[Buildroot] uClibc, gcc get rebuilt on every make

[Buildroot] uClibc, gcc get rebuilt on every make

[Patrick McNeil]

I'm using svn24232, arm920t, with gcc and headers on the target. Every  
time I run make, even if I haven't made any changes, the cross  
compiler toolchain, as well as the target gcc, and uClibc gets  
rebuilt, taking about half an hour. Any ideas what this is depending  
on? I know that buildroot from a few months ago did not do this.

Also, what is the proper way to rebuild root if I have changed  
something - ie changed files in target_skeleton, updates something in  
a package directory, etc.? Right now I run this script and it seems to  
work ok. If I don't delete /root and .root, the system doesn't seem to  
realize that things have changed.


#!/bin/bash

[ -e project_build_arm/phidget_sbc/.root ] && {
	echo "Removing Root..."
	rm -rf project_build_arm/phidget_sbc/root/
	rm project_build_arm/phidget_sbc/.root
}
[ -e build_arm/webif-0.2/.built ] && {
	echo "Removing webif..."
	rm build_arm/webif-0.2/.built
}
for stamp in $( find build_arm/ | grep .stamp_staging_installed ); do
	echo "Removing $stamp..."
	rm $stamp
done
for stamp in $( find project_build_arm/phidget_sbc/autotools-stamps - 
type f ); do
	echo "Removing $stamp..."
	rm $stamp
done

--
Patrick McNeil
Lead Developer | Phidgets Inc.
patrick at phidgets.com | www.phidgets.com
--
There are two ways to write error-free programs; only the third works.
-Alan J. Perlis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://busybox.net/lists/buildroot/attachments/20081209/2686a70e/attachment.htm 

[Buildroot] uClibc, gcc get rebuilt on every make

[Zac Wheeler]

On Tue, Dec 9, 2008 at 10:23 AM, Patrick McNeil <patrick@phidgets.com>wrote:

> I'm using svn24232, arm920t, with gcc and headers on the target. Every time
> I run make, even if I haven't made any changes, the cross compiler
> toolchain, as well as the target gcc, and uClibc gets rebuilt, taking about
> half an hour. Any ideas what this is depending on? I know that buildroot
> from a few months ago did not do this.
>

Just in case the issue is still there, this is what I did to resolve it:

diff --git a/toolchain/uClibc/uclibc.mk b/toolchain/uClibc/uclibc.mk
index c8da464..6dff66c 100644
--- a/toolchain/uClibc/uclibc.mk
+++ b/toolchain/uClibc/uclibc.mk
@@ -481,7 +482,7 @@ uclibc-menuconfig: host-sed $(UCLIBC_DIR)/.config
        touch -c $(UCLIBC_DIR)/.config


-$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/lib/libc.a
+$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/.configured $(gcc_initial)
$(LIBFLOAT_TARGET)
 ifneq ($(BR2_TOOLCHAIN_SYSROOT),y)
        $(MAKE1) -C $(UCLIBC_DIR) \
                PREFIX= \

This also fixes the issue you sent in another mail where libc.a would get
modified after the initial build and so simply typing "make" again was
enough to break symbols in libc.a (e.g. python's division).

The better way would probably to figure out what is touching
$(UCLIBC_DIR)/lib/libc.a after the initial build and fix that so that make
doesn't see a dependency that is newer than the
$(STAGING_DIR)/usr/lib/libc.a target, but I'm not sure that is easily
possible given the build sequence.

Zac
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20090702/ede55513/attachment.htm>

[Buildroot] uClibc, gcc get rebuilt on every make

[Bernhard Reutner-Fischer]

On Thu, Jul 02, 2009 at 01:38:23PM -0700, Zac Wheeler wrote:
>On Tue, Dec 9, 2008 at 10:23 AM, Patrick McNeil <patrick@phidgets.com>wrote:
>
>> I'm using svn24232, arm920t, with gcc and headers on the target. Every time
>> I run make, even if I haven't made any changes, the cross compiler
>> toolchain, as well as the target gcc, and uClibc gets rebuilt, taking about
>> half an hour. Any ideas what this is depending on? I know that buildroot
>> from a few months ago did not do this.
>>
>
>Just in case the issue is still there, this is what I did to resolve it:
>
>diff --git a/toolchain/uClibc/uclibc.mk b/toolchain/uClibc/uclibc.mk
>index c8da464..6dff66c 100644
>--- a/toolchain/uClibc/uclibc.mk
>+++ b/toolchain/uClibc/uclibc.mk
>@@ -481,7 +482,7 @@ uclibc-menuconfig: host-sed $(UCLIBC_DIR)/.config
>        touch -c $(UCLIBC_DIR)/.config
>
>
>-$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/lib/libc.a
>+$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/.configured $(gcc_initial)

This doesn't sounds clean to me. My current working theory:
http://repo.or.cz/w/buildroot.git?a=blob;f=toolchain/uClibc/uclibc.mk;h=5b4e0c6a02daff5a5c498a2d277eb628ca15aad5;hb=HEAD#l509
http://repo.or.cz/w/buildroot.git?a=blob;f=toolchain/uClibc/uclibc.mk;h=5b4e0c6a02daff5a5c498a2d277eb628ca15aad5;hb=HEAD#l550

Again, i invite people to port improvements from my tree to the repo..

