-
Notifications
You must be signed in to change notification settings - Fork 28
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Need some way to lock the version of boost to what we build against #71
Comments
@rvianello : the immediately obvious way to do this would be to just directly match the version numbers in the |
Since a The section with title "Pinning at the variant level" at https://conda.io/docs/user-guide/tasks/build-packages/variants.html provides an example that specifically addresses the boost case. The following section ("Pinning at the recipe level") illustrates how to obtain a similar result in the meta.yaml file alone. |
Does this already work? I merged all changes @greglandrum and @rvianello did in their forks and build in a fresh container, it build RDKit 2018.03.3, but at the end it is missing boost-1.65.1 and the container somehow has 1.67 installed: `The following NEW packages will be INSTALLED:
Preparing transaction: ...working... done |
@markussitzmann : when I do a fresh build using the development branch of this recipe everything works fine. Here's what gets installed for me on a centos6 container:
|
@greglandrum Thanks for the testing and reply - although I hate it :-) ... I definitely have put boost-1.65.1 everywhere and diff'ed with yours and Riccardo's branches/forks. But somehow it ends up with boost-1.67. I can retry only tonight. On which distro did you build? Centos? I build on debian:stretch |
@markussitzmann : It also didn't work for me when I tried to force boost 1.65.1. The current branch is set to use 1.65, which grabs 1.65.1 by default. I'm using centos, but that shouldn't make a difference. Can you point me to the code you're using to do the build? |
While looking for the code I might have found the error I made but as I said I can only try tonight again (did the fixes in my repository but checked out the wrong branch during the Docker build - and my attempts were revolving around this 1.65 vs 1.65.1 thingy). Thanks again - I only will come back if I still struggle ;-). |
@greglandrum: hmm, when I run the development branch of this recipe here, the build works now, however now I receive boost 1.63 as built boost package and then starting rdkit fails because it is missing 1.65.1 ... a bit mysterious. |
I run this Dockerfile: https://github.com/chembience/chembience/blob/master/context/base/rdkit-conda-compile/Dockerfile after replacing line 44-46 with the standard conda-rdkit repository and branch development: RUN git clone https://github.com/rdkit/conda-rdkit.git |
Unless you need it for something else, you can probably skip the boost build now (line 51 in your Dockerfile); that will use the boost build recipe from the RDKit channel, which will use 1.63. You almost certainly don't need to build eigen either. I don't see anything else obviously wrong in the dockerfile. When you get failures: do they happen during the build or later when you try and use the contents of the image? |
FWIW: these two containers are what I am using to do conda RDKit builds: |
@greglandrum Thanks for your hints. The builds fail in different places depending what I try. Sometimes during conda is packaging, sometimes when I start RDKit which then misses the right boost version (this happens with a subsequent image which copies all compiled packages from the image I build with the Dockerfile above). Yep, and I need to give my Dockerfiles another clean up, they have collected old, overcautious or unnecessary stuff. I am also considering to just switch to your already available anaconda packages; the reason for compiling RDKit myself was that very early last year, when I started to turn my preliminary works into something real, the anaconda RDKit packages still were on Python 3.5 and I had decided to go for Python 3.6 (BTW, Python 3.7 is out now for a few months already :-)) |
Yeah, assuming that there is an anaconda distrib that supports it by then, I will do Python 3.7 builds (and probably retire 3.5) for the 2018.09 release. |
I have exactly same problem :ImportError: libboost_python3.so.1.65.1: cannot open shared object file: No such file or directory |
Boost version numbers end up in the library names, so using DