[PATCH] xorg-proto-common: allow the empty package

[PATCH] xorg-proto-common: allow the empty package

[jackie.huang]

From: Jackie Huang <jackie.huang@windriver.com>

Allow the ${PN} package to be created empty to avoid error like:
ERROR: bigreqsproto not found in the base feeds (intel_x86_64 core2-64
x86_64 noarch any all).

Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
---
 meta/recipes-graphics/xorg-proto/xorg-proto-common.inc | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/meta/recipes-graphics/xorg-proto/xorg-proto-common.inc b/meta/recipes-graphics/xorg-proto/xorg-proto-common.inc
index 7a417c9..3f43189 100644
--- a/meta/recipes-graphics/xorg-proto/xorg-proto-common.inc
+++ b/meta/recipes-graphics/xorg-proto/xorg-proto-common.inc
@@ -22,5 +22,7 @@ PACKAGECONFIG ??= ""
 PACKAGECONFIG[xmlto] = "--with-xmlto, --without-xmlto, xmlto-native docbook-xml-dtd4-native docbook-xsl-stylesheets-native"
 
 # ${PN} is empty so we need to tweak -dev and -dbg package dependencies
+# and allow ${PN} to be created empty
+ALLOW_EMPTY_${PN} = "1"
 RDEPENDS_${PN}-dev = ""
 RRECOMMENDS_${PN}-dbg = "${PN}-dev (= ${EXTENDPKGV})"
-- 
1.9.1

Re: [PATCH] xorg-proto-common: allow the empty package

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 579 bytes --]

On 1 July 2015 at 08:52, <jackie.huang@windriver.com> wrote:

>  # ${PN} is empty so we need to tweak -dev and -dbg package dependencies
> +# and allow ${PN} to be created empty
> +ALLOW_EMPTY_${PN} = "1"
>  RDEPENDS_${PN}-dev = ""
>  RRECOMMENDS_${PN}-dbg = "${PN}-dev (= ${EXTENDPKGV})"
>

As the comment and surrounding lines state, the dependencies on PN from
PN-dev and PN-dbg are already removed.  What other package has a runtime
dependency on PN, and when you find it a better solution is to remove the
dependency instead of creating empty packages.

Ross

[-- Attachment #2: Type: text/html, Size: 1027 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Huang, Jie (Jackie)]

[-- Attachment #1: Type: text/plain, Size: 1270 bytes --]

Yeah I know it’s not good to have the empty package and we can remove dependencies to void error, but
our customer may add the package through IMAGE_INSTALL in local.conf and blame it fails until do_rootfs:

ERROR: bigreqsproto not found in the base feeds (intel_x86_64 core2-64 x86_64 noarch any all).

Do we have a way to find out a PN package is empty at an earlier time so that we may warn the user not
to add this package into image?

Thanks,
Jackie

From: Burton, Ross [mailto:ross.burton@intel.com]
Sent: Friday, July 03, 2015 3:40 AM
To: Huang, Jie (Jackie)
Cc: OE-core
Subject: Re: [OE-core] [PATCH] xorg-proto-common: allow the empty package


On 1 July 2015 at 08:52, <jackie.huang@windriver.com<mailto:jackie.huang@windriver.com>> wrote:
 # ${PN} is empty so we need to tweak -dev and -dbg package dependencies
+# and allow ${PN} to be created empty
+ALLOW_EMPTY_${PN} = "1"
 RDEPENDS_${PN}-dev = ""
 RRECOMMENDS_${PN}-dbg = "${PN}-dev (= ${EXTENDPKGV})"

As the comment and surrounding lines state, the dependencies on PN from PN-dev and PN-dbg are already removed.  What other package has a runtime dependency on PN, and when you find it a better solution is to remove the dependency instead of creating empty packages.

Ross

[-- Attachment #2: Type: text/html, Size: 5629 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1880 bytes --]

On Fri, Jul 03, 2015 at 01:34:31AM +0000, Huang, Jie (Jackie) wrote:
> Yeah I know it’s not good to have the empty package and we can remove dependencies to void error, but
> our customer may add the package through IMAGE_INSTALL in local.conf and blame it fails until do_rootfs:
> 
> ERROR: bigreqsproto not found in the base feeds (intel_x86_64 core2-64 x86_64 noarch any all).
> 
> Do we have a way to find out a PN package is empty at an earlier time so that we may warn the user not
> to add this package into image?

