| Message ID | 1395583071-10016-1-git-send-email-tremyfr@yahoo.fr |
|---|---|
| State | Superseded |
| Headers | show |
>>>>> "Philippe" == Philippe Reynes <tremyfr@yahoo.fr> writes: > The file external.mk was included before fs/common.mk, > so it was impossible to add rootfs target to external. > This change move fs/common.mk before external.mk, so > now rootfs target may be added to external. > Signed-off-by: Philippe Reynes <tremyfr@yahoo.fr> > --- > Makefile | 3 +-- > 1 files changed, 1 insertions(+), 2 deletions(-) > diff --git a/Makefile b/Makefile > index d49d7bf..07e6c74 100644 > --- a/Makefile > +++ b/Makefile > @@ -391,6 +391,7 @@ include $(sort $(wildcard package/*/*.mk)) > include boot/common.mk > include linux/linux.mk > include system/system.mk > +include fs/common.mk > include $(BR2_EXTERNAL)/external.mk > @@ -408,8 +409,6 @@ ifeq ($(BR2_ECLIPSE_REGISTER),y) > TARGETS+=toolchain-eclipse-register > endif > -include fs/common.mk > - Hmm, does that really work? Does the locale data get correctly purged / generated before the filesystem images are generated?
Hi Peter, I've done a little test before sending this patch. I've used this useless fs : define ROOTFS_USERFS_CMD echo "Nothing to do" ; \ touch $@ endef $(eval $(call ROOTFS_TARGET,userfs)) In the file $(BR2_EXTERNAL)/fs/common.mk Without my patch, the file rootfs.userfs is not created. With my patch, the file rootfs.userfs is created. I could have missed something. Do you have an idea of what may failed ? Regards, Philippe Le Mercredi 26 mars 2014 22h39, Peter Korsgaard <jacmet@uclibc.org> a écrit : >>>>> "Philippe" == Philippe Reynes <tremyfr@yahoo.fr> writes: > The file external.mk was included before fs/common.mk, > so it was impossible to add rootfs target to external. > This change move fs/common.mk before external.mk, so > now rootfs target may be added to external. > Signed-off-by: Philippe Reynes <tremyfr@yahoo.fr> > --- > Makefile | 3 +-- > 1 files changed, 1 insertions(+), 2 deletions(-) > diff --git a/Makefile b/Makefile > index d49d7bf..07e6c74 100644 > --- a/Makefile > +++ b/Makefile > @@ -391,6 +391,7 @@ include $(sort $(wildcard package/*/*.mk)) > include boot/common.mk > include linux/linux.mk > include system/system.mk > +include fs/common.mk > include $(BR2_EXTERNAL)/external.mk > @@ -408,8 +409,6 @@ ifeq ($(BR2_ECLIPSE_REGISTER),y) > TARGETS+=toolchain-eclipse-register > endif > -include fs/common.mk > - Hmm, does that really work? Does the locale data get correctly purged / generated before the filesystem images are generated?
>>>>> "Philippe" == Philippe Reynes <tremyfr@yahoo.fr> writes: > Hi Peter, > I've done a little test before sending this patch. > I've used this useless fs : > define ROOTFS_USERFS_CMD > echo "Nothing to do" ; \ > touch $@ > endef > $(eval $(call ROOTFS_TARGET,userfs)) > In the file $(BR2_EXTERNAL)/fs/common.mk > Without my patch, the file rootfs.userfs is not created. > With my patch, the file rootfs.userfs is created. > I could have missed something. > Do you have an idea of what may failed ? The specific issue I mentioned was about the externally defined rootfs getting created before target-purgelocales / target-generatelocales gets called. The easiest way to check that is to use a glibc toolchain and enable BR2_ENABLE_LOCALE_PURGE / BR2_GENERATE_LOCALE, and then stick an echo statement in each of them.
