Bug#779183: pcl: sometimes FTBFS - fatal error: pcl/visualization/pcl_visualizer.h: No such file or directory
James Cowgill
james410 at cowgill.org.uk
Wed Feb 25 09:22:52 UTC 2015
Source: pcl
Version: 1.7.2-6
Severity: serious
Tags: patch
Hi,
pcl seems to sometimes FTBFS with this error:
> [ 52%] Building CXX object tools/CMakeFiles/pcl_mesh2pcd.dir/mesh2pcd.cpp.o
> cd /tmp/buildd/pcl-1.7.2/build/tools && /usr/bin/c++ -DEIGEN_USE_NEW_STDVECTOR -DEIGEN_YES_I_KNOW_SPARSE_MODULE_IS_NOT_STABLE_YET -DQT_CORE_LIB -DQT_GUI_LIB -DQT_NO_DEBUG -Dqh_QHpointer -g -O2 -fstack-protector-strong -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -D_FORTIFY_SOURCE=2 -pthread -fopenmp -Wno-deprecated -isystem /usr/include/eigen3 -isystem /usr/include/ni -I/tmp/buildd/pcl-1.7.2/recognition/include/pcl/recognition/3rdparty -isystem /usr/include/qt4 -isystem /usr/include/qt4/QtGui -isystem /usr/include/qt4/QtCore -I/tmp/buildd/pcl-1.7.2/build/include -I/tmp/buildd/pcl-1.7.2/common/include -I/tmp/buildd/pcl-1.7.2/io/include -I/tmp/buildd/pcl-1.7.2/filters/include -I/tmp/buildd/pcl-1.7.2/sample_consensus/include -I/tmp/buildd/pcl-1.7.2/segmentation/include -I/tmp/buildd/pcl-1.7.2/search/include -I/tmp/buildd/pcl-1.7.2/kdtree/include -I/tmp/buildd/pcl-1.7.2/features/include -I/tmp/buildd/pcl-1.7.2/surface/include -I/tmp/buildd/pcl-1.7.2/octree/include -I/tmp/buildd/pcl-1.7.2/registration/include -I/tmp/buildd/pcl-1.7.2/recognition/include -I/tmp/buildd/pcl-1.7.2/geometry/include -I/tmp/buildd/pcl-1.7.2/keypoints/include -I/usr/include/vtk-5.8 -I/tmp/buildd/pcl-1.7.2/tools/include -o CMakeFiles/pcl_mesh2pcd.dir/mesh2pcd.cpp.o -c /tmp/buildd/pcl-1.7.2/tools/mesh2pcd.cpp
> /tmp/buildd/pcl-1.7.2/tools/mesh2pcd.cpp:38:46: fatal error: pcl/visualization/pcl_visualizer.h: No such file or directory
> #include <pcl/visualization/pcl_visualizer.h>
> ^
> compilation terminated.
Note that -I/.../visualization/include does not appear in the compiler
command like it does on the successful builds.
The top of the log contains this which seems a bit strange:
> -- The following subsystems will be built:
> -- common
> -- kdtree
> -- octree
> -- search
> -- sample_consensus
> -- filters
> -- features
> -- geometry
> -- registration
> -- io
> -- segmentation
> -- surface
> -- recognition
> -- keypoints
> -- visualization
> -- tracking
> -- apps
> building:
> |_ point_cloud_editor
> |_ in_hand_scanner
> |_ modeler
> not building:
> |_ cloud_composer: No reason
> |_ optronic_viewer: FZAPI was not found.
> -- people
> -- outofcore
> -- examples
> -- The following subsystems will not be built:
> -- tools: Requires visualization.
> -- global_tests: No reason
I had a look at the build system and it seems that "tools" does not
declare a dependency on "visualization" so the order they're built is
random. I think the order depends on the order of directory entries as
chosen by the filesystem driver. This could explain why the build worked
on some buildds.
I haven't managed to build the current version on my amd64 machine at
all (in various different chroots).
Some logs of failed builds (the arm64 one later built successfully):
https://buildd.debian.org/status/fetch.php?pkg=pcl&arch=sparc&ver=1.7.2-6&stamp=1417974163
https://launchpadlibrarian.net/193721902/buildlog_ubuntu-vivid-ppc64el.pcl_1.7.2-6_FAILEDTOBUILD.txt.gz
https://buildd.debian.org/status/fetch.php?pkg=pcl&arch=arm64&ver=1.7.2-3&stamp=1416536853
I've attached a patch which adds the dependency which fixes this issue
for me.
Thanks,
James
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0005-tools-depends-on-visualization.patch
Type: text/x-patch
Size: 501 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debian-science-maintainers/attachments/20150225/de8af322/attachment.bin>
More information about the debian-science-maintainers
mailing list