[PATCH] systemd: disable 'libdir' QA check

[PATCH] systemd: disable 'libdir' QA check

[Mark Asselstine]

When building systemd with multilib support enabled in your build you
will get the following QA warnings (if the 'libdir' QA check is
enabled.)

WARNING: systemd-1_232-r0 do_package_qa: QA Issue: systemd-dbg: found \
 library in wrong location: /lib/systemd/.debug/libsystemd-shared-232.so
systemd: found library in wrong location: /lib/systemd/libsystemd-shared.so
systemd: found library in wrong location: /lib/systemd/libsystemd-shared-232.so [libdir]

Since systemd 231 upstream has included an 'internal' library which
they explicitly place in the application specific /lib/systemd
directory. You can see some of the discussion about this placement
here https://github.com/systemd/systemd/issues/3810

This placement is being picked up by the QA checker since when
multilibs are enabled it expects all libraries to be in lib32 or
lib64. Since the systemd and systemd-dbg packages don't contain any
other libraries we can respect the upstream placement and skip this QA
check for these packages. Unfortunately the QA mechanism doesn't allow
us to specify individual files so this approach is the best we can do.

Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
---
 meta/recipes-core/systemd/systemd_232.bb | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/meta/recipes-core/systemd/systemd_232.bb b/meta/recipes-core/systemd/systemd_232.bb
index baee02e..c86badb 100644
--- a/meta/recipes-core/systemd/systemd_232.bb
+++ b/meta/recipes-core/systemd/systemd_232.bb
@@ -485,7 +485,8 @@ RRECOMMENDS_${PN} += "${@bb.utils.contains('PACKAGECONFIG', 'serial-getty-genera
                       os-release \
 "
 
-INSANE_SKIP_${PN} += "dev-so"
+INSANE_SKIP_${PN} += "dev-so libdir"
+INSANE_SKIP_${PN}-dbg += "libdir"
 INSANE_SKIP_${PN}-doc += " libdir"
 
 PACKAGES =+ "udev udev-hwdb"
-- 
2.7.4

Re: [PATCH] systemd: disable 'libdir' QA check

[Khem Raj]

> On Dec 12, 2016, at 11:36 AM, Mark Asselstine <mark.asselstine@windriver.com> wrote:
> 
> When building systemd with multilib support enabled in your build you
> will get the following QA warnings (if the 'libdir' QA check is
> enabled.)
> 
> WARNING: systemd-1_232-r0 do_package_qa: QA Issue: systemd-dbg: found \
> library in wrong location: /lib/systemd/.debug/libsystemd-shared-232.so
> systemd: found library in wrong location: /lib/systemd/libsystemd-shared.so
> systemd: found library in wrong location: /lib/systemd/libsystemd-shared-232.so [libdir]

Can we check if systemd can be a bit more flexible and accept it to go into multilib libdir ?
this patch can be then applied if its not acceptable to systemd community.

> 
> Since systemd 231 upstream has included an 'internal' library which
> they explicitly place in the application specific /lib/systemd
> directory. You can see some of the discussion about this placement
> here https://github.com/systemd/systemd/issues/3810
> 
> This placement is being picked up by the QA checker since when
> multilibs are enabled it expects all libraries to be in lib32 or
> lib64. Since the systemd and systemd-dbg packages don't contain any
> other libraries we can respect the upstream placement and skip this QA
> check for these packages. Unfortunately the QA mechanism doesn't allow
> us to specify individual files so this approach is the best we can do.
> 
> Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
> ---
> meta/recipes-core/systemd/systemd_232.bb | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/meta/recipes-core/systemd/systemd_232.bb b/meta/recipes-core/systemd/systemd_232.bb
> index baee02e..c86badb 100644
> --- a/meta/recipes-core/systemd/systemd_232.bb
> +++ b/meta/recipes-core/systemd/systemd_232.bb
> @@ -485,7 +485,8 @@ RRECOMMENDS_${PN} += "${@bb.utils.contains('PACKAGECONFIG', 'serial-getty-genera
>                       os-release \
> "
> 
> -INSANE_SKIP_${PN} += "dev-so"
> +INSANE_SKIP_${PN} += "dev-so libdir"
> +INSANE_SKIP_${PN}-dbg += "libdir"
> INSANE_SKIP_${PN}-doc += " libdir"
> 
> PACKAGES =+ "udev udev-hwdb"
> -- 
> 2.7.4
> 
> -- 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core

Re: [PATCH] systemd: disable 'libdir' QA check

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 399 bytes --]

On 12 December 2016 at 19:59, Khem Raj <raj.khem@gmail.com> wrote:

> Can we check if systemd can be a bit more flexible and accept it to go
> into multilib libdir ?
> this patch can be then applied if its not acceptable to systemd community.
>

Agreed, whilst I get that systemd really wants to use /lib as "libexec in
/" the libraries most definitely should be following $libdir.

Ross

[-- Attachment #2: Type: text/html, Size: 794 bytes --]

Re: [PATCH] systemd: disable 'libdir' QA check

[Mark Asselstine]

On Monday, December 12, 2016 11:59:04 AM EST Khem Raj wrote:
> > On Dec 12, 2016, at 11:36 AM, Mark Asselstine
> > <mark.asselstine@windriver.com> wrote:
> > 
> > When building systemd with multilib support enabled in your build you
> > will get the following QA warnings (if the 'libdir' QA check is
> > enabled.)
> > 
> > WARNING: systemd-1_232-r0 do_package_qa: QA Issue: systemd-dbg: found \
> > library in wrong location: /lib/systemd/.debug/libsystemd-shared-232.so
> > systemd: found library in wrong location:
> > /lib/systemd/libsystemd-shared.so
> > systemd: found library in wrong location:
> > /lib/systemd/libsystemd-shared-232.so [libdir]
> Can we check if systemd can be a bit more flexible and accept it to go into
> multilib libdir ? this patch can be then applied if its not acceptable to
> systemd community.

