fakeroot: bump version to 1.21-2

In this version, the configure and Makefile are not provided so we need to generate them using the 'bootstrap' script included in the source.

Hello,

Quoted 2 lines · Maxime Hadjinlian

In this version, the configure and Makefile are not provided so we need to generate them using the 'bootstrap' script included in the source.

This is super annoying.

First, your patch is wrong because the bootstrap script most likely needs autoconf and automake, possibly libtool as well, and you haven't added these to the fakeroot dependencies.

But above all, it means that you would now be adding the build of autoconf and automake to *every* Buildroot build (since essentially host-fakeroot is always built). This is really annoying.

So I'd prefer if we just took the Debian patch that works around the original problem.

Best regards,

Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Quoted 13 lines · Thomas Petazzoni

Hello,

Quoted 2 lines · Maxime Hadjinlian

In this version, the configure and Makefile are not provided so we need to generate them using the 'bootstrap' script included in the source.

This is super annoying.

First, your patch is wrong because the bootstrap script most likely needs autoconf and automake, possibly libtool as well, and you haven't added these to the fakeroot dependencies.

Quoted 6 lines

But above all, it means that you would now be adding the build of autoconf and automake to *every* Buildroot build (since essentially host-fakeroot is always built). This is really annoying.

So I'd prefer if we just took the Debian patch that works around the original problem.

OK, will resend.

Quoted 8 lines

Best regards,

Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

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

I have another reason to bump to 1.21, but experience the same difficulties in getting the configure step right. What would be a sensible way to avoid depending on autoconf/automake?

Hello,

Quoted 3 lines · Andreas Naumann

I have another reason to bump to 1.21, but experience the same difficulties in getting the configure step right. What would be a sensible way to avoid depending on autoconf/automake?

Is your reason to depend on automake/autoconf still the same as what it was in Maxime's patch?

Makefile.am. We normally don't like that very much, but I personally prefer this solution over having fakeroot depending on automake/autoconf.

Best regards,

Thomas

Am 13.03.2017 um 13:32 schrieb Thomas Petazzoni:

Quoted 8 lines

Hello,

Quoted 3 lines · Andreas Naumann

I have another reason to bump to 1.21, but experience the same difficulties in getting the configure step right. What would be a sensible way to avoid depending on autoconf/automake?

Quoted 2 lines

Is your reason to depend on automake/autoconf still the same as what it was in Maxime's patch?

I think so, but in order test it I used FAKEROOT_AUTORECONF instead of running the bootstrap script. However, that does create the same dependency on automake/autoconf right?

If so, then we can patch Makefile.in instead of

Quoted 3 lines

Makefile.am. We normally don't like that very much, but I personally prefer this solution over having fakeroot depending on automake/autoconf.

There is no Makefile.in, only .am and a configure.ac. Wouldnt a patch need to add configure + the contents of build-aux as well? Can you point out a package that I can use as an example?

regards,
Andreas

Quoted 5 lines

Best regards,

Thomas

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

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Hello,

On Mon, 13 Mar 2017 17:33:06 +0100, Yann E. MORIN wrote:

Quoted 6 lines

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Andreas, could you test Yann's patch? Does it fix your problem?

Thanks,

Thomas

Hi,

Am 14.03.2017 um 09:16 schrieb Thomas Petazzoni:

Quoted 4 lines

Hello,

On Mon, 13 Mar 2017 17:33:06 +0100, Yann E. MORIN wrote:

Quoted 8 lines

Quoted 6 lines

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Andreas, could you test Yann's patch? Does it fix your problem?

I just did, but unfortunately it does not help. However, even with plain 1.21 I no longer get the correct behaviour. Only thing that helps is going back to 1.18.4.
It was late Friday when I thought 1.21 was working, but I may have tested an Ubuntu version of it (1.21-3.1 or so). I will try to track that down.

regards,
Andreas

Quoted 5 lines

Thanks,

Thomas

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

Hi again,

I cannot confirm that fakeroot 1.21 solves the problem. What works is going back to 1.18.4 or using the packaged fakeroot of my host (1.20-3ubuntu2).

