[Buildroot] Autobuilder site back online

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Hello,

As many of you have noticed, the autobuild.buildroot.org web site has
been down for a number of days. It was hosted on a Virtual Private
Server that I have been using since 10+ years, but it was getting short
on disk space, was not very powerful, and more importantly, I did not
have access to the host machine, which means that whenever something
bad/wrong was happening, there was nothing I could do, except rely on
people who didn't always had enough time to take care of the issue in a
timely fashion.

Therefore, I have migrated the service to a dedicated server that I
own. I have direct access to the support team, I can reboot the server,
access the logs. And this server has more disk space (1 TB) and is much
more powerful than the previous one.

This new service is temporarily available at:

	http://autobuild.humanoidz.org

The build results from the Free Electrons build server have already
started to flow into this new infrastructure.

There are a few things that remain to be done:

 * Changing the autobuild.buildroot.org to point to this new server.
   I've already asked Peter to do so.

 * Changing the configuration of the build servers maintained by Peter
   so that they send their results to the new infrastructure. I've also
   asked Peter to do so.

 * Re-add the daily e-mails to the new infrastructure. However, since I
   couldn't access the old server to turn them off, the old server will
   start sending them (empty probably when it goes up). So as soon as
   I'll have access to the old server to turn it off, I will migrate
   the daily e-mails to the new infrastructure.

 * Import the history of build results. This also requires me to have
   access to the old server and transfer all the results to the new
   server. It might take a bit of time.

Sorry again for the noise. At least, the build results are back up,
which is the most important during this -rc cycle.

Best regards,

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

[Buildroot] Autobuilder site back online

[Thomas De Schampheleire]

Hi Thomas,

On Sat, Nov 16, 2013 at 10:36 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Hello,
>
> As many of you have noticed, the autobuild.buildroot.org web site has
> been down for a number of days. It was hosted on a Virtual Private
> Server that I have been using since 10+ years, but it was getting short
> on disk space, was not very powerful, and more importantly, I did not
> have access to the host machine, which means that whenever something
> bad/wrong was happening, there was nothing I could do, except rely on
> people who didn't always had enough time to take care of the issue in a
> timely fashion.
>
> Therefore, I have migrated the service to a dedicated server that I
> own. I have direct access to the support team, I can reboot the server,
> access the logs. And this server has more disk space (1 TB) and is much
> more powerful than the previous one.
>
> This new service is temporarily available at:
>
>         http://autobuild.humanoidz.org
>
> The build results from the Free Electrons build server have already
> started to flow into this new infrastructure.
>
> There are a few things that remain to be done:
>
>  * Changing the autobuild.buildroot.org to point to this new server.
>    I've already asked Peter to do so.
>
>  * Changing the configuration of the build servers maintained by Peter
>    so that they send their results to the new infrastructure. I've also
>    asked Peter to do so.
>
>  * Re-add the daily e-mails to the new infrastructure. However, since I
>    couldn't access the old server to turn them off, the old server will
>    start sending them (empty probably when it goes up). So as soon as
>    I'll have access to the old server to turn it off, I will migrate
>    the daily e-mails to the new infrastructure.
>
>  * Import the history of build results. This also requires me to have
>    access to the old server and transfer all the results to the new
>    server. It might take a bit of time.
>
> Sorry again for the noise. At least, the build results are back up,
> which is the most important during this -rc cycle.

That is all great to hear, thanks a lot for taking these actions!

Best regards,
Thomas

[Buildroot] Autobuilder site back online

[Yann E. MORIN]

Thomas, All,

On 2013-11-16 22:36 +0100, Thomas Petazzoni spake thusly:
[--SNIP--]
> Therefore, I have migrated the service to a dedicated server that I
> own. I have direct access to the support team, I can reboot the server,
> access the logs. And this server has more disk space (1 TB) and is much
> more powerful than the previous one.
> 
> This new service is temporarily available at:
> 
> 	http://autobuild.humanoidz.org

Here are a few analysis of some of the build failures:

http://autobuild.humanoidz.org/results/a60/a60a63163288af14f701adfdfa9f62229c983984/
  -> unable to download the toolchain from mentor.com
  -> I was able to download the toolchain OK from here
  -> temporary glitch in network or on the server?

http://autobuild.humanoidz.org/results/889/889b3870c30d027af1aedb843648c4ddb7ed79f0/
  -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/microblaze-linux-gcc'

http://autobuild.humanoidz.org/results/33f/33ff18fef818bba47eedaff2662142c25b53b4da/
  -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/aarch64-linux-gcc'

http://autobuild.humanoidz.org/results/6aa/6aa47858a511e89b0bd6a3a472276baab65afb1c/
  -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/nios2-linux-gcc'

For the last three failures above, can it be that the .config file is
improperly tweaked by the test framework?