I think the discussion I pointed to in the commit log closes the door on any 
such change. Specific the comment from Lennart --
https://github.com/systemd/systemd/issues/3810#issuecomment-235290526

They don't want the library to be found in the default search path, they want 
to maintain this as a "hidden, internal resource".

Being on the record that I am aware of this discussion/conclusion if I send a 
patch now I would look like a bit of ass for wasting their time. If someone 
else wants to push the issue, feel free. Unless there is a suitable 
recommendation that would satisfy their wishes and ours that I am missing.

Mark


> > Since systemd 231 upstream has included an 'internal' library which
> > they explicitly place in the application specific /lib/systemd
> > directory. You can see some of the discussion about this placement
> > here https://github.com/systemd/systemd/issues/3810
> > 
> > This placement is being picked up by the QA checker since when
> > multilibs are enabled it expects all libraries to be in lib32 or
> > lib64. Since the systemd and systemd-dbg packages don't contain any
> > other libraries we can respect the upstream placement and skip this QA
> > check for these packages. Unfortunately the QA mechanism doesn't allow
> > us to specify individual files so this approach is the best we can do.
> > 
> > Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
> > ---
> > meta/recipes-core/systemd/systemd_232.bb | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/meta/recipes-core/systemd/systemd_232.bb
> > b/meta/recipes-core/systemd/systemd_232.bb index baee02e..c86badb 100644
> > --- a/meta/recipes-core/systemd/systemd_232.bb
> > +++ b/meta/recipes-core/systemd/systemd_232.bb
> > @@ -485,7 +485,8 @@ RRECOMMENDS_${PN} +=
> > "${@bb.utils.contains('PACKAGECONFIG', 'serial-getty-genera> 
> >                       os-release \
> > 
> > "
> > 
> > -INSANE_SKIP_${PN} += "dev-so"
> > +INSANE_SKIP_${PN} += "dev-so libdir"
> > +INSANE_SKIP_${PN}-dbg += "libdir"
> > INSANE_SKIP_${PN}-doc += " libdir"
> > 
> > PACKAGES =+ "udev udev-hwdb"

Re: [PATCH] systemd: disable 'libdir' QA check

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 740 bytes --]

On 12 December 2016 at 20:51, Mark Asselstine <mark.asselstine@windriver.com
> wrote:

> I think the discussion I pointed to in the commit log closes the door on
> any
> such change. Specific the comment from Lennart --
> https://github.com/systemd/systemd/issues/3810#issuecomment-235290526
>
> They don't want the library to be found in the default search path, they
> want
> to maintain this as a "hidden, internal resource".
>

Oh if it's an implementation detail of systemd and not a user-facing
library then I can see their argument I guess.

Don't agree with it though.  But, the point is that as long as the parts of
systemd that could realistically be multilibd are in $libdir then this
isn't a problem.

Ross

[-- Attachment #2: Type: text/html, Size: 1368 bytes --]

Re: [PATCH] systemd: disable 'libdir' QA check

[Mark Asselstine]

On Monday, December 12, 2016 9:17:14 PM EST Burton, Ross wrote:
> On 12 December 2016 at 20:51, Mark Asselstine <mark.asselstine@windriver.com
> > wrote:
> > 
> > I think the discussion I pointed to in the commit log closes the door on
> > any
> > such change. Specific the comment from Lennart --
> > https://github.com/systemd/systemd/issues/3810#issuecomment-235290526
> > 
> > They don't want the library to be found in the default search path, they
> > want
> > to maintain this as a "hidden, internal resource".
> 
> Oh if it's an implementation detail of systemd and not a user-facing
> library then I can see their argument I guess.
> 
> Don't agree with it though.  But, the point is that as long as the parts of
> systemd that could realistically be multilibd are in $libdir then this
> isn't a problem.

It is definitely an internal infra item for systemd and not seen as a 'public' 
library so I think we are best to run with it. I am with you that I don't 
necessarily agree with it. Khem, does this satisfy your concerns?

Mark

> 
> Ross

Re: [PATCH] systemd: disable 'libdir' QA check

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1335 bytes --]

This library may be private however this does not mean it should be hidden
in a dedicated path. But I guess its better to lower qa guards then carry a
patch for life

On Dec 13, 2016 7:20 AM, "Mark Asselstine" <mark.asselstine@windriver.com>
wrote:

On Monday, December 12, 2016 9:17:14 PM EST Burton, Ross wrote:
> On 12 December 2016 at 20:51, Mark Asselstine <
mark.asselstine@windriver.com
> > wrote:
> >
> > I think the discussion I pointed to in the commit log closes the door on
> > any
> > such change. Specific the comment from Lennart --
> > https://github.com/systemd/systemd/issues/3810#issuecomment-235290526
> >
> > They don't want the library to be found in the default search path, they
> > want
> > to maintain this as a "hidden, internal resource".
>
> Oh if it's an implementation detail of systemd and not a user-facing
> library then I can see their argument I guess.
>
> Don't agree with it though.  But, the point is that as long as the parts
of
> systemd that could realistically be multilibd are in $libdir then this
> isn't a problem.

It is definitely an internal infra item for systemd and not seen as a
'public'
library so I think we are best to run with it. I am with you that I don't
necessarily agree with it. Khem, does this satisfy your concerns?

Mark

>
> Ross

[-- Attachment #2: Type: text/html, Size: 2078 bytes --]