Closed rajat2004 closed 4 years ago
Interesting
And on the contrary, I would say exporting the variable into the current environment is more elegant than reloading the environment as a whole.
Will try to reproduce in a docker container and check if this works
And on the contrary, I would say exporting the variable into the current environment is more elegant than reloading the environment as a whole.
On second thought, yeah, that's true. I'll mostly soon open a PR for this
Will try to reproduce in a docker container and check if this works
Great. thanks! I'll also try, though will take some time.
Fixed via #53
I'm reopening the issue since the problem is still occurring with the latest code changes (Another user, not mine)
Some more problems also occurred, with libegl.so
(though this possibly was a manual installation issue with Nvidia drivers),libz
(caused by Anaconda)
Update: I've tried to recreate the problem in Travis with some success, branch - https://github.com/rajat2004/Ubuntu-Setup-Scripts/tree/test-opencv
Travis build - https://travis-ci.com/rajat2004/Ubuntu-Setup-Scripts/jobs/271557371
It builds and installs, but the warning is present
and a checkinstall
packaging problem
Googling this shows it happens sometimes with Java due to some reason, OpenCV builds Java wrappers by default
Finally did a make install
so as to not go through the whole thing again
Since OpenCV throws errors if it's not able to find the FFmpeg libs, isn't it already using the compiled libraries rather than the system ones (First thing mentioned in https://github.com/rsnk96/Ubuntu-Setup-Scripts/issues/20)
The example of zlib
given there isn't actually being compiled from source, no such flags are being passed to it during the configuration
An error which I encountered while running
opencvDirectInstall.sh
. This was done on a fresh installation of Ubuntu 18.04 with both the 1st and 2nd scripts used before this.In the
1-BasicSetup.sh
, Anaconda installation was commented out, nvidia-drivers were enabled. No errors in running it and ZSH was working as expected. No changes in 2, everything ran smoothly. Cuda 10.0 was installed before the OpenCV installation. Reboots between running each script and Cuda.Error -
FFmpeg building and installation worked, but after going back and looking through the output, saw this -
Ok, so for some reason,
zsh-autosuggestions
was loading, which was pretty strange in itself since it was working correctly till then. Due to this, theLD_LIBRARY_PATH
and the associated ones didn't get loaded. Opening a new terminal, it was loading correctly and the new FFmpeg paths were also set.So started a new terminal, checked the paths, cleared OpenCV build and ran it again, still the same problem. Reboot, same thing. Was getting quite frustrated by now. FFmpeg was working.
Now what I did that fixed it (Some more things that I tried, but these seem to be relevant)-
Will try to reproduce it, and to figure out what to do to prevent it
One idea from the ZSH error - Export the required PATHS and then echo them to the SHELLRC, rather than to source the SHELLRC, it's an inelegant solution, but should work.