To describe the issue again, the problem is doing 'cp -a' in the fakeroot environment (genimage does this, so I cant easily change this). However, I'm not so sure if what so far worked is to be expected. After all, -a means --preserve=all.

I have a simple test, running the following script under fakeroot after building a defconfig, which provides a rootfs.tar:
rm -rf .tmpdir* rootfs.2.tar
mkdir .tmpdir
tar -xzf output/images/rootfs.tar -C .tmpdir
cp -a .tmpdir .tmpdir2
tar -cf rootfs.2.tar -C .tmpdir2 .

When done, a
tar -tvf rootfs.2.tar
shows changed permissions for everything and the original owner for files:

drwx------ root/root         0 2017-03-14 13:37 ./
drwx------ root/root         0 2016-10-06 09:03 ./mnt/
drwxrwxrwt root/root         0 2017-03-10 14:44 ./tmp/
drwx------ root/root         0 2017-03-14 13:37 ./tmp/ldconfig/
-rw------- joe/somegroup     33 2017-03-14 13:37 ./tmp/ldconfig/aux-cache

drwx------ root/root 0 2017-03-10 14:35 ./tmp/fontconfig/
...

I tried to apply the patches that Ubuntu has for the host package, but that doesnt help either. What else can I do? Can anybody shed some light on this?

thanks,
Andreas

Am 14.03.2017 um 12:04 schrieb Andreas Naumann:

Quoted 28 lines

Hi,

Am 14.03.2017 um 09:16 schrieb Thomas Petazzoni:

Quoted 4 lines

Hello,

On Mon, 13 Mar 2017 17:33:06 +0100, Yann E. MORIN wrote:

Quoted 8 lines

Quoted 6 lines

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Andreas, could you test Yann's patch? Does it fix your problem?

I just did, but unfortunately it does not help. However, even with plain 1.21 I no longer get the correct behaviour. Only thing that helps is going back to 1.18.4.
It was late Friday when I thought 1.21 was working, but I may have tested an Ubuntu version of it (1.21-3.1 or so). I will try to track that down.

regards,
Andreas

Quoted 1 line

Quoted 9 lines

Quoted 4 lines

Thanks,

Thomas

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

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

For the record, I can reproduce this on my side. I just do "fakeroot
/bin/bash", then do the commands from the shell. Even an ls -l of the copied files within the fakeroot shell already shows the "arnout" instead of "root" user. My host fakeroot (Debian 1.21-3.1).

I have also tried removing the ac_cv_header_sys_capability_h=y, that doesn't
make a difference.

I have not yet tried applying the Debian patches to the Buildroot-built fakeroot.

Regards,
Arnout

Quoted 68 lines

When done, a
tar -tvf rootfs.2.tar
shows changed permissions for everything and the original owner for files:

drwx------ root/root         0 2017-03-14 13:37 ./
drwx------ root/root         0 2016-10-06 09:03 ./mnt/
drwxrwxrwt root/root         0 2017-03-10 14:44 ./tmp/
drwx------ root/root         0 2017-03-14 13:37 ./tmp/ldconfig/
-rw------- joe/somegroup     33 2017-03-14 13:37 ./tmp/ldconfig/aux-cache

drwx------ root/root 0 2017-03-10 14:35 ./tmp/fontconfig/
...

I tried to apply the patches that Ubuntu has for the host package, but that doesnt help either. What else can I do? Can anybody shed some light on this?

thanks,
Andreas

Am 14.03.2017 um 12:04 schrieb Andreas Naumann:

Quoted 28 lines

Hi,

Am 14.03.2017 um 09:16 schrieb Thomas Petazzoni:

Quoted 4 lines

Hello,

On Mon, 13 Mar 2017 17:33:06 +0100, Yann E. MORIN wrote:

Quoted 8 lines

Quoted 6 lines

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Andreas, could you test Yann's patch? Does it fix your problem?

