* Re: [OE-core] [PATCH 1/1] eglibc: remove runtime dependency of perl from eglibc-mtrace
@ 2012-06-11 22:07 ` Saul Wold
0 siblings, 0 replies; 10+ messages in thread
From: Saul Wold @ 2012-06-11 22:07 UTC (permalink / raw)
To: Darren Hart; +Cc: yocto, Patches and discussions about the oe-core layer
On 06/08/2012 04:02 PM, Darren Hart wrote:
>
>
> On 06/08/2012 03:23 PM, nitin.a.kamble@intel.com wrote:
>> From: Nitin A Kamble<nitin.a.kamble@intel.com>
>>
>> perl needs eglibc to build.
>> The presence of runtime dependency of
>> perl for eglibc-mtrace caused bitbake to build perl before eglibc,
>> which causes build failure of perl with poky-tiny distro
>>
>
> So is this a circular dependency chain?
>
> perl DEPENDS on eglibc
> eglibc (because of eglibc-mtrace) RDEPENDS on perl?
>
> If so, doesn't this solution leave eglibc-mtrace with an incomplete set
> of dependencies in it's final package meta-data?
>
> Would the correct solution be to break eglibc-mtrace out into a separate
> recipe.
>
I am not sure if this would be more correct or have the PACKAGES contain
something like ${EGLIBC_PACKAGE_MTRACE}, which could the be over-ridden
by the yocto-tiny distro.
EGLIBC_PACKAGE_MTRACE ?= "${PN}-mtrace"
and then change ${PN}-mtrace in the PACKAGES list to
${EGLIBC_PACKAGE_MTRACE}.
This exposes another "global", so I am not sure which is better.
Sau!
> eglibc-mtrace.bb could then DEPENDS=eglibc and RDEPENDS=perl and
> poky-tiny would need to be able to exclude eglibc-mtrace.
>
>
>> This fixes bug: [YOCTO #2523]
>>
>> Signed-off-by: Nitin A Kamble<nitin.a.kamble@intel.com>
>> ---
>> meta/recipes-core/eglibc/eglibc-package.inc | 1 -
>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>
>> diff --git a/meta/recipes-core/eglibc/eglibc-package.inc b/meta/recipes-core/eglibc/eglibc-package.inc
>> index ce37155..423729a 100644
>> --- a/meta/recipes-core/eglibc/eglibc-package.inc
>> +++ b/meta/recipes-core/eglibc/eglibc-package.inc
>> @@ -55,7 +55,6 @@ FILES_${PN}-dbg += "${libexecdir}/*/.debug ${libdir}/audit/.debug"
>> FILES_catchsegv${PKGSUFFIX} = "${bindir}/catchsegv"
>> RDEPENDS_catchsegv${PKGSUFFIX} = "libsegfault"
>> RDEPENDS_${PN}-utils += "bash"
>> -RDEPENDS_${PN}-mtrace += "perl"
>> FILES_${PN}-pcprofile = "${base_libdir}/libpcprofile.so"
>> FILES_eglibc-thread-db${PKGSUFFIX} = "${base_libdir}/libthread_db.so.* ${base_libdir}/libthread_db-*.so"
>> RPROVIDES_${PN}-dev += "libc-dev"
>
^ permalink raw reply [flat|nested] 10+ messages in thread* Re: [yocto] [PATCH 1/1] eglibc: remove runtime dependency of perl from eglibc-mtrace
2012-06-11 22:07 ` [OE-core] " Saul Wold
@ 2012-06-11 22:08 ` Darren Hart
-1 siblings, 0 replies; 10+ messages in thread
From: Darren Hart @ 2012-06-11 22:08 UTC (permalink / raw)
To: Saul Wold; +Cc: yocto, Patches and discussions about the oe-core layer
On 06/11/2012 03:07 PM, Saul Wold wrote:
> On 06/08/2012 04:02 PM, Darren Hart wrote:
>>
>>
>> On 06/08/2012 03:23 PM, nitin.a.kamble@intel.com wrote:
>>> From: Nitin A Kamble<nitin.a.kamble@intel.com>
>>>
>>> perl needs eglibc to build.
>>> The presence of runtime dependency of
>>> perl for eglibc-mtrace caused bitbake to build perl before eglibc,
>>> which causes build failure of perl with poky-tiny distro
>>>
>>
>> So is this a circular dependency chain?
>>
>> perl DEPENDS on eglibc
>> eglibc (because of eglibc-mtrace) RDEPENDS on perl?
