Closed claretway closed 7 years ago
is this master or develop branch?
I was on master
Output says -j48 right before it fails not sure what behavior to expect but I'm doubting you have 48 cores (although it's not impossible) also -l48 seems suspect perhaps some typo in build script?
On Jul 21, 2017 6:34 PM, "claretway" notifications@github.com wrote:
I was on master
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/CPFL/Autoware/issues/737#issuecomment-317128067, or mute the thread https://github.com/notifications/unsubscribe-auth/AGXckF3IiaUzIFeN4XLKvEZ1h-iabXjFks5sQSeOgaJpZM4Of7tn .
I have built successfully on 16.04 and kinetic with the same command
./catkin_make_release --make-args -j16
@amc-nu who is most familiar with build issues in Autoware? Do you guys have any nightly or PR builds for different:
I think there are 24 cores and 24 virtual cores
@claretway We don't find any problems while building with the script maybe it's something related to your environment. Please try compiling manually While in the ros directory:
@dejanpan afaik, it supports ARM architecture, but that's already included in the scripts.
It also supports ROS hydro or Ubuntu 13.04, 13.10, but that's a very old version of Autoware.
thx. What I meant is do you build on clean versions of all this variations before new pull request gets merged into master?
@dejanpan We check it builds and works on 14.04 and 16.04. ARM is also verified.
Do you have any documentation on how building and working is tested? Is it manual?
I am asking because my plan was to release Autoware through ROS build farm: http://wiki.ros.org/bloom/Tutorials/FirstTimeRelease which would check these things where at least building would happen automatically.
@dejanpan We do it manually on each of our different machines. We add the new dependencies, clean the environment using the catkin_make_release script and test on each of our different environments before merging to master (since all of us work on the develop branch)
Would definitely try -j4 or -j8. On my 32gb ram workstation I have ran into memory issues when building with a lot of cores, but sounds like there might be something else going on as well.
On Jul 21, 2017 8:48 PM, "Abraham Monrroy" notifications@github.com wrote:
@dejanpan https://github.com/dejanpan We do it manually on each of our different machines. We add the new dependencies, clean the environment using the catkin_make_release script and test on each of our different environments before merging to master (since all of us work on the develop branch)
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/CPFL/Autoware/issues/737#issuecomment-317142148, or mute the thread https://github.com/notifications/unsubscribe-auth/AGXckMOb9VkwaJlhKlgE1avl5peQfRZAks5sQUbvgaJpZM4Of7tn .
Okay, now I'm thinking it's an issue with opencv, of which I have version 3.3.0 installed. I got this error, I guess after I ran catkin_make clean
and built with reduced cores:
make[2]: *** No rule to make target `/usr/lib/x86_64-linux-gnu/libopencv_videostab.so.2.4.8', needed by `/home/driver/ccui/planner/Autoware/ros/devel/lib/vectacam/vectacam_node'. Stop.
Is Autoware not working with OpenCV3 still @syohex? I guess I missed this. I was under the impression that "OpenCV 2.4.10 or higher" under the README requirements meant that OpenCV3 was supported as well.
Still running into issues, but earlier issue was due to wrong OpenCV version.
I know that this was covered here but I ran into this same issue and the suggestions for debugging did not work for me. I'm using Ubuntu 14.04 with ROS Indigo. I was never able to selectively build the orb_localizer package with
catkin_make --pkg orb_localizer
even after adding the lines to the CMakeLists.txt suggested in the link above.This is the output I get after trying to build with
./catkin_make_release --make-args -j16
This is the output I get without any flags, with just
./catkin_make_release
. It appears that I can actually get further without specifying the number of cores also.