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] Autobuilder site back online

[Yann E. MORIN]

Thomas, All,

On 2013-11-17 18:29 +0100, Yann E. MORIN spake thusly:
> Here are a few analysis of some of the build failures:

Here are a few more (note: I'm only doing analysis for now):

http://autobuild.humanoidz.org/results/5f7/5f7003615207adc737c482edc8271557d9f2ce79/
  -> valloc missing
  -> config is using uClibc snapshot
  -> since e806570 (2013-04-04), valloc is hidden behind SUSV2_LEGACY
     which is missing from our uclibc-snapshot.config

http://autobuild.humanoidz.org/results/6d4/6d4c476421b2a2620c52b4cfc6189f65325ee026/
  -> error: conversion to 'unsigned int' from 'long unsigned int' may
     alter its value [-Werror=conversion]
  -> cc1plus: all warnings being treated as errors
  -> ola code is not 64-bit clean

http://autobuild.humanoidz.org/results/95f/95fe1ed8b2e51e7eb1712958d9e4d891bb38e35b/
  -> gzip: /home/test/test/2/dl/aiccu_20070115.tar.gz: not in gzip format
  -> the file is 6.4KiB in size, which is wrong
  -> downloaded from here, the file is 68KiB, and is a correct .tar.gz
  -> spurious network/server error?

http://autobuild.humanoidz.org/results/a50/a5078a8759242ac8198bf55fe6c61da3c793522c/
  -> error: #warning requested reentrant code, but thread support was
     disabled [-Werror=cpp]
  -> but BR2_TOOLCHAIN_EXTERNAL_HAS_THREADS is not set
  -> libpfm4-4.3.0 needs threads?

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] Autobuilder site back online

[Peter Korsgaard]

>>>>> "Yann" == Yann E MORIN <yann.morin.1998@free.fr> writes:

Hi,

 > Here are a few analysis of some of the build failures:

 > http://autobuild.humanoidz.org/results/a60/a60a63163288af14f701adfdfa9f62229c983984/
 -> unable to download the toolchain from mentor.com
 -> I was able to download the toolchain OK from here
 -> temporary glitch in network or on the server?

It also works here, so I guess so.

 > http://autobuild.humanoidz.org/results/6aa/6aa47858a511e89b0bd6a3a472276baab65afb1c/
 -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/nios2-linux-gcc'

 > For the last three failures above, can it be that the .config file is
 > improperly tweaked by the test framework?

It seems like it. This /path/to/toolchain/usr is the default path when a
custom external toolchain is used.

-- 
Bye, Peter Korsgaard

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Dear Yann E. MORIN,

On Sun, 17 Nov 2013 18:29:17 +0100, Yann E. MORIN wrote:

> Here are a few analysis of some of the build failures:
> 
> http://autobuild.humanoidz.org/results/a60/a60a63163288af14f701adfdfa9f62229c983984/
>   -> unable to download the toolchain from mentor.com
>   -> I was able to download the toolchain OK from here
>   -> temporary glitch in network or on the server?

Appears to be a temporary glitch, because I've just tested, and I'm
able to wget this URL from the build server right now.

> http://autobuild.humanoidz.org/results/889/889b3870c30d027af1aedb843648c4ddb7ed79f0/
>   -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/microblaze-linux-gcc'
> 
> http://autobuild.humanoidz.org/results/33f/33ff18fef818bba47eedaff2662142c25b53b4da/
>   -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/aarch64-linux-gcc'
> 
> http://autobuild.humanoidz.org/results/6aa/6aa47858a511e89b0bd6a3a472276baab65afb1c/
>   -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/nios2-linux-gcc'
> 
> For the last three failures above, can it be that the .config file is
> improperly tweaked by the test framework?

That's really weird. If you look at the config files of all of those
builds, they have BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y, and
BR2_TOOLCHAIN_EXTERNAL_PREINSTALLED=y. Which doesn't make any sense
when you look at the defconfig used as the base for those
configurations:

  http://autobuild.buildroot.org/toolchains/configs/free-electrons/microblazeel-unknown-linux-gnu.config
  http://autobuild.buildroot.org/toolchains/configs/free-electrons/linaro-aarch64.config
  http://autobuild.buildroot.org/toolchains/configs/free-electrons/sourcery-nios2.config

Take any of these base configuration, run 'make menuconfig' on them,
and you will definitely not have BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y
defined, but instead the proper usage of an existing toolchain profile.

Moreover, there have been some successful builds of Microblaze, AArch64
and NIOS 2 later on. Well, the builds were not necessarily successful
until the end, but they built some packages, which means that the
toolchain was successfully installed. See for example:

  Aarch64
  http://autobuild.buildroot.org/results/450/4506a24e72505949772ce61745d78c7b5754c93e/build-end.log

  Microblaze
  http://autobuild.buildroot.org/results/c7e/c7e5797d0c1e5164ead67c0dcf1d2ab293fb15e7/build-end.log

  NIOS 2
  http://autobuild.buildroot.org/results/351/35108c3a4a128371a55fdda13f14c032bc118c3a/build-end.log

And for those three architectures, there is only one defconfig for
each, so it really means that those builds could only originate from
the same base configuration as the builds who failed in a bizarre way.

