[Buildroot] Some build failures to solve for volunteers

[Buildroot] Some build failures to solve for volunteers

[Thomas Petazzoni]

Hello,

I have done a set of test builds, with a random choice of toolchain
amongst a set of internal and external toolchains I have and with a
random choice of packages. I have fixed some of the issues that occurred
multiple times, but there are many remaining issues and I won't have
the time to go through all of them.

Therefore, I've put online the results of the failed tests, with
the .config and the complete build log :

 http://free-electrons.com/~thomas/buildroot/test-output-2011-05-22/

Unfortunately, some of the builds have been done with external
toolchains, and the page above doesn't give details about the custom
external toolchains I'm using. But anyway, if people have time to go
through some of these failures and contribute fixes, that'd be really
great.

Regards,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Some build failures to solve for volunteers

[Daniel Nyström]

2011/5/22 Thomas Petazzoni <thomas.petazzoni@free-electrons.com>:
> I have done a set of test builds, with a random choice of toolchain
> amongst a set of internal and external toolchains I have and with a
> random choice of packages. I have fixed some of the issues that occurred
> multiple times, but there are many remaining issues and I won't have
> the time to go through all of them.

Great initiative!

> Unfortunately, some of the builds have been done with external
> toolchains, and the page above doesn't give details about the custom
> external toolchains I'm using. But anyway, if people have time to go
> through some of these failures and contribute fixes, that'd be really
> great.

How about setting up a couple of buildbots with different toolchains
and all options enabled? We can all help running build slaves?

[Buildroot] Some build failures to solve for volunteers

[Thomas Petazzoni]

Hello,

On Mon, 23 May 2011 14:22:19 +0200
Daniel Nystr?m <daniel.nystrom@timeterminal.se> wrote:

> 2011/5/22 Thomas Petazzoni <thomas.petazzoni@free-electrons.com>:
> > I have done a set of test builds, with a random choice of toolchain
> > amongst a set of internal and external toolchains I have and with a
> > random choice of packages. I have fixed some of the issues that
> > occurred multiple times, but there are many remaining issues and I
> > won't have the time to go through all of them.
> 
> Great initiative!

Thanks :-)

> > Unfortunately, some of the builds have been done with external
> > toolchains, and the page above doesn't give details about the custom
> > external toolchains I'm using. But anyway, if people have time to go
> > through some of these failures and contribute fixes, that'd be
> > really great.
> 
> How about setting up a couple of buildbots with different toolchains
> and all options enabled? We can all help running build slaves?

I have no experience with BuildBot, but as far I as could see, BuildBot
is oriented towards building the same software in the same
configuration over and over again, for continuous integration. From
what I could see, it assumes that it has a finite set of build
configurations, and run them all.

Here, what I do is random configuration testing, so there is
potentially an unlimited set of build configurations, and I almost
never test the same configuration again. So it's different from
continuous integration. I don't know if/how BuildBot can handle that
kind of thing.

And finally, my build set up uses some Buildroot internal toolchain or
the CT-NG backend, but also uses a lot of preinstalled external
toolchains. So each BuildBot slave would have to have all those
external toolchains installed for the thing to work.

So, yes, having a BuildBot thing running would be nice, but I don't
know how easy it will be to setup. I know Peter wanted to set up
something like that, but I think he is also doing his randpackageconfig
tests with a custom script.

Regards,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Some build failures to solve for volunteers

[Daniel Nyström]

>> > Unfortunately, some of the builds have been done with external
>> > toolchains, and the page above doesn't give details about the custom
>> > external toolchains I'm using. But anyway, if people have time to go
>> > through some of these failures and contribute fixes, that'd be
>> > really great.
>>
>> How about setting up a couple of buildbots with different toolchains
>> and all options enabled? We can all help running build slaves?
>
> I have no experience with BuildBot, but as far I as could see, BuildBot
> is oriented towards building the same software in the same
> configuration over and over again, for continuous integration. From
> what I could see, it assumes that it has a finite set of build
> configurations, and run them all.

There's no problem running Buildroot builds with Buildbot. I do
actually have one running here locally for our local needs. :)

One could set "make randpackageconfig" as one step of the build
process in the Factory, but that makes it hard to find patterns in the
failing builds, hence my suggestion of one toolchain per slave and all
options enabled. Maybe there's other advantages of random configs?

[Buildroot] Some build failures to solve for volunteers

[Thomas Petazzoni]

On Mon, 23 May 2011 14:42:12 +0200
Daniel Nystr?m <daniel.nystrom@timeterminal.se> wrote:

> > I have no experience with BuildBot, but as far I as could see,
> > BuildBot is oriented towards building the same software in the same
> > configuration over and over again, for continuous integration. From
> > what I could see, it assumes that it has a finite set of build
> > configurations, and run them all.
> 
> There's no problem running Buildroot builds with Buildbot. I do
> actually have one running here locally for our local needs. :)

Sure running Buildroot builds with BuildBot is possible. What I'm
questionning is if running *random* builds with BuildBot does make
sense. I think BuildBot tracks whether a particular build configuration
was building at a time, then no longer building. This only makes sense
if the tested build configurations are identical over time.

> One could set "make randpackageconfig" as one step of the build
> process in the Factory, but that makes it hard to find patterns in the
> failing builds, hence my suggestion of one toolchain per slave and all
> options enabled. Maybe there's other advantages of random configs?

Well "make randpackageconfig" creates a completely random set of
package set and options. This allows to test an enormous number of
possible configurations.

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] Some build failures to solve for volunteers

[Peter Korsgaard]

>>>>> "Daniel" == Daniel Nystr?m <daniel.nystrom@timeterminal.se> writes:

Hi,

 Daniel> There's no problem running Buildroot builds with Buildbot. I do
 Daniel> actually have one running here locally for our local needs. :)

Here as well.

 Daniel> One could set "make randpackageconfig" as one step of the build
 Daniel> process in the Factory, but that makes it hard to find patterns in the
 Daniel> failing builds, hence my suggestion of one toolchain per slave and all
 Daniel> options enabled. Maybe there's other advantages of random configs?

Well, it's a trade off between us wanting reproducable builds and
wanting to test as many configurations as possible.

My current (private) setup involves randpackageconfig and then archiving
the .config and the entire build log, but that easily grows to something
quite unhandy for querying (E.G. in what other builds with a specific
toolchain config and package A+B but not C did it succeed).

-- 
Bye, Peter Korsgaard