Where does anaconda install cudatoolkit and cudnn? [duplicate] - cuda

I've done my homework: searched, tried and read conda documentation. However, I could not find the answer to this seeming common and simple task: List files that belong to an installed package. How do I do that?
My conda version: conda 4.3.30
I've looked at list, info, search, and package subcommands.
My use cases for this:
When a package 'A' installed another package 'B' as its dependency. But B has a bug, or the installation somehow broken, I want to check which files B installed.
An extended use: when commands of A calls some command b of B, but I don't know the exact name of B. A 'reverse' search based on b to find out B and lets me read more about it.

Assuming you have activated the relevant anaconda environment, you can look at the file ${CONDA_PREFIX}/conda-meta/<package-name-and-version>-<hash>.json and look for the files element.
(this works with Miniconda on Linux)

Found out that all packages that conda installed are stored under <root_environment>/pkgs. One can find out about root environment and other information by running conda info.
Then to list files that a package has:
tree <root_environment>/pkgs/<package_name>-<package_version>
or with find, one can also find which downloaded package has the command:
find <root_environment> -type f -iname 'somecommand'
<root_environment> here is a placeholder for something like ~/anaconda if one installed anaconda into ~/anaconda
This solution is rather *nix-specific but it is good enough to me.

Related

FPC IDE: Can't find unit system

I just installed FPC 3.2.0 (on Linux Mint 19.3) and trying to use FPC IDE. I launched the IDE from the bin installation subfolder as ./fp, written a simplest program in IDE
program hello;
begin
writeln('hello');
end.
saved the program as hello.pas and when I compile it I get
(3,1) Fatal: Can't find unit system used by hello
(0) Fatal: Compilation aborted
This is strange because system is the compiler's unit, not a user's unit.
I tried to google the issue, found several posts in different forums, one question was dated back to 2007 year, and none of the questions was answered.
Is where a way to solve the issue or FPC IDE is dead for about 15 years?
Check the unit directories in options->directories -> unit directories.
The package configuration should have put a line like
/usr/lib/fpc/$FPCVERSION/units/$FPCTARGET/*
The dollar values are builtins (respectively 3.2.0 and i386-linux or x86_64-linux). Please verify that the directories with prebuilt .ppu .o are there.
Try to fix your installation of FPC or you can fix it manually by calling fpcmkcfg
fpcmkcfg -d "basepath=path/to/fpc/3.3" -o path/to/fpc.cfg
For instance, if the compiler is in /usr/local/bin, it will look in
/usr/local/etc.
Had the same problem on Ubuntu 20.04. To solve it, I had to create the file fp.cfg in my project's folder with the following content:
-Fu/usr/lib/x86_64-linux-gnu/fpc/$fpcversion/units/$fpctarget
-Fu/usr/lib/x86_64-linux-gnu/fpc/$fpcversion/units/$fpctarget/*
If this will not work, it means unit path on your system is different. You may use something like
dpkg -L fp-units-base-3.0.4 | grep .ppu
to find out where units are actually stored. Replace 3.0.4 with your actual Free Pascal version. If you do not know exact version, use
sudo apt install apt-show-versions
apt-show-versions fpc
It will show something like
fpc:all/focal 3.0.4+dfsg-23 uptodate
3.0.4 is the version number you need.
Another way of getting needed paths is running this command:
cat /etc/fpc.cfg | grep -- -Fu | head -2
So you may just run
cat /etc/fpc.cfg | grep -- -Fu | head -2 > fp.cfg
in your project's folder to have the problem solved.
How to use Free Pascal (Lazarus) on Linux - no install solution:
If not already installed, install "GNU binutils" (sudo apt-get install binutils)
Dowload the latest version of Free Pascal (Lazarus): from the Downloads section of the Lazarus Home page (Lazarus is a professional open-source cross platform IDE powered by Free Pascal):
2.1. In case of a Debian based Linux system (such as Ubuntu): the .deb package could be downloaded ("fpc-laz_<version> ... .deb" (replace <version> with an actual version number))
2.2. In case of a RPM based Linux system (such as Fedora): the .rpm package could be downloaded ("fpc-<version> ... .rpm" (replace <version> with an actual version number))
Extract the downloaded package:
3.1. Place the downloaded .deb / .rpm package in the folder where you are going to extract it
3.2. Extract the downloaded .deb / .rpm package by Right-Click-ing it and then choosing Extract Here
3.3. Rename the extracted folder, as you desire, for example "fpc"; Further we denote this folder as "<BASEFOLDER>" and its path as - "<BASEFOLDER_PATH>"
In the case of a .deb package: Open the extracted <BASEFOLDER> and then, extract "data.tar.xz" - by Right-Click-ing it and then choosing Extract Here
Launch the Free Pascal window by typing the next command in the terminal - this command may vary according to the terminal used - replace <terminal> with the terminal used (by default: for Gnome based Desktop Environments -> gnome-terminal; for KDE based Desktop Environments -> konsole; for Xfce based Desktop Environments -> xfce4-terminal; for the MATE Desktop Environment -> mate-terminal; ...):
<terminal> -e '<BASEFOLDER_PATH>/usr/bin/fp'
OR:
<terminal> -- '<BASEFOLDER_PATH>/usr/bin/fp'
To add the units folders: From the Free Pascal Menu, access: Options -> Directories -> and copy (CTRL + C to Copy from outside the IDE, and (FN + )SHIFT + Insert to Paste inside the IDE) to the "Units" tab, the next three lines, each on a new line (replace <BASEFOLDER_PATH> with its corresponding value, and replace <LIB_FOLDER> with: lib in case of a .deb package or lib64 in case of a .rpm package):
<BASEFOLDER_PATH>/usr/<LIB_FOLDER>/fpc/$fpcversion\units\$fpctarget
<BASEFOLDER_PATH>/usr/<LIB_FOLDER>/fpc/$fpcversion\units\$fpctarget\*
<BASEFOLDER_PATH>/usr/<LIB_FOLDER>/fpc/$fpcversion\units\$fpctarget\rtl
Please note that by moving the files from the <BASEFOLDER> directory to other <BASEFOLDER_PATH>, you'll have to update the above three lines so that they contain the new <BASEFOLDER_PATH>, otherwise you might get errors like:
"Fatal: Can't find unit ..."
For navigation inside the IDE using keyboard - the following shortcuts can be used: Alt + Highlighted-letter, Highlighted-letter(s), Tab / Shift + Tab, Arrows, Enter, Esc
For Official Online Documentation see:
Online Documentation