>>
>> If so, doesn't this solution leave eglibc-mtrace with an incomplete set
>> of dependencies in it's final package meta-data?
>>
>> Would the correct solution be to break eglibc-mtrace out into a separate
>> recipe.
>>
> I am not sure if this would be more correct or have the PACKAGES contain
> something like ${EGLIBC_PACKAGE_MTRACE}, which could the be over-ridden
> by the yocto-tiny distro.
>
> EGLIBC_PACKAGE_MTRACE ?= "${PN}-mtrace"
>
> and then change ${PN}-mtrace in the PACKAGES list to
> ${EGLIBC_PACKAGE_MTRACE}.
>
> This exposes another "global", so I am not sure which is better.
This doesn't seem inconsistent with other solutions to similar sorts of
problems. This is certainly less work and a smaller change.
--
Darren
>
> Sau!
>
>
>> eglibc-mtrace.bb could then DEPENDS=eglibc and RDEPENDS=perl and
>> poky-tiny would need to be able to exclude eglibc-mtrace.
>>
>>
>>> This fixes bug: [YOCTO #2523]
>>>
>>> Signed-off-by: Nitin A Kamble<nitin.a.kamble@intel.com>
>>> ---
>>> meta/recipes-core/eglibc/eglibc-package.inc | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/meta/recipes-core/eglibc/eglibc-package.inc b/meta/recipes-core/eglibc/eglibc-package.inc
>>> index ce37155..423729a 100644
>>> --- a/meta/recipes-core/eglibc/eglibc-package.inc
>>> +++ b/meta/recipes-core/eglibc/eglibc-package.inc
>>> @@ -55,7 +55,6 @@ FILES_${PN}-dbg += "${libexecdir}/*/.debug ${libdir}/audit/.debug"
>>> FILES_catchsegv${PKGSUFFIX} = "${bindir}/catchsegv"
>>> RDEPENDS_catchsegv${PKGSUFFIX} = "libsegfault"
>>> RDEPENDS_${PN}-utils += "bash"
>>> -RDEPENDS_${PN}-mtrace += "perl"
>>> FILES_${PN}-pcprofile = "${base_libdir}/libpcprofile.so"
>>> FILES_eglibc-thread-db${PKGSUFFIX} = "${base_libdir}/libthread_db.so.* ${base_libdir}/libthread_db-*.so"
>>> RPROVIDES_${PN}-dev += "libc-dev"
>>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 10+ messages in thread* Re: [OE-core] [PATCH 1/1] eglibc: remove runtime dependency of perl from eglibc-mtrace
@ 2012-06-11 22:08 ` Darren Hart
0 siblings, 0 replies; 10+ messages in thread
From: Darren Hart @ 2012-06-11 22:08 UTC (permalink / raw)
To: Saul Wold; +Cc: yocto, Patches and discussions about the oe-core layer
On 06/11/2012 03:07 PM, Saul Wold wrote:
> On 06/08/2012 04:02 PM, Darren Hart wrote:
>>
>>
>> On 06/08/2012 03:23 PM, nitin.a.kamble@intel.com wrote:
>>> From: Nitin A Kamble<nitin.a.kamble@intel.com>
>>>
>>> perl needs eglibc to build.
>>> The presence of runtime dependency of
>>> perl for eglibc-mtrace caused bitbake to build perl before eglibc,
>>> which causes build failure of perl with poky-tiny distro
>>>
>>
>> So is this a circular dependency chain?
>>
>> perl DEPENDS on eglibc
>> eglibc (because of eglibc-mtrace) RDEPENDS on perl?
>>
>> If so, doesn't this solution leave eglibc-mtrace with an incomplete set
>> of dependencies in it's final package meta-data?
>>
>> Would the correct solution be to break eglibc-mtrace out into a separate
>> recipe.
>>
> I am not sure if this would be more correct or have the PACKAGES contain
> something like ${EGLIBC_PACKAGE_MTRACE}, which could the be over-ridden
> by the yocto-tiny distro.
>
> EGLIBC_PACKAGE_MTRACE ?= "${PN}-mtrace"
>
> and then change ${PN}-mtrace in the PACKAGES list to
> ${EGLIBC_PACKAGE_MTRACE}.
>
> This exposes another "global", so I am not sure which is better.
This doesn't seem inconsistent with other solutions to similar sorts of
problems. This is certainly less work and a smaller change.
--
Darren
>
> Sau!
