || [ -z "$2" ]; then
>>> + echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
>>> +fi
>>> +
>>> +rpmbuild --target $1 --without debuginfo --without perf --without tools
>>> --rebuild $2
>>> --
>>
>> Is the --target really necessary? Why not just do $(arch)?
>
$(arch) should return the native target.
I was assuming that the script would only be run natively ... but Laura brings
up the important case of x86 64-bit vs 32-bit.
Yes. Having a quick build shortcut reduces the burden on users building
their own kernel.
Agreed (although I reserve the right to complain about minuscule things in
future and have jboyer ignore those too :) :) ).
P.
>
> Thanks,
> Laura
> _______________________________________________
> kernel mailing list
> kernel(a)lists.fedoraproject.org
> https://lists.fedoraproject.org/admin/lists/kernel@lists.fedoraproject.org
8:16 p.m.
On Mon, Jun 13, 2016 at 4:34 PM, Prarit Bhargava <prarit(a)redhat.com> wrote:
On 06/13/2016 03:35 PM, Laura Abbott wrote:
> On 06/13/2016 11:41 AM, Josh Boyer wrote:
>> On Mon, Jun 13, 2016 at 2:32 PM, Prarit Bhargava <prarit(a)redhat.com>
wrote:
>>> On 06/10/2016 03:42 PM, Miguel Flores Silverio wrote:
>>>> Signed-off-by: Miguel Flores Silverio <floresmigu3l(a)gmail.com>
>>>> ---
>>>> scripts/fast-build.sh | 13 +++++++++++++
>>>> 1 file changed, 13 insertions(+)
>>>> create mode 100755 scripts/fast-build.sh
>>>>
>>>> diff --git a/scripts/fast-build.sh b/scripts/fast-build.sh
>>>> new file mode 100755
>>>> index 0000000..19eaa4d
>>>> --- /dev/null
>>>> +++ b/scripts/fast-build.sh
>>>> @@ -0,0 +1,13 @@
>>>> +#! /bin/sh
>>>> +# Description:
>>>> +# rpmbuild combo to build the given architecture with
>>>> +# no debugging information, perf and tools.
>>>> +#
>>>> +# Sample usage:
>>>> +# ./fast-build.sh x86_64 kernel-4.7.0-0.rc1.git1.2.fc25.src.rpm
>>>> +
>>>> +if [ -z "$1" ] || [ -z "$2" ]; then
>>>> + echo "usage: $0 [ arch ] [ kernel-x.x.x.fcxx.src.rpm ] "
>>>> +fi
>>>> +
>>>> +rpmbuild --target $1 --without debuginfo --without perf --without tools
>>>> --rebuild $2
>>>> --
>>>
>>> Is the --target really necessary? Why not just do $(arch)?
>>
$(arch) should return the native target.
I was assuming that the script would only be run natively ... but Laura brings
up the important case of x86 64-bit vs 32-bit.
Fedora ships cross compilers as well. And yes, people use them to
test build. And yes, that's totally valid :).
> Yes. Having a quick build shortcut reduces the burden on users
building
> their own kernel.
Agreed (although I reserve the right to complain about minuscule things in
future and have jboyer ignore those too :) :) ).
I only ignore them if they're misplaced. If we were suggesting this
script was "THE" way to do builds then you'd have totally valid
arguments. But that's not the case. It's a helper, for arguably
community outreach purposes.
At any rate, it's a friendly ignore if you can conceive of it that way.
josh
7:10 a.m.
On 06/13/2016 09:16 PM, Josh Boyer wrote:
On Mon, Jun 13, 2016 at 4:34 PM, Prarit Bhargava
<prarit(a)redhat.com> wrote:
> Agreed (although I reserve the right to complain about minuscule things in
> future and have jboyer ignore those too :) :) ).
I only ignore them if they're misplaced. If we were suggesting this
script was "THE" way to do builds then you'd have totally valid
arguments. But that's not the case. It's a helper, for arguably
community outreach purposes.
FWIW ... a real comment on the patch:
I've used --target in two cases. The first is to do specific noarch builds that
bundle non-arch specific stuff like a bunch of scripts, and the second is to
cross build.
For example (taken from a script of mine) to check builds across different arches:
CROSS_RPMFLAGS = $(RPMBUILD) --define "_sourcedir $(SOURCES)" --define
"_builddir $(RPM)/BUILD" --define "_srcrpmdir $(RPM)/SRPMS" --define
"_rpmdir
$(RPM)/RPMS" --define "_specdir $(RPM)/SPECS" --define "dist
$(DIST)"
$(CROSS_RPMFLAGS) --target ppc64 --with cross --without debuginfo -bc
$(RPM)/SOURCES/kernel.spec
$(CROSS_RPMFLAGS) --target s390x --with cross --without debuginfo -bc
$(RPM)/SOURCES/kernel.spec
$(CROSS_RPMFLAGS) -ba $(RPM)/SOURCES/kernel.spec
--target implies to me that we can cross build. I think the script should
probably be updated with a note that the script is not meant to be used for
cross building. The other option is to add with "--with cross", although
I'm
not sure what the ramification of "--with cross" is when building natively.