TCL_LIB_SPEC missing in modules install

I'm trying to install modules from SourceForge and I'm getting an error that a variable named TCL_LIB_SPEC is not set. What is this supposed to be set to?
The answer here is to install the TCL development package
% yum install tcl-devel
This gives you a tclConfig.sh file at /usr/lib64/tcl8.5
So
% cd /usr/lib64/tcl8.5
% . ./tclConfig.sh
Then configure works.
It's supposed to be set to the instructions to use with your compiler for linking against the Tcl C library, and it should (conventionally) be generated by running the relevant configure script inside a Tcl source distribution. Or a distribution of Tcl (e.g., on Linux perhaps called tcl-dev) may have alternate correct values already set up.

"Fatal error: 'EXTERN.h' file not found" while installing Perl modules

While trying to install Perl modules like JSON::XS or YAML::XS, i receive the same error:
XS.xs:1:10: fatal error: 'EXTERN.h' file not found
I use MacBook, xCode is up to date, everything else that could help is up to date too.
Since OS X El Capitan, Apple introduced System Integrity Protection which restricts writing to /usr/lib /usr/bin and other sensitive directories (even to root or sudo user) that are used by the installation of Perl bundled with the Operating System. This can cause issues when it comes to installing new modules and also if trying to install XS modules ( those linked to external C libraries ).
For this reason you should not consider the default Perl installation as a working development environment, especially if you are installing custom modules.
Check out this thread on PM and others. I had since El-Capitan managed to solve this before by manually building from tarball and adding a few params or environment variables to set the paths believing that it would be best to retain use of the system Perl but this is not the way to go. This makes your environment difficult to build but also brittle and sensitive to OS updates that may either break things in many different ways.
The best practice seems to be starting with a Perl using brew install perl and work in this environment, remembering to setup your bash_profile as directed by the installer.
Also worth remembering to do a brew link perl. If you receive warnings about this clobbering what looks like system Perl libraries don't worry - these are likely modules that were installed by you over the top and it will cause you less trouble to link over these. If you have concerns, make a note of which module installs will be cleared and re-install them once your environment is configured ( ie your module installer approach is configured using cpanm or sticking with the old perl -MCPAN -e shell etc)
This new Perl setup from brew eliminates the need to continuing running sudo which adds another layer of things that can go wrong as environment variables don't follow through and permission conflicts arise etc.
Finally to simplify package/module installation I suggest doing a brew install cpanminus. If you had previously already installed this, you can ensure the paths etc are configured by doing a brew reinstall cpanminus
If you want to take it another step further then you can install perlbrew as well which will give you the ability to run multiple versions of Perl as your user and configure these with their own libs and modules which can be very useful particularly if aligning with your production environment for testing etc.
One problem you may face if moving from system Perl to this kind of approach is needing to deal with any hangovers from installing things with sudo. It wis worth taking a little time to get all this set up right though and your issues going forward will be greatly reduced and you won't be left with that nagging feeling that you don't want to change anything for fear of it all breaking.
I have also come across a Perl Blog Article that suggests a fix for XS issues with perlbrew on Mojave
This Gist described updating your cpan shell install root though this shouldn't be necessary unless your cpan is stuck in an old config after taking steps above.
I've also raised this as a new issue on PerlMonks
After reading https://developer.apple.com/documentation/xcode_release_notes/xcode_10_release_notes#3035624 and installing the Additional headers via
sudo installer -pkg /Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg -target /
I successfully compiled without the missing 'EXTERN.h' error
In order to follow the common advice I also tried with Perlbrew to install a dedicated development version of Perl. Especially with the advice in mind First, do not use the system Perl on MacOS. The installed version is for Apple, not for you (see the discussion here: https://www.perlmonks.org/?node_id=1224727).
Unfortunately, the following error occurred:
Test Summary Report
-------------------
porting/libperl.t (Wstat: 65280 Tests: 35 Failed: 0)
Non-zero exit status: 255
Parse errors: No plan found in TAP output
Files=2653, Tests=1217766, 708 wallclock secs (52.74 usr 9.40 sys + 395.38 cusr 49.90 csys = 507.42 CPU)
Result: FAIL
make: *** [test_harness] Error 1
##### Brew Failed #####
Therefore, I decided to install it the following way (and not following the advice due to the error).
Even after having the above mentioned macOS SDK headers already installed on Catalina (macOS 10.15.2) it didn't work for me. I faced the issue during the installation of the Perl module Mac-SystemDirectory-0.13. The following steps (by identifying the missing file in hope of having a more generic approach for more or less equivalent issues) did the trick:
Locate the header file (in this case EXTERN.h)
sudo find /Library -type f -name EXTERN.h
/Library/Developer/CommandLineTools/SDKs/MacOSX10.14.sdk/System/Library/Perl/5.18/darwin-thread-multi-2level/CORE/EXTERN.h
/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/System/Library/Perl/5.28/darwin-thread-multi-2level/CORE/EXTERN.h
/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/System/Library/Perl/5.18/darwin-thread-multi-2level/CORE/EXTERN.h
Ensure the installed Perl version (here 5.18) match the header file:
perl -v | grep version
This is perl 5, version 18, subversion 4 (v5.18.4) built for darwin-thread-multi-2level
Export the path for the C-Compiler (note MacOSX10.15.sdk for Catalina and Perl Version 5.18)
export CPATH=/Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/System/Library/Perl/5.18/darwin-thread-multi-2level/CORE
Invoke the Makefile.PL with perl
perl Makefile.PL
BTW — For anybody who's still struggling with this, my workaround was:
bash% module="Sub::Util" # For example
bash% cpanm --configure-args="INC=-I/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Perl/5.18/darwin-thread-multi-2level/CORE" "$module"
Please try this
CPATH=$(dirname $(find /usr/local/Cellar/ -name EXTERN.h)) cpan JSON::XS
For Big Sur and perl 5.30, EXTERN.h is at /Library/Developer/CommandLineTools/SDKs/MacOSX11.3.sdk/System/Library/Perl/5.30/darwin-thread-multi-2level/CORE
I'm trying to upgrade CPAN itself and got that error. But I have /usr/bin/cpan and I can't write there so I have to tweak it to write the updated version to /usr/local/bin/cpan.
No promises, but yum install perl-devel worked for me.
As #huyz has helpfully pointed out, if you hit this error on a Mac, you don't have this option, even though this is probably your issue, and you need to follow one of the above methods of getting a version of Perl that isn't missing important chunks, as per other answers.
But if, dear reader, you hit this error on a linux host, as I did, then this might be an option for you.
Building on what E Lisse suggested, you might also have luck looking in
/System/Volumes/Data/Library/Developer/CommandLineTools/SDKs/
For example:
CPATH=$(dirname $(find /System/Volumes/Data/Library/Developer/CommandLineTools/SDKs/ -name EXTERN.h)) cpan JSON::XS
You could also find where EXTERN.h is located and add that to your shell by default, e.g. in your .bashrc or .zshrc file:
export CPATH=/System/Volumes/Data/Library/Developer/CommandLineTools/SDKs/MacOSX12.sdk/System/Library/Perl/5.18/darwin-thread-multi-2level/CORE/

