Re: [Paparazzi-devel] Continous Integration Server

[Felix Ruess]

Hi Gareth,

> Interesting, some of the makefiles don't trigger an error.  Halfway through
> a testbuild, I realised gcc-arm from debian was shadowing -arm-multilib and
> removed it.  Despite missing GCC and throwing make errors all over the
> place, the build 'succeeded'!

Sorry, I'm not quite following.... which gcc-arm package do you mean,
the one from the paparazzi repo or some other debian package?
The toolchain from paparazzi-arm-multilib install to
/opt/paparazzi/arm-multilib, so I don't think that this is shadowed by
some other package.

So in the makefile we look for the toolchain in
/opt/paparazzi/arm-multilib, then in ~/sat, then try finding it in
PATH (and in the case of lpc21 we then try to find the old
arm-elf-gcc).

The old gcc-arm was installed to /usr , but in making the
paparazzi-arm-multilib package I followed the paparazzi-stm32 package
and put the toolchain in /opt/paparazzi to avoid clashes with other
arm-none-eabi compilers possibly installed under /usr ..... We could
of course change the package to install to /usr again as well and just
pick it up from the path as the old arm-elf-gcc.
It would be a bit nicer at the risk of possible problems... not sure
if it is worth it..

> Not sure what's going on, you can see the log @
> http://ec2-79-125-28-43.eu-west-1.compute.amazonaws.com:8080/job/Paparazzi/14/console
>
> Additionally, after I removed gcc-arm there is no gcc found any more, as per
> http://ec2-79-125-28-43.eu-west-1.compute.amazonaws.com:8080/job/Paparazzi/15/console

For older branches the makefiles have not been updated to search for
arm-multilib, so these will of course fail if only
paparazzi-arm-multilib is installed.
Also when a toolchain was not found (either the old stm32 or the new
arm-multlib) in some older makefiles no arguments were passed to
dirname and that of course throws errors.

So I'm not really sure if we really want to build all branches all the
time or instead specify a few (master, dev, ...)?

E.g. for the HB branch (really old, but containing some stuff from the
guys in Bremen for their HB autopilot which was never finished and
merged is currently still kept for reference... also the campaign2010
branch is one that Martin maintains for some meteorologists and
doesn't contain the updated makefiles for the new toolchain.

Cheers, Felix