Hi peter, I think I understand the issue now (thanks to Thomas and Yann). The filesystem target depend on target-finalize, and target-finalize depends on targets. grep -n target-finalize Makefile 513:$(TARGETS_ROOTFS): target-finalize 515:target-finalize: $(TARGETS) So even if we move fs/common.mk above external.mk, the filesystem targets will always be done after targets. So I think that this change should be safe. Best regards, Philippe Le , Peter Korsgaard <jacmet@uclibc.org> a écrit : >>>>> "Philippe" == Philippe Reynes <tremyfr@yahoo.fr> writes: > Hi Peter, > I've done a little test before sending this patch. > I've used this useless fs : > define ROOTFS_USERFS_CMD > echo "Nothing to do" ; \ > touch $@ > endef > $(eval $(call ROOTFS_TARGET,userfs)) > In the file $(BR2_EXTERNAL)/fs/common.mk > Without my patch, the file rootfs.userfs is not created. > With my patch, the file rootfs.userfs is created. > I could have missed something. > Do you have an idea of what may failed ? The specific issue I mentioned was about the externally defined rootfs getting created before target-purgelocales / target-generatelocales gets called. The easiest way to check that is to use a glibc toolchain and enable BR2_ENABLE_LOCALE_PURGE / BR2_GENERATE_LOCALE, and then stick an echo statement in each of them.
Hi Peter, On Thu, Mar 27, 2014 at 10:32 PM, Peter Korsgaard <jacmet@uclibc.org> wrote: >>>>>> "Philippe" == Philippe Reynes <tremyfr@yahoo.fr> writes: > > > Hi Peter, > > I've done a little test before sending this patch. > > I've used this useless fs : > > > define ROOTFS_USERFS_CMD > > echo "Nothing to do" ; \ > > touch $@ > > endef > > > $(eval $(call ROOTFS_TARGET,userfs)) > > > In the file $(BR2_EXTERNAL)/fs/common.mk > > > Without my patch, the file rootfs.userfs is not created. > > With my patch, the file rootfs.userfs is created. > > > I could have missed something. > > Do you have an idea of what may failed ? > > The specific issue I mentioned was about the externally defined rootfs > getting created before target-purgelocales / target-generatelocales gets > called. The easiest way to check that is to use a glibc toolchain and > enable BR2_ENABLE_LOCALE_PURGE / BR2_GENERATE_LOCALE, and then stick an > echo statement in each of them. > I performed this test, and can confirm that all is well: purging/generating of locale data is done before the filesystem is created, in the order generate / purge / rootfs. So moving that line does not impact the behavior with respect to locales. As the patch does no longer cleanly apply, I'll send a new version. Do you still consider this for 2014.05 or is it too late/risky in your opinion? Thanks, Thomas
>>>>> "Thomas" == Thomas De Schampheleire <patrickdepinguin@gmail.com> writes: Hi, >> The specific issue I mentioned was about the externally defined rootfs >> getting created before target-purgelocales / target-generatelocales gets >> called. The easiest way to check that is to use a glibc toolchain and >> enable BR2_ENABLE_LOCALE_PURGE / BR2_GENERATE_LOCALE, and then stick an >> echo statement in each of them. >> > I performed this test, and can confirm that all is well: > purging/generating of locale data is done before the filesystem is > created, in the order generate / purge / rootfs. > So moving that line does not impact the behavior with respect to locales. Thanks! > As the patch does no longer cleanly apply, I'll send a new version. > Do you still consider this for 2014.05 or is it too late/risky in your opinion? If you tested it then I'm Ok with it ;)
diff --git a/Makefile b/Makefile index d49d7bf..07e6c74 100644 --- a/Makefile +++ b/Makefile @@ -391,6 +391,7 @@ include $(sort $(wildcard package/*/*.mk)) include boot/common.mk include linux/linux.mk include system/system.mk +include fs/common.mk include $(BR2_EXTERNAL)/external.mk @@ -408,8 +409,6 @@ ifeq ($(BR2_ECLIPSE_REGISTER),y) TARGETS+=toolchain-eclipse-register endif -include fs/common.mk - TARGETS_SOURCE:=$(patsubst %,%-source,$(TARGETS)) TARGETS_DIRCLEAN:=$(patsubst %,%-dirclean,$(TARGETS))
The file external.mk was included before fs/common.mk, so it was impossible to add rootfs target to external. This change move fs/common.mk before external.mk, so now rootfs target may be added to external. Signed-off-by: Philippe Reynes <tremyfr@yahoo.fr> --- Makefile | 3 +-- 1 files changed, 1 insertions(+), 2 deletions(-)