How do I find out what version of a bower package is actually installed?

Normally a bower.json file specifies some dependencies, but these are typically expressed so that they allow a range of versions of a bower package to be used (e.g. >=1.0, which means anything higher than version 1.0).
I have an automated process which needs to find what version of a bower package is actually installed on this system right now.
How can I find this out programmatically (just the version itself), ideally using standard Unix command line tools / the bower command?
bower info <thepackagename> does not show this - it shows information about what is currently available from the bower repository (for example, even if I do bower info apackageIdonthaveinstalled it will still show a valid JSON structure containing a version number).
cat bower_components/thepackagename/bower.json | node_modules/json/lib/json.js version works for some packages (assuming the npm package json is installed), but not all (e.g. jquery 2.2.0's bower package does not contain a bower.json).
Here's a grep command to do that:
grep "version\"\:" bower_components/thepackagename/.bower.json
Also, a command to see versions of all bower components for the project - this list can be a handy CI artefact:
grep "version\"\:" bower_components/*/.bower.json
Have you ever tried "bower list --json=0 --offline".
It would list all bower packages info.
The best approach I've now found, which seems to work for every package I've come across so far, is:
cat bower_components/thepackagename/.bower.json | node_modules/json/lib/json.js version
(note the extra . in .bower.json).
It would appear that bower stores some metadata about the installed package in .bower.json, and that includes the installed version.
The best I've come up with so far is:
bower list | grep jquery | perl -pe 's/.*jquery#(.*?) .*$/$1/'
(if, for example, the package I was interested in was jquery).
That's pretty ugly for a variety of reasons:
I have to repeat the package name (although this could probably be improved
with a better Perl script which filters lines too, I'm just being lazy).
bower list gets information about all installed packages, not just the one I'm interested in - the rest of the information is discarded.
bower list seems to require internet connectivity to check the registry, otherwise it fails.
Would be interested to see if this could be improved upon, particularly the last point.

Weird Cuda C installation (Ubuntu14.04)

I just came across a weird situation after installing cuda... I literally followed every single step suggested by the nvidia website: http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-linux/#axzz3H0tm46yY.... everything seems to be fine, even the samples work... however when I try to run the "hello world" program presented in Cuda by Example documentation, the terminal displays:
The program 'nvcc' is currently not installed. You can install it by typing:
sudo apt-get install nvidia-cuda-toolkit
This is quite strange since nvcc should already be installed during the procedure suggested by the nvidia website....
Does anyone have any idea?
The PATH variable needs to include your cuda /bin directory (by default it is /usr/local/cuda-6.5/bin)
On Ubuntu you can edit the .bashrc file on your $HOME directory, adding the following;
export PATH=/usr/local/cuda-6.5/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib64:$LD_LIBRARY_PATH
If you are running a 32-bit Ubuntu version, then the paths are:
export PATH=/usr/local/cuda-6.5/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib:$LD_LIBRARY_PATH
After editing the bashrc file, source it (or simply restart the terminal):
source ~/.bashrc
Of course, if you have changed the default location of your CUDA directory you must change the PATHS accordingly.
PLUS: there are another options to tell the system where to find a library, a program, etc. like using enviromental modules, specify the complete path to the libraries and programs manually are another options. There are many methods out there!