[U-Boot] Please pull u-boot-fdt, take 2

[Simon Glass]

Hi Tom,

On 26 May 2017 at 06:49, Tom Rini <trini at konsulko.com> wrote:
>
> On Thu, May 25, 2017 at 09:15:33PM -0600, Simon Glass wrote:
> > Hi Tom,
> >
> > On 25 May 2017 at 11:42, Tom Rini <trini at konsulko.com> wrote:
> > > On Thu, May 25, 2017 at 11:27:10AM -0600, Simon Glass wrote:
> > >> Hi Tom,
> > >>
> > >> On 25 May 2017 at 05:19, Tom Rini <trini at konsulko.com> wrote:
> > >> >
> > >> > On Wed, May 24, 2017 at 06:15:25PM -0600, Simon Glass wrote:
> > >> >
> > >> > > Hi Tom,
> > >> > >
> > >> > > This incorporates the v2 patch for 'fdt: Build the new python libfdt
> > >> > > module' which should fix the problem with the original pull request.
> > >> > >
> > >> > >
> > >> > > The following changes since commit be62fbf376261ab3a4ed5db3bf54d5df9e216d9f:
> > >> > >
> > >> > >   Merge branch 'rmobile' of git://git.denx.de/u-boot-sh (2017-05-23
> > >> > > 16:22:03 -0400)
> > >> > >
> > >> > > are available in the git repository at:
> > >> > >
> > >> > >   git://git.denx.de/u-boot-fdt.git
> > >> > >
> > >> > > for you to fetch changes up to da9c601049eb7c993c7f6e33ae10af7a847a483d:
> > >> > >
> > >> > >   fdt: Drop fdt_select.py (2017-05-24 18:12:31 -0600)
> > >> >
> > >> > NAK.  travis-ci blows up quite badly:
> > >> > https://travis-ci.org/trini/u-boot/builds/235861889
> > >>
> > >> I'm not sure how to repeat this problem. When I try this:
> > >
> > > Your best bet is likely:
> > > https://docs.travis-ci.com/user/common-build-problems/#Troubleshooting-Locally-in-a-Docker-Image
> >
> > Sadly still no luck. I installed this one:
> >
> > travisci/ci-garnet:packer-1478744932
> >
> > It includes python-dev but not swig, so should not be able to build
> > the module. As expected I get this error:
> >
> > NO_SDL=1 ./tools/buildman/buildman -P sandbox_spl
> > boards.cfg is up to date. Nothing to do.
> > Building current source for 1 boards (1 thread, 8 jobs per thread)
> >    sandbox:  +   sandbox_spl
> > +Traceback (most recent call last):
> > +  File "<stdin>", line 1, in <module>
> > +ImportError: No module named libfdt
> > +make[2]: *** [checkdtoc] Error 1
> > +make[1]: *** [spl/u-boot-spl] Error 2
> > +make: *** [sub-make] Error 2
> >
> >
> > If I install swig then all is well. I don't see the same error.
> >
> > I'm also unsure how you get it to pass with some boards but not others...
> >
> > There is obviously something odd going on. I also cannot understand
> > why in this one:
> >
> > https://travis-ci.org/trini/u-boot/jobs/235861899#L769
> >
> > I see it trying to compile libfdt_wrap.c. That should be handled by
> > setup.py - I just cannot figure out why it would try to compile it
> > itself:
> >
> >       arm:  +  mx6sabresd_spl
> > +mv: cannot stat lib/libfdt/pylibfdt/libfdt.py: No such file or directory
> > +make[2]: *** [tools/_libfdt.so] Error 1
> > +make[1]: *** [tools] Error 2
> > +make: *** [sub-make] Error 2
> >       arm:  +  ls1021aqds_nor_lpuart
> > +x86_64-linux-gnu-gcc: error: lib/libfdt/pylibfdt/libfdt_wrap.c: No
> > such file or directory
> > +x86_64-linux-gnu-gcc: fatal error: no input files
> > +compilation terminated.
> > +error: command 'x86_64-linux-gnu-gcc' failed with exit status 4
> > +make[2]: *** [tools/_libfdt.so] Error 1
> > +make[1]: *** [tools] Error 2
> > +make: *** [sub-make] Error 2
> >       arm:  +  mx6slevk
> > +x86_64-linux-gnu-gcc: error: lib/libfdt/pylibfdt/libfdt_wrap.c: No
> > such file or directory
> > +x86_64-linux-gnu-gcc: fatal error: no input files
> > +compilation terminated.
> > +error: command 'x86_64-linux-gnu-gcc' failed with exit status 4
> > +make[2]: *** [tools/_libfdt.so] Error 1
> > +make[1]: *** [tools] Error 2
> > +make: *** [sub-make] Error 2
> >   55    0    3 /58     mx6qsabreauto
> > boards.cfg is up to date. Nothing to do.
> > Summary of current source for 58 boards (2 threads, 1 job per thread)
> >       arm:  +  mx6sabresd_spl ls1021aqds_nor_lpuart mx6slevk
> > +mv: cannot stat lib/libfdt/pylibfdt/libfdt.py: No such file or directory
> >
> >
> > Are you able to run with V=1 to get the full make output? That might
> > help me narrow it down. Or, if you have a particular docket image,
> > please point me to it. There seems to be some init going on though
> > (e.g. /tmp/dtc).
>
> Anyone can use travis-ci :)  In the past I've had some luck debugging
> these issues (since I didn't want to mess with docker) by hacking up the
> .travis.yml file to just have a single matrix entry for whatever I was
> debugging and hack things as needed, including 'cat'ing files.

OK I found that with the prerequisites installed I can repeat this
with something like:

buildman -P sandbox

(which only builds 3 boards)

The problem is that setup.py is writing to the source directory. It
seems to be hard to stop it doing that. I was aware of it but did not
imagine the race condition it would create. I've updated the patches
and sent v3. I have a passing run on travis-ci [1]. If all looks good
I'll pull it in and try a take three early next week.

Most of my testing is with buildman running on a branch. In that case
each thread has its own source tree, so the race condition does not
occur.

Once again I'm really impressed with your test setup!

Regards,
Simon

[1] https://travis-ci.org/sglass68/u-boot/builds/236586312