Up till now we have been installing arm-gcc using apt from the team-gcc-arm-embedded ppa.
As discussed in #1453 and implemented in #1504, we have decided to lock travis to use the specific version of the toolchain that is documented in the READMEs of relevant platforms. However, the PPA no longer hosts this version, apt fails to install arm-gcc and as a result the job fails too.
This commit changes the travis job to wget an installation tarball for the desired version, instead of trying to install via apt.
Make Travis CI use the GNU ARM Embedded toolchain version specified by
the READMEs, instead of automatically using the latest version provided
by the PPA. In this way, the READMEs will still be correct after a PPA
upgrade, and the version used by Contiki is under control.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau.dev@gmail.com>
Up to now we were using the LibreOffice 4.3 ppa (ppa:libreoffice/libreoffice-4-4) to install doxygen. The LibreOffice Packaging team appear to have removed this ppa, resulting in our doxygen build failing.
This changes the ppa we use to LibreOffice 4.4.x.
It makes sense to test cc2538dk with the toolchain recommended in the README (GNU Tools for ARM Embedded Processors) rather than the one which was being used two years ago (Sourcery G++ Lite)
This helps reduce the chance of a build failure due to transient
download error. This also switches to consistent use of wget
throughout, which reports download errors more cleanly in the travis
build logs.
Besides general whitespace and quoting cleanup, commands that test
variables were changed from:
[ $B = a ] && command || true
to
if [ $B = a ] ; then command ; fi
This form correctly returns the exit code of "command".
For travis tests, we host pre-built bundles of the arm-gcc and sdcc
toolchains on github, and we download them with curl
Apparently, the URL for those bundles has been changed and the
HTTP response is a 301 redirect. curl doesn't follow it
We change .travis.yml to pull the bundles form the new location
- build cc65 libraries only for target supported by Contiki
- instead of building all examples for all 6502 targets build
- all examples for one target (c64)
- the most demanding client example (webbrowser) for all 6502 targets
- the most demanding server examples (webserver) for all 6502 targets
* The previous 01-compile test has been renamed to compile-base. It
now only tests {msp430, avr, native}-based platforms
* New test builds for arm-based and 8051-based ports (one build per arch)
* Only install arm-none-eabi, sdcc and srecord for the builds which need
them. This will somewhat speed up all other tests
* Test er-rest-example and webserver-ipv6 for the cc2538
* An erroneous comment was fixed and another comment was somewhat
clarified
1. there was an odd error when this environment variable wasn't set
(`[: =: unary operator expected`)
2. this environment variable is not used anywhere else
Removing it makes the logic look much simpler - do make `cooja.jar`
if build type is any other then `compile`.
Antonio Lignan
George Oikonomou
George Oikonomou
Robert Olsson
Wojciech Bober
Benoît Thébaudeau
Simon Duquennoy
Lars Schmertmann
Adam Dunkels
George Oikonomou
Moritz 'Morty' Strübe
Nicolas Tsiftes
Jonas Olsson
Jim Paris
Oliver Schmidt
Ian Martin
Sébastien Dawans
Adam Dunkels
George Oikonomou
Ilya Dmitrichenko
Mariano Alvira