So, it was failing for some time, and is now "fixed". But I don't see
any commit today that could be causing this.

Ideas?

Best regards,

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

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Dear Yann E. MORIN,

On Sun, 17 Nov 2013 18:50:35 +0100, Yann E. MORIN wrote:

> Here are a few more (note: I'm only doing analysis for now):
> 
> http://autobuild.humanoidz.org/results/5f7/5f7003615207adc737c482edc8271557d9f2ce79/
>   -> valloc missing
>   -> config is using uClibc snapshot
>   -> since e806570 (2013-04-04), valloc is hidden behind SUSV2_LEGACY
>      which is missing from our uclibc-snapshot.config

This has been discussed recently. See
http://lists.busybox.net/pipermail/buildroot/2013-November/082106.html
and the following e-mails.

> http://autobuild.humanoidz.org/results/95f/95fe1ed8b2e51e7eb1712958d9e4d891bb38e35b/
>   -> gzip: /home/test/test/2/dl/aiccu_20070115.tar.gz: not in gzip format
>   -> the file is 6.4KiB in size, which is wrong
>   -> downloaded from here, the file is 68KiB, and is a correct .tar.gz
>   -> spurious network/server error?

File removed from the server.

> http://autobuild.humanoidz.org/results/a50/a5078a8759242ac8198bf55fe6c61da3c793522c/
>   -> error: #warning requested reentrant code, but thread support was
>      disabled [-Werror=cpp]
>   -> but BR2_TOOLCHAIN_EXTERNAL_HAS_THREADS is not set
>   -> libpfm4-4.3.0 needs threads?

Only dependency is oprofile on PowerPC, so adding threads as a
dependency doesn't seem horrible :)

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

[Buildroot] Autobuilder site back online

[Arnout Vandecappelle]

On 17/11/13 23:38, Thomas Petazzoni wrote:
> Dear Yann E. MORIN,
>
> On Sun, 17 Nov 2013 18:29:17 +0100, Yann E. MORIN wrote:
>
[snip]
>> http://autobuild.humanoidz.org/results/889/889b3870c30d027af1aedb843648c4ddb7ed79f0/
>>    -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/microblaze-linux-gcc'
>>
>> http://autobuild.humanoidz.org/results/33f/33ff18fef818bba47eedaff2662142c25b53b4da/
>>    -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/aarch64-linux-gcc'
>>
>> http://autobuild.humanoidz.org/results/6aa/6aa47858a511e89b0bd6a3a472276baab65afb1c/
>>    -> Cannot execute cross-compiler '/path/to/toolchain/usr/bin/nios2-linux-gcc'
>>
>> For the last three failures above, can it be that the .config file is
>> improperly tweaked by the test framework?
>
> That's really weird. If you look at the config files of all of those
> builds, they have BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y, and
> BR2_TOOLCHAIN_EXTERNAL_PREINSTALLED=y. Which doesn't make any sense
> when you look at the defconfig used as the base for those
> configurations:
>
>    http://autobuild.buildroot.org/toolchains/configs/free-electrons/microblazeel-unknown-linux-gnu.config
>    http://autobuild.buildroot.org/toolchains/configs/free-electrons/linaro-aarch64.config
>    http://autobuild.buildroot.org/toolchains/configs/free-electrons/sourcery-nios2.config
>
> Take any of these base configuration, run 'make menuconfig' on them,
> and you will definitely not have BR2_TOOLCHAIN_EXTERNAL_CUSTOM=y
> defined, but instead the proper usage of an existing toolchain profile.

  I read:

config BR2_TOOLCHAIN_EXTERNAL_XILINX_MICROBLAZEEL_14_3
         bool "Xilinx Little Endian Microblaze GNU Tools 14.3"
         depends on BR2_microblazeel
         depends on BR2_HOSTARCH = "x86_64" || BR2_HOSTARCH = "x86"
         depends on !BR2_PREFER_STATIC_LIB

  Is it possible that the autobuilder scripts randomly set 
BR2_PREFER_STATIC_LIB?  If that is set, then none of the official 
external toolchain profiles work and the default will be a custom 
preinstalled one...