>
>
>> eglibc-mtrace.bb could then DEPENDS=eglibc and RDEPENDS=perl and
>> poky-tiny would need to be able to exclude eglibc-mtrace.
>>
>>
>>> This fixes bug: [YOCTO #2523]
>>>
>>> Signed-off-by: Nitin A Kamble<nitin.a.kamble@intel.com>
>>> ---
>>> meta/recipes-core/eglibc/eglibc-package.inc | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/meta/recipes-core/eglibc/eglibc-package.inc b/meta/recipes-core/eglibc/eglibc-package.inc
>>> index ce37155..423729a 100644
>>> --- a/meta/recipes-core/eglibc/eglibc-package.inc
>>> +++ b/meta/recipes-core/eglibc/eglibc-package.inc
>>> @@ -55,7 +55,6 @@ FILES_${PN}-dbg += "${libexecdir}/*/.debug ${libdir}/audit/.debug"
>>> FILES_catchsegv${PKGSUFFIX} = "${bindir}/catchsegv"
>>> RDEPENDS_catchsegv${PKGSUFFIX} = "libsegfault"
>>> RDEPENDS_${PN}-utils += "bash"
>>> -RDEPENDS_${PN}-mtrace += "perl"
>>> FILES_${PN}-pcprofile = "${base_libdir}/libpcprofile.so"
>>> FILES_eglibc-thread-db${PKGSUFFIX} = "${base_libdir}/libthread_db.so.* ${base_libdir}/libthread_db-*.so"
>>> RPROVIDES_${PN}-dev += "libc-dev"
>>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [yocto] [PATCH 1/1] eglibc: remove runtime dependency of perl from eglibc-mtrace
2012-06-11 22:07 ` [OE-core] " Saul Wold
@ 2012-06-11 22:25 ` Khem Raj
-1 siblings, 0 replies; 10+ messages in thread
From: Khem Raj @ 2012-06-11 22:25 UTC (permalink / raw)
To: Saul Wold
Cc: yocto, Darren Hart,
Patches and discussions about the oe-core layer
On Mon, Jun 11, 2012 at 3:07 PM, Saul Wold <sgw@linux.intel.com> wrote:
>> Would the correct solution be to break eglibc-mtrace out into a separate
>> recipe.
>>
> I am not sure if this would be more correct or have the PACKAGES contain
> something like ${EGLIBC_PACKAGE_MTRACE}, which could the be over-ridden by
> the yocto-tiny distro.
>
> EGLIBC_PACKAGE_MTRACE ?= "${PN}-mtrace"
>
> and then change ${PN}-mtrace in the PACKAGES list to
> ${EGLIBC_PACKAGE_MTRACE}.
>
> This exposes another "global", so I am not sure which is better.
there are other tools e.g. memusage etc. which could be built in a
second pass and IMO
we should separate the eglibc-bin into a separate recipe since
everything depends on system libc
its desirable that it should have few dependencies to build it so
builds can be more parallel and
eglibc wont be a bottleneck.
-Khem
^ permalink raw reply [flat|nested] 10+ messages in thread* Re: [OE-core] [PATCH 1/1] eglibc: remove runtime dependency of perl from eglibc-mtrace
@ 2012-06-11 22:25 ` Khem Raj
0 siblings, 0 replies; 10+ messages in thread
From: Khem Raj @ 2012-06-11 22:25 UTC (permalink / raw)
To: Saul Wold
Cc: yocto, Darren Hart,
Patches and discussions about the oe-core layer
On Mon, Jun 11, 2012 at 3:07 PM, Saul Wold <sgw@linux.intel.com> wrote:
>> Would the correct solution be to break eglibc-mtrace out into a separate
>> recipe.
>>
> I am not sure if this would be more correct or have the PACKAGES contain
> something like ${EGLIBC_PACKAGE_MTRACE}, which could the be over-ridden by
> the yocto-tiny distro.
>
> EGLIBC_PACKAGE_MTRACE ?= "${PN}-mtrace"
>
> and then change ${PN}-mtrace in the PACKAGES list to
> ${EGLIBC_PACKAGE_MTRACE}.
>
> This exposes another "global", so I am not sure which is better.
there are other tools e.g. memusage etc. which could be built in a
second pass and IMO
we should separate the eglibc-bin into a separate recipe since
everything depends on system libc
its desirable that it should have few dependencies to build it so
builds can be more parallel and
eglibc wont be a bottleneck.
-Khem
^ permalink raw reply [flat|nested] 10+ messages in thread