HTH,
>$(LIBFLOAT_TARGET)
> ifneq ($(BR2_TOOLCHAIN_SYSROOT),y)
>        $(MAKE1) -C $(UCLIBC_DIR) \
>                PREFIX= \
>
>This also fixes the issue you sent in another mail where libc.a would get
>modified after the initial build and so simply typing "make" again was
>enough to break symbols in libc.a (e.g. python's division).
>
>The better way would probably to figure out what is touching
>$(UCLIBC_DIR)/lib/libc.a after the initial build and fix that so that make
>doesn't see a dependency that is newer than the
>$(STAGING_DIR)/usr/lib/libc.a target, but I'm not sure that is easily
>possible given the build sequence.
>
>Zac

>_______________________________________________
>buildroot mailing list
>buildroot at busybox.net
>http://lists.busybox.net/mailman/listinfo/buildroot

[Buildroot] uClibc, gcc get rebuilt on every make

[Zac Wheeler]

On Thu, Jul 2, 2009 at 2:07 PM, Bernhard Reutner-Fischer <
rep.dot.nop@gmail.com> wrote:

> On Thu, Jul 02, 2009 at 01:38:23PM -0700, Zac Wheeler wrote:
> >On Tue, Dec 9, 2008 at 10:23 AM, Patrick McNeil <patrick@phidgets.com
> >wrote:
> >
> >> I'm using svn24232, arm920t, with gcc and headers on the target. Every
> time
> >> I run make, even if I haven't made any changes, the cross compiler
> >> toolchain, as well as the target gcc, and uClibc gets rebuilt, taking
> about
> >> half an hour. Any ideas what this is depending on? I know that buildroot
> >> from a few months ago did not do this.
> >>
> >
> >Just in case the issue is still there, this is what I did to resolve it:
> >
> >diff --git a/toolchain/uClibc/uclibc.mk b/toolchain/uClibc/uclibc.mk
> >index c8da464..6dff66c 100644
> >--- a/toolchain/uClibc/uclibc.mk
> >+++ b/toolchain/uClibc/uclibc.mk
> >@@ -481,7 +482,7 @@ uclibc-menuconfig: host-sed $(UCLIBC_DIR)/.config
> >        touch -c $(UCLIBC_DIR)/.config
> >
> >
> >-$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/lib/libc.a
> >+$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/.configured $(gcc_initial)
>
> This doesn't sounds clean to me. My current working theory:
>
> http://repo.or.cz/w/buildroot.git?a=blob;f=toolchain/uClibc/uclibc.mk;h=5b4e0c6a02daff5a5c498a2d277eb628ca15aad5;hb=HEAD#l509
>
> http://repo.or.cz/w/buildroot.git?a=blob;f=toolchain/uClibc/uclibc.mk;h=5b4e0c6a02daff5a5c498a2d277eb628ca15aad5;hb=HEAD#l550
>

Cool, I didn't know about order-only prerequisites. Or are the additional CC
and FLAGS options important for fixing this issue as well?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20090702/a555cc23/attachment.htm>

[Buildroot] uClibc, gcc get rebuilt on every make

[Bernhard Reutner-Fischer]

On Thu, Jul 02, 2009 at 04:37:04PM -0700, Zac Wheeler wrote:
>On Thu, Jul 2, 2009 at 2:07 PM, Bernhard Reutner-Fischer <

>Cool, I didn't know about order-only prerequisites. Or are the additional CC
>and FLAGS options important for fixing this issue as well?

No, they are just a matter of correctness and a different issue.

[Buildroot] uClibc, gcc get rebuilt on every make

[bogus]

It would be great if you could submit a little patch...

For now, I keep using the first (not clean) option, no time to dive into
the second one right now.


Regards,



Simon


>Le jeu, 02 Jul 2009 ? 04:37:04PM -0700, Zac Wheeler ?crivit:

>Cool, I didn't know about order-only prerequisites. Or are the
>additional CC
>and FLAGS options important for fixing this issue as well?

No, they are just a matter of correctness and a different issue.

[Buildroot] uClibc, gcc get rebuilt on every make

[Zac Wheeler]

On Sun, Jul 5, 2009 at 6:12 AM, <simon.addareyss@online.fr> wrote:

> It would be great if you could submit a little patch...
>
> For now, I keep using the first (not clean) option, no time to dive into
> the second one right now.
>

This is it - just a pipe before the libc.a dependency to make it order-only:

diff --git a/toolchain/uClibc/uclibc.mk b/toolchain/uClibc/uclibc.mk
index 6dff66c..f9eef88 100644
--- a/toolchain/uClibc/uclibc.mk
+++ b/toolchain/uClibc/uclibc.mk
@@ -482,7 +482,7 @@ uclibc-menuconfig: host-sed $(UCLIBC_DIR)/.config
        touch -c $(UCLIBC_DIR)/.config


-$(STAGING_DIR)/usr/lib/libc.a: $(UCLIBC_DIR)/lib/libc.a
+$(STAGING_DIR)/usr/lib/libc.a: | $(UCLIBC_DIR)/lib/libc.a
 ifneq ($(BR2_TOOLCHAIN_SYSROOT),y)
        $(MAKE1) -C $(UCLIBC_DIR) \
                PREFIX= \
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20090705/8a34d5e5/attachment.htm>