I as a customer would definitely prefer error message even later in the
build then getting completely useless empty package installed in my
image and blame you for installing it when I was expecting to find
bigreqsproto in my rootfs.

> From: Burton, Ross [mailto:ross.burton@intel.com]
> Sent: Friday, July 03, 2015 3:40 AM
> To: Huang, Jie (Jackie)
> Cc: OE-core
> Subject: Re: [OE-core] [PATCH] xorg-proto-common: allow the empty package
> 
> 
> On 1 July 2015 at 08:52, <jackie.huang@windriver.com<mailto:jackie.huang@windriver.com>> wrote:
>  # ${PN} is empty so we need to tweak -dev and -dbg package dependencies
> +# and allow ${PN} to be created empty
> +ALLOW_EMPTY_${PN} = "1"
>  RDEPENDS_${PN}-dev = ""
>  RRECOMMENDS_${PN}-dbg = "${PN}-dev (= ${EXTENDPKGV})"
> 
> As the comment and surrounding lines state, the dependencies on PN from PN-dev and PN-dbg are already removed.  What other package has a runtime dependency on PN, and when you find it a better solution is to remove the dependency instead of creating empty packages.
> 
> Ross

> -- 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core


-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 547 bytes --]

On 3 July 2015 at 10:04, Martin Jansa <martin.jansa@gmail.com> wrote:

> I as a customer would definitely prefer error message even later in the
> build then getting completely useless empty package installed in my
> image and blame you for installing it when I was expecting to find
> bigreqsproto in my rootfs.
>

Agreed.  Installing bigreqsproto and not getting anything is far more
annoying than the error and following epiphany that if you want the
development headers for bigreqsproto you should install bigreqsproto-dev.

Ross

[-- Attachment #2: Type: text/html, Size: 956 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Huang, Jie (Jackie)]

[-- Attachment #1: Type: text/plain, Size: 1207 bytes --]

I see your point, I think we can make the empty PN rdepends on PN-dev(but the dev-deps QA check need to be skipped),

then when user try to add PN, PN-dev will also be installed, so it’s not an useless empty package.



I see that openssh has similar situation, PN is allowed empty, but it rdepends on “${PN}-scp ${PN}-ssh ${PN}-sshd ${PN}-keygen”.



Does it make sense to you?



Thanks,

Jackie


From: Burton, Ross [mailto:ross.burton@intel.com]
Sent: Friday, July 03, 2015 5:14 PM
To: Martin Jansa
Cc: Huang, Jie (Jackie); OE-core
Subject: Re: [OE-core] [PATCH] xorg-proto-common: allow the empty package


On 3 July 2015 at 10:04, Martin Jansa <martin.jansa@gmail.com<mailto:martin.jansa@gmail.com>> wrote:
I as a customer would definitely prefer error message even later in the
build then getting completely useless empty package installed in my
image and blame you for installing it when I was expecting to find
bigreqsproto in my rootfs.

Agreed.  Installing bigreqsproto and not getting anything is far more annoying than the error and following epiphany that if you want the development headers for bigreqsproto you should install bigreqsproto-dev.

Ross

[-- Attachment #2: Type: text/html, Size: 5225 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 1289 bytes --]

On 3 July 2015 at 10:34, Huang, Jie (Wind River) <jackie.huang@windriver.com
> wrote:

> I see your point, I think we can make the empty PN rdepends on PN-dev(but
> the dev-deps QA check need to be skipped),
>
> then when user try to add PN, PN-dev will also be installed, so it’s not
> an useless empty package.
>

bigreqsproto is development headers.  Therefore its in the bigreqsproto-dev
package.  This is not hard to understand, nor should we start filling the
feeds up with empty packages just to please users who can't verify the
package names.

Didn't we have this discussion a few months ago?  There are some recipes
where PN not existing is unusual (avahi was a good example, and I had a
partial branch to fix that), but for recipes which entirely consist of
development files, there only being a development package is reasonable and
predictable.

(If I *had* to do this, I'd have RPROVIDES_${PN}-dev += ${PN})

> I see that openssh has similar situation, PN is allowed empty, but it
> rdepends on “${PN}-scp ${PN}-ssh ${PN}-sshd ${PN}-keygen”.
>
That's a different use case entirely: pulling in the entire SSH suite in a
single package whilst still allowing resource-constrained or secure systems
to pull in only the parts they want.

