Installation/setup instructions?

[bmp192529]

Hello,

I have tried many different ways to get this package to compile, but, even if I can get the initial build process to work, I always run into issues during the run the shared libraries not linking.
Does the RDKit licensing prevent you from including RDKit as part of the GitHub directory itself? It seems like it would be simpler and more self-contained to just have all of the RDKit dependencies as part of the git and not reference outside libraries. You might have a lot of good reasons why not to do this, I just thought I should mention it.
I haven't found any installation instructions that specify where the custom RDKit build is to be installed anywhere on the git, are any available? I even tried reworking the old script that was being used to generate the RDKit Debian package from the RDKit git and that did not work either.
I could be doing something completely wrong, but "dynamic-linking-from-conda" does not work at all for me either. I looked at build.rs under rdkit-sys and it seems to be requesting directories that aren't there in my conda files.
I was previously able to get the other rdkit rust interface -rdkit cffi - to compile, but that interface is relatively limited in features compared to yours.

I would greatly appreciate any help with this issue.

[bmp192529]

So, I figured it out myself after a lot of digging through random files here as well as trial and error. Here are some instructions for anyone else looking to install. My system is Fedora 41. This should work for other linux distributions, potentially with some small changes.

1. Download the appropriate precompiled tar file for your CPU architecture.
2. Extract the tar file. In the extracted output folder "rdkit-Release_2024_09_1", you should see the folders "Code" and "build". Open a terminal window inside the "rdkit-Release_2024_09_1" directory and move the contents of these folders with terminal commands:
   "sudo mv Code /usr/local/include/"
   "sudo mv build/lib/* /usr/lib/"

• The first command moves all of the code into the include directory, and the second command moves all of the .so "shared object" executables as individual files into the /usr/lib folder

3. After hitting cargo clean and cargo run, everything now compiled, but failed to launch since it was missing some boost .so files (I believe it was libboost_iostream.so.1.74.0 ?). I resolved this by running this terminal command before doing cargo run again:
4. "export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib"
5. This still didn't run for me because the compiler was looking for libboost_iostream.so.1.74.0 specifically for some reason (and not the generic symlink?) and I didn't have that version of boost installed on my system; I have 1.83.0 installed. The universal answer from a quick google search was that you can't or really really shouldn't have two versions of boost installed simultaneously, so I didn't go that route. I resolved the problem by going into my /usr/lib folder and creating a symlink named "libboost_iostream.so.1.74.0" that actually points to /usr/lib64/libboost_iosteam.so.1.83.0 and rerunning cargo run. This felt sketchy honestly but it works! Hopefully I'll eventually figure out how to get the compiler to outright use the correct version but this works for now. This issue likely relates to the rdkit tar file being compiled on someone else's computer, who may have been using the 1.74.0 version of boost when they compiled it. I'm not sure what modifications they made prior to compiling, so recompiling myself isn't really an option.

Use at your own risk of course and be careful messing with your system files.

I used the "test_suite.yml" file available at /.github/workflows to help figure this out.

[bmp192529]

Here is a much better explanation/procedure that should fix any issues with Boost dependencies by building the code yourself from source:

1:
Install the required dependencies of boost libraries, cmake, make, etc. The example they gave in the workflow was the following:
sudo apt-get update | sudo apt-get install -y build-essential libssl-dev libboost-iostreams-dev libboost-serialization-dev pkg-config curl tar
I used the boost packages for Rocky Linux 9.5 (Version 1.75.0-10.el9) from the standard @appstream repositories and it worked fine without modification; I think using the pre-compiled libraries creates a lot of issues for people who aren't using that specific version of ubuntu with that specific version of boost and CPU architecture. In a way, it is much easier to just build from source, so that it will work with what is currently installed on your system.

2:
Download an RDKit Release. Go to https://github.com/rdkit/rdkit/releases and download an archive of a suitable release. I choose to use 2024_09_6 since it was similar to the 2024_09_1 version they used but contained some bug fixes.

3:
Extract said archive somewhere, in a temporary location if you want. In my example, we'll call it folder:
/folder/rdkit-Release_2024_09_6 would be the extracted contents of my example archive.

4:
In terminal, navigate to /folder/rdkit-Release_2024_09_6 or the appropriate version/location for you. Enter the commands:

mkdir build

cd build

cmake .. -D RDK_BUILD_PYTHON_WRAPPERS=OFF \
             -D RDK_OPTIMIZE_POPCNT=OFF \
             -D RDK_INSTALL_COMIC_FONTS=OFF \
             -D RDK_BUILD_FREETYPE_SUPPORT=OFF \
             -D RDK_INSTALL_STATIC_LIBS=ON \
             -D RDK_INSTALL_INTREE=OFF \
             -D RDK_BUILD_SWIG_JAVA_WRAPPER=OFF \
             -D RDK_BUILD_CPP_TESTS=OFF

and, finally:
sudo make install -j 10
The number after -j meaning how many CPU cores you want to allocate to the build. This part I modified from this dockerfile for the old debian package for this project. Keep this terminal window open to see where the installed files went.

5:
In your cargo project (with rdkit already added as a dependency) in another terminal screen, run "cargo clean" and then "cargo run". If you get an error about missing shared libraries, you just need to add the containing folder to your LD_LIBRARY_PATH. For example, I got this error:

target/debug/rdkit-rust: error while loading shared libraries: libRDKitDataStructs.so.1: cannot open shared object file: No such file or directory

I solved this by adding the /usr/local/lib directory to my LD_LIBRARY_PATH in terminal:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib

You can put that line in the end of your .bashrc file in your home folder if you don't want to have to put that line through every time you open a terminal to work on your project. Now, cargo run with an example program should work. I hope this helps someone.