At any rate, it's a friendly ignore if you can conceive of it that way.
I absolutely took it that way! :)
P.
2:54 p.m.
On vr, 2016-06-10 at 12:42 -0700, Miguel Flores Silverio wrote:
+rpmbuild --target $1 --without debuginfo --without perf --without
tools --rebuild $2
When I build locally I also have:
--with baseonly
--without debug
--without doc
I've been using these additional flags for years. My hope is that doing
this speeds up building the kernel rpms. Of course I've never actually
measured anything so this might be cargo cult behavior on my side.
(I also build with CONFIG_DEBUG_INFO unset, for a year or two now. See
https://lkml.org/lkml/2014/1/28/402 . I mainly noticed that it saves a
few GB of space in my rpmbuild directory, if memory serves me right.
Cargo cult again?)
Paul Bolle
2:47 p.m.
On 06/13/2016 12:54 PM, Paul Bolle wrote:
On vr, 2016-06-10 at 12:42 -0700, Miguel Flores Silverio wrote:
> +rpmbuild --target $1 --without debuginfo --without perf --without tools --rebuild
$2
When I build locally I also have:
--with baseonly
--without debug
--without doc
I've been using these additional flags for years. My hope is that doing
this speeds up building the kernel rpms. Of course I've never actually
measured anything so this might be cargo cult behavior on my side.
(I also build with CONFIG_DEBUG_INFO unset, for a year or two now. See
https://lkml.org/lkml/2014/1/28/402 . I mainly noticed that it saves a
few GB of space in my rpmbuild directory, if memory serves me right.
Cargo cult again?)
Thanks for this info. CONFIG_DEBUG_INFO does save a bunch of disk space.
It goes from ~7GB -> ~1.5GB which is much more reasonable.
I think --without debug would be a good addition to this. Miguel, can
you send out another version of this patch with --without debug added?
Thanks,
Laura
2:51 p.m.
On Tue, Jun 14, 2016 at 3:47 PM, Laura Abbott <labbott(a)redhat.com> wrote:
On 06/13/2016 12:54 PM, Paul Bolle wrote:
>
> On vr, 2016-06-10 at 12:42 -0700, Miguel Flores Silverio wrote:
>>
>> +rpmbuild --target $1 --without debuginfo --without perf --without tools
>> --rebuild $2
>
>
> When I build locally I also have:
> --with baseonly
> --without debug
> --without doc
>
> I've been using these additional flags for years. My hope is that doing
> this speeds up building the kernel rpms. Of course I've never actually
> measured anything so this might be cargo cult behavior on my side.
>
> (I also build with CONFIG_DEBUG_INFO unset, for a year or two now. See
> https://lkml.org/lkml/2014/1/28/402 . I mainly noticed that it saves a
> few GB of space in my rpmbuild directory, if memory serves me right.
> Cargo cult again?)
>
Thanks for this info. CONFIG_DEBUG_INFO does save a bunch of disk space.
It goes from ~7GB -> ~1.5GB which is much more reasonable.
To be clear, that means tools like crash, gdb, and the
kernel-debuginfo package won't work at all because the information
isn't part of the build. If you just want to test compiling, that's
fine. If you're actively developing something and just want a faster
build but need to debug issues, it might not be the best choice.
I think --without debug would be a good addition to this. Miguel,
can
you send out another version of this patch with --without debug added?
Yes, that is probably a good addition.
josh
4:45 p.m.
On Tue, 2016-06-14 at 12:47 -0700, Laura Abbott wrote:
On 06/13/2016 12:54 PM, Paul Bolle wrote:
> On vr, 2016-06-10 at 12:42 -0700, Miguel Flores Silverio wrote:
> > +rpmbuild --target $1 --without debuginfo --without perf --
> > without tools --rebuild $2
>
> When I build locally I also have:
> --with baseonly
> --without debug
> --without doc
>
> I've been using these additional flags for years. My hope is that
> doing
> this speeds up building the kernel rpms. Of course I've never
> actually
> measured anything so this might be cargo cult behavior on my side.
>
> (I also build with CONFIG_DEBUG_INFO unset, for a year or two now.
> See
> https://lkml.org/lkml/2014/1/28/402 . I mainly noticed that it
> saves a
> few GB of space in my rpmbuild directory, if memory serves me
> right.
> Cargo cult again?)
>
Thanks for this info. CONFIG_DEBUG_INFO does save a bunch of disk
space.
It goes from ~7GB -> ~1.5GB which is much more reasonable.
I think --without debug would be a good addition to this. Miguel, can
you send out another version of this patch with --without debug
added?
Yes! I can do that.
Thanks,
Laura
Thanks everyone for your feedback! I am starting to get very familiar
with the different tools and components in the kernel. I do have to
admit that the spec file is an intimidating file at first not only
that, but the kernel in general. I am eager to keep on learning and
contribute to the best of my abilities.
Regards,
Miguel
2874
days inactive
2878
days old
kernel@lists.fedoraproject.org
16 comments
8 participants
participants (8)
-
Jarod Wilson -
Josh Boyer -
Josh Boyer -
Laura Abbott -
Miguel A. Flores Silverio -
Paul Bolle -
Peter Robinson -
Prarit Bhargava