  Regards,
  Arnout

[snip]
-- 
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:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Dear Arnout Vandecappelle,

On Mon, 18 Nov 2013 13:55:20 +0100, Arnout Vandecappelle wrote:

>   I read:
> 
> config BR2_TOOLCHAIN_EXTERNAL_XILINX_MICROBLAZEEL_14_3
>          bool "Xilinx Little Endian Microblaze GNU Tools 14.3"
>          depends on BR2_microblazeel
>          depends on BR2_HOSTARCH = "x86_64" || BR2_HOSTARCH = "x86"
>          depends on !BR2_PREFER_STATIC_LIB
> 
>   Is it possible that the autobuilder scripts randomly set 
> BR2_PREFER_STATIC_LIB?  If that is set, then none of the official 
> external toolchain profiles work and the default will be a custom 
> preinstalled one...

Aaah, you got it! Indeed the autobuilder randomly sets
BR2_PREFER_STATIC_LIB, and this isn't going to work properly when the
external toolchain is glibc based, since I've excluded glibc/eglibc
toolchains from the static build thing.

I guess I need to improve the autobuild script to only use
BR2_PREFER_STATIC_LIB when !glibc && !eglibc (i.e at the moment, when
uClibc is used).

Thanks for spotting this!

Best regards,

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

[Buildroot] Autobuilder site back online

[Arnout Vandecappelle]

On 17/11/13 18:50, Yann E. MORIN wrote:
> Thomas, All,
> 
> On 2013-11-17 18:29 +0100, Yann E. MORIN spake thusly:
>> >Here are a few analysis of some of the build failures:
> Here are a few more (note: I'm only doing analysis for now):

 One more:

http://autobuild.humanoidz.org/results/05f/05ff0f5ea99df806589c9d2c0aa5d991ae299f5c/
 -> passing 'const QAtomicPointer<pcre16_extra>' as 'this' argument of 'T* QBasicAtomicPointer<T>::loadAcquire() [with T = pcre16_extra]' discards qualifiers
 -> upstream bug, fixed in Fedora https://bugzilla.redhat.com/attachment.cgi?id=812643

 Unfortunately, seems Fedora didn't upstream it.

 I don't have time to test the patch, so perhaps someone else can take
it over?

 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:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Dear Arnout Vandecappelle,

On Tue, 19 Nov 2013 00:15:40 +0100, Arnout Vandecappelle wrote:

>  One more:
> 
> http://autobuild.humanoidz.org/results/05f/05ff0f5ea99df806589c9d2c0aa5d991ae299f5c/
>  -> passing 'const QAtomicPointer<pcre16_extra>' as 'this' argument of 'T* QBasicAtomicPointer<T>::loadAcquire() [with T = pcre16_extra]' discards qualifiers
>  -> upstream bug, fixed in Fedora https://bugzilla.redhat.com/attachment.cgi?id=812643
> 
>  Unfortunately, seems Fedora didn't upstream it.
> 
>  I don't have time to test the patch, so perhaps someone else can take
> it over?

Since this is a Qt5 issue, let's Cc Fatih. He has been very efficient
on fixing Qt5 problems, and on upstreaming the relevant patches.

Fatih, do you think you could have a look at the above problem?

Thanks a lot!

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

[Buildroot] Autobuilder site back online

[Fatih Aşıcı]

On Tuesday 19 November 2013 01:30:49 Thomas Petazzoni wrote:
> Dear Arnout Vandecappelle,
> 
> On Tue, 19 Nov 2013 00:15:40 +0100, Arnout Vandecappelle wrote:
> >  One more:
> > http://autobuild.humanoidz.org/results/05f/05ff0f5ea99df806589c9d2c0aa5d9
> > 91ae299f5c/
> > 
> >  -> passing 'const QAtomicPointer<pcre16_extra>' as 'this' argument of
> >  'T* QBasicAtomicPointer<T>::loadAcquire() [with T = pcre16_extra]'
> >  discards qualifiers -> upstream bug, fixed in Fedora
> >  https://bugzilla.redhat.com/attachment.cgi?id=812643
> >  
> >  Unfortunately, seems Fedora didn't upstream it.
> >  
> >  I don't have time to test the patch, so perhaps someone else can take
> > 
> > it over?
> 
> Since this is a Qt5 issue, let's Cc Fatih. He has been very efficient
> on fixing Qt5 problems, and on upstreaming the relevant patches.
> 
> Fatih, do you think you could have a look at the above problem?

Sent a patch to the list.

Also reported to upstream: https://codereview.qt-project.org/71763

[Buildroot] Autobuilder site back online

[Thomas Petazzoni]

Dear Fatih A??c?,

On Wed, 20 Nov 2013 13:45:51 +0200, Fatih A??c? wrote:

> > Since this is a Qt5 issue, let's Cc Fatih. He has been very
> > efficient on fixing Qt5 problems, and on upstreaming the relevant
> > patches.
> > 
> > Fatih, do you think you could have a look at the above problem?
> 
> Sent a patch to the list.
> 
> Also reported to upstream: https://codereview.qt-project.org/71763

Awesome, thanks a lot! Really great to have you looking after the Qt5
packaging.

Best regards,

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