I just did, but unfortunately it does not help. However, even with plain 1.21 I no longer get the correct behaviour. Only thing that helps is going back to 1.18.4.
It was late Friday when I thought 1.21 was working, but I may have tested an Ubuntu version of it (1.21-3.1 or so). I will try to track that down.

regards,
Andreas

Quoted 1 line

Quoted 9 lines

Quoted 4 lines

Thanks,

Thomas

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

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

Hi Arnout,

Am 15.03.2017 um 01:20 schrieb Arnout Vandecappelle:

Quoted 6 lines

For the record, I can reproduce this on my side. I just do "fakeroot
/bin/bash", then do the commands from the shell. Even an ls -l of the copied files within the fakeroot shell already shows the "arnout" instead of "root" user. My host fakeroot (Debian 1.21-3.1).

So are you saying you see this behaviour with your host fakeroot?

Quoted 7 lines

I have also tried removing the ac_cv_header_sys_capability_h=y, that doesn't
make a difference.

I have not yet tried applying the Debian patches to the Buildroot-built
fakeroot.

I have done this with the Ubuntu patches for 1.20.2 and Debian patches for 1.21-3.1 to no avail.

thanks and regards,
Andreas

Quoted 73 lines

Regards,
Arnout

Quoted 68 lines

When done, a
tar -tvf rootfs.2.tar
shows changed permissions for everything and the original owner for files:

drwx------ root/root         0 2017-03-14 13:37 ./
drwx------ root/root         0 2016-10-06 09:03 ./mnt/
drwxrwxrwt root/root         0 2017-03-10 14:44 ./tmp/
drwx------ root/root         0 2017-03-14 13:37 ./tmp/ldconfig/
-rw------- joe/somegroup     33 2017-03-14 13:37 ./tmp/ldconfig/aux-cache

drwx------ root/root 0 2017-03-10 14:35 ./tmp/fontconfig/
...

I tried to apply the patches that Ubuntu has for the host package, but that doesnt help either. What else can I do? Can anybody shed some light on this?

thanks,
Andreas

Am 14.03.2017 um 12:04 schrieb Andreas Naumann:

Quoted 28 lines

Hi,

Am 14.03.2017 um 09:16 schrieb Thomas Petazzoni:

Quoted 4 lines

Hello,

On Mon, 13 Mar 2017 17:33:06 +0100, Yann E. MORIN wrote:

Quoted 8 lines

Quoted 6 lines

I had a look at the diff between 1.20.2 (which we have right now) and 1.21, and the only noticeable delta is the change of a bufgfer size in mesage.h, the rest of the delta being the removal of the generated autostuff.

So we can quite safely backport that single change.

Andreas, could you test Yann's patch? Does it fix your problem?

I just did, but unfortunately it does not help. However, even with plain 1.21 I no longer get the correct behaviour. Only thing that helps is going back to 1.18.4.
It was late Friday when I thought 1.21 was working, but I may have tested an Ubuntu version of it (1.21-3.1 or so). I will try to track that down.

regards,
Andreas

Quoted 1 line

Quoted 9 lines

Quoted 4 lines

Thanks,

Thomas

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

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

Quoted 9 lines · Andreas Naumann

Hi Arnout,

Am 15.03.2017 um 01:20 schrieb Arnout Vandecappelle:

Quoted 6 lines

For the record, I can reproduce this on my side. I just do "fakeroot
/bin/bash", then do the commands from the shell. Even an ls -l of the copied files within the fakeroot shell already shows the "arnout" instead of "root" user. My host fakeroot (Debian 1.21-3.1).

Quoted 1 line

So are you saying you see this behaviour with your host fakeroot?

That was only half a sentence, it should have been:

My host fakeroot (Debian 1.21-3.1) doesn't show the problem.