Ross

[-- Attachment #2: Type: text/html, Size: 1943 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Huang, Jie (Jackie)]

[-- Attachment #1: Type: text/plain, Size: 1693 bytes --]

Thanks for the explanation. I think you’re suggesting that we do nothing for such issue or
just tell the user to use PN-dev for this kind of packages, right?

Thanks,
Jackie

From: Burton, Ross [mailto:ross.burton@intel.com]
Sent: Friday, July 03, 2015 5:39 PM
To: Huang, Jie (Jackie)
Cc: Martin Jansa; OE-core
Subject: Re: [OE-core] [PATCH] xorg-proto-common: allow the empty package


On 3 July 2015 at 10:34, Huang, Jie (Wind River) <jackie.huang@windriver.com<mailto:jackie.huang@windriver.com>> wrote:

I see your point, I think we can make the empty PN rdepends on PN-dev(but the dev-deps QA check need to be skipped),

then when user try to add PN, PN-dev will also be installed, so it’s not an useless empty package.

bigreqsproto is development headers.  Therefore its in the bigreqsproto-dev package.  This is not hard to understand, nor should we start filling the feeds up with empty packages just to please users who can't verify the package names.

Didn't we have this discussion a few months ago?  There are some recipes where PN not existing is unusual (avahi was a good example, and I had a partial branch to fix that), but for recipes which entirely consist of development files, there only being a development package is reasonable and predictable.

(If I *had* to do this, I'd have RPROVIDES_${PN}-dev += ${PN})

I see that openssh has similar situation, PN is allowed empty, but it rdepends on “${PN}-scp ${PN}-ssh ${PN}-sshd ${PN}-keygen”.
That's a different use case entirely: pulling in the entire SSH suite in a single package whilst still allowing resource-constrained or secure systems to pull in only the parts they want.

Ross

[-- Attachment #2: Type: text/html, Size: 6012 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 286 bytes --]

On 3 July 2015 at 11:16, Huang, Jie (Wind River) <jackie.huang@windriver.com
> wrote:

> Thanks for the explanation. I think you’re suggesting that we do nothing
> for such issue or
>
> just tell the user to use PN-dev for this kind of packages, right?
>
>
Correct.

Ross

[-- Attachment #2: Type: text/html, Size: 948 bytes --]

Re: [PATCH] xorg-proto-common: allow the empty package

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1950 bytes --]

On Fri, Jul 03, 2015 at 10:16:09AM +0000, Huang, Jie (Jackie) wrote:
> Thanks for the explanation. I think you’re suggesting that we do nothing for such issue or
> just tell the user to use PN-dev for this kind of packages, right?

Yes, if user wants to install PN-dev, then he should add PN-dev to
IMAGE_INSTALL not PN.

> From: Burton, Ross [mailto:ross.burton@intel.com]
> Sent: Friday, July 03, 2015 5:39 PM
> To: Huang, Jie (Jackie)
> Cc: Martin Jansa; OE-core
> Subject: Re: [OE-core] [PATCH] xorg-proto-common: allow the empty package
> 
> 
> On 3 July 2015 at 10:34, Huang, Jie (Wind River) <jackie.huang@windriver.com<mailto:jackie.huang@windriver.com>> wrote:
> 
> I see your point, I think we can make the empty PN rdepends on PN-dev(but the dev-deps QA check need to be skipped),
> 
> then when user try to add PN, PN-dev will also be installed, so it’s not an useless empty package.
> 
> bigreqsproto is development headers.  Therefore its in the bigreqsproto-dev package.  This is not hard to understand, nor should we start filling the feeds up with empty packages just to please users who can't verify the package names.
> 
> Didn't we have this discussion a few months ago?  There are some recipes where PN not existing is unusual (avahi was a good example, and I had a partial branch to fix that), but for recipes which entirely consist of development files, there only being a development package is reasonable and predictable.
> 
> (If I *had* to do this, I'd have RPROVIDES_${PN}-dev += ${PN})
> 
> I see that openssh has similar situation, PN is allowed empty, but it rdepends on “${PN}-scp ${PN}-ssh ${PN}-sshd ${PN}-keygen”.
> That's a different use case entirely: pulling in the entire SSH suite in a single package whilst still allowing resource-constrained or secure systems to pull in only the parts they want.
> 
> Ross

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]