Hi, the 11.08 Android toolchain RC is ready:
https://android-build.linaro.org/builds/~linaro-android/toolchain-4.6-2011.0... https://android-build.linaro.org/jenkins/job/linaro-android_toolchain-4.6-20...
The good news is that this build has been tested and seems to work fine for everything so far. The bad news is that we did that by reverting the switch from the arm-eabi to the arm-linux-androideabi target, losing support for OpenMP and friends. My theory is that the combination of an arm-linux-androideabi toolchain and the mess in the Android build system that works around the toolchain not knowing anything about the OS simply don't go well together. We should investigate this further for the next release.
Other than that, this toolchain build seems to be ready for release. Let me know if you run into any problems.
ttyl bero
Hello Bernhard,
On Fri, 19 Aug 2011 23:47:26 -0700 Bernhard Rosenkranzer bernhard.rosenkranzer@linaro.org wrote:
Hi, the 11.08 Android toolchain RC is ready:
https://android-build.linaro.org/builds/~linaro-android/toolchain-4.6-2011.0... https://android-build.linaro.org/jenkins/job/linaro-android_toolchain-4.6-20...
The good news is that this build has been tested and seems to work fine for everything so far.
Have you guys worked all night - when I went to bed, android-build was busy with the builds, and I woke up and it's busy either ;-). Great work!
Did you have a chance to look at 4.5 build: https://android-build.linaro.org/jenkins/job/pfalcon_toolchain-4.5-2011.08-l... ? Or would you have one? Just in case, I'd like to discuss contingency plan for 4.5 release either. 4.5 builds against last-month's toolchain bundling: https://android-build.linaro.org/jenkins/job/pfalcon_toolchain-4.5-2011.08-l... Releasing 4.6 & 4.5 against different tags/branches of toolchain/* is not ideal, but we could do that if needed (assuming it would be fixed next milestone).
Thanks, Paul
The bad news is that we did that by reverting the switch from the arm-eabi to the arm-linux-androideabi target, losing support for OpenMP and friends. My theory is that the combination of an arm-linux-androideabi toolchain and the mess in the Android build system that works around the toolchain not knowing anything about the OS simply don't go well together. We should investigate this further for the next release.
Other than that, this toolchain build seems to be ready for release. Let me know if you run into any problems.
ttyl bero
On Sat, Aug 20, 2011 at 8:47 AM, Bernhard Rosenkranzer < bernhard.rosenkranzer@linaro.org> wrote:
Hi, the 11.08 Android toolchain RC is ready:
https://android-build.linaro.org/builds/~linaro-android/toolchain-4.6-2011.0...
https://android-build.linaro.org/jenkins/job/linaro-android_toolchain-4.6-20...
The good news is that this build has been tested and seems to work fine for everything so far.
This is remarkable Bernhard. Thanks for that!
The bad news is that we did that by reverting the switch from the arm-eabi to the arm-linux-androideabi target, losing support for OpenMP and friends. My theory is that the combination of an
That's OK! We have latest Linaro toolchain and afaik you also updated binutils and friends which might give us more goodies/optimizations on top of what we got from TCWG gcc update.
Remember to talk to Andy about benchmarking these builds and maybe give some hints what he should pay extra attention to.
arm-linux-androideabi toolchain and the mess in the Android build system that works around the toolchain not knowing anything about the OS simply don't go well together. We should investigate this further for the next release.
++ I agree that the magic they do for using bare-metal for whole platform might contribute to that.
Another idea I had is that there seem to be two ways to build androideabi: one is the way we use; the other is to use the NDK method of building TC. Maybe we tried to use the wrong one?
Other than that, this toolchain build seems to be ready for release. Let me know if you run into any problems.
I kicked off a build for leb-panda and panda tip with the new toolchain to see if that succeeds in the validation lab. If that works I will move the release builds for panda and leb-panda as well.
Unfortunately, beagle isn't really set up properly in the validation lab it seems; even deploy fails, look here: http://validation.linaro.org/lava-server/dashboard/streams/anonymous/android...
And all other tip builds we run seem to be not hooked up with the lab (or dont have support in the lab yet). I will move those to new toolchain once I see green light on panda/leb-panda hoping that that means they won't regress.
so following up:
panda tip build succeeded and the android test results showed up as expected and all looks fine (manually eyeballed the serial.log and 0xbench results seem to show up nicely).
here where the results show up: - android results: http://validation.linaro.org/lava-server/dashboard/reports/android-runs/ - panda result: http://validation.linaro.org/lava-server/dashboard/permalink/test-run/e843ef... build https://android-build.linaro.org/jenkins/job/linaro-android_panda/237/)
leb-panda succeeded build as well, I can also see that LAVA was run with it when going to the job ( http://validation.linaro.org/lava-server/scheduler/job/599) but I cannot figure what happened after that. Is there any way to find out what happened to that job (maybe serial.log for a scheduler job?)?
-
On Sat, Aug 20, 2011 at 8:47 AM, Bernhard Rosenkranzer < bernhard.rosenkranzer@linaro.org> wrote:
Hi, the 11.08 Android toolchain RC is ready:
https://android-build.linaro.org/builds/~linaro-android/toolchain-4.6-2011.0...
https://android-build.linaro.org/jenkins/job/linaro-android_toolchain-4.6-20...
The good news is that this build has been tested and seems to work fine for everything so far. The bad news is that we did that by reverting the switch from the arm-eabi to the arm-linux-androideabi target, losing support for OpenMP and friends. My theory is that the combination of an arm-linux-androideabi toolchain and the mess in the Android build system that works around the toolchain not knowing anything about the OS simply don't go well together. We should investigate this further for the next release.
Other than that, this toolchain build seems to be ready for release. Let me know if you run into any problems.
ttyl bero
On Sat, Aug 20, 2011 at 7:32 PM, Alexander Sack asac@linaro.org wrote:
so following up:
panda tip build succeeded and the android test results showed up as expected and all looks fine (manually eyeballed the serial.log and 0xbench results seem to show up nicely).
here where the results show up:
- android results:
http://validation.linaro.org/lava-server/dashboard/reports/android-runs/
- panda result:
http://validation.linaro.org/lava-server/dashboard/permalink/test-run/e843ef... build https://android-build.linaro.org/jenkins/job/linaro-android_panda/237/)
leb-panda succeeded build as well, I can also see that LAVA was run with it when going to the job ( http://validation.linaro.org/lava-server/scheduler/job/599) but I cannot figure what happened after that. Is there any way to find out what happened to that job (maybe serial.log for a scheduler job?)?
the fact that panda tip showed up as working in LAVA gave me enough comfort to move the panda 11.08-release builds to new toolchain as well. Builds will show up as:
- https://android-build.linaro.org/builds/~linaro-android/panda-11.08-release/... - https://android-build.linaro.org/builds/~linaro-android/leb-panda-11.08-rele...
Let's see if they get properly tested in lava etc.
-
Alexander Sack -
Bernhard Rosenkranzer -
Paul Sokolovsky