Buildroot fakeroot works fine when I have libacl1-dev installed on my system -
the Debian package has a build dependency on libacl1-dev. So we would need to add host-acl to FAKEROOT_DEPENDENCIES. That also pulls in host-attr. On my system, it brings the time for 'make host-fakeroot' on a clean output from

real    0m12.230s
user    0m8.500s
sys     0m1.132s

to

real    0m21.945s
user    0m17.724s
sys     0m2.948s

so a significant slowdown... Although in absolute numbers it is just 10 seconds, so maybe it is acceptable.

Anyway, I'll send a patch that does this, let's see if others consider this
overhead acceptable.

Regards,
Arnout

Signature · 6 lines

Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF

Quoted 1 line

Quoted 1 line

Quoted 1 line

Quoted 1 line

"Yann" == Yann E MORIN <[email protected]> writes:

Arnout, All,
On 2017-03-15 20:46 +0100, Arnout Vandecappelle spake thusly:

Quoted 1 line · Arnout Vandecappelle

[--SNIP--]

Quoted 15 lines

Buildroot fakeroot works fine when I have libacl1-dev installed on my system - the Debian package has a build dependency on libacl1-dev. So we would need to add host-acl to FAKEROOT_DEPENDENCIES. That also pulls in host-attr. On my system, it brings the time for 'make host-fakeroot' on a clean output from

real    0m12.230s
user    0m8.500s
sys     0m1.132s

to

real    0m21.945s
user    0m17.724s
sys     0m2.948s

Quoted 2 lines

so a significant slowdown... Although in absolute numbers it is just 10 seconds, so maybe it is acceptable.

It's either correctness and a bit of overhead, or inconsistencies.

I know where my vote lies: correctness. ;-)

Quoted 1 line

I know where my vote lies: correctness. ;-)

Yes, naturally. I vaguely remember something about an earlier discussion related to host-acl/host-attr, but that was perhaps about pseudo?

Quoted 1 line

Quoted 1 line

Quoted 1 line

Quoted 1 line

"Thomas" == Thomas Petazzoni <[email protected]> writes:

Hello,
On Wed, 15 Mar 2017 21:41:23 +0100, Peter Korsgaard wrote:

Quoted 4 lines

Quoted 1 line

I know where my vote lies: correctness. ;-)

Yes, naturally. I vaguely remember something about an earlier discussion related to host-acl/host-attr, but that was perhaps about pseudo?

I think it was about host-libcap, not host-acl/host-attr. See commit c994c3db1f7d76b5dfe79e3b1133d66fdd06212c.

Arnout, All,

On 2017-03-15 20:46 +0100, Arnout Vandecappelle spake thusly:

Quoted 1 line · Arnout Vandecappelle

[--SNIP--]

Quoted 1 line

Buildroot fakeroot works fine when I have libacl1-dev installed on my system -

It's either correctness and a bit of overhead, or inconsistencies.

I know where my vote lies: correctness. ;-)

Regards,
Yann E. MORIN.

Quoted 2 lines

Anyway, I'll send a patch that does this, let's see if others consider this
overhead acceptable.

Quoted 11 lines

Regards,
Arnout

--

Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF

Hi Arnout,

Am 15.03.2017 um 21:09 schrieb Yann E. MORIN:

Quoted 6 lines

Arnout, All,

On 2017-03-15 20:46 +0100, Arnout Vandecappelle spake thusly:

Quoted 1 line · Arnout Vandecappelle

[--SNIP--]

Quoted 1 line

Buildroot fakeroot works fine when I have libacl1-dev installed on my system -

This is great! And of course, I dont mind the 10 seconds :-)

[snip]

Quoted 3 lines

It's either correctness and a bit of overhead, or inconsistencies.

Quoted 7 lines

I know where my vote lies: correctness. ;-)

Regards,
Yann E. MORIN.

Quoted 2 lines

Anyway, I'll send a patch that does this, let's see if others consider this
overhead acceptable.

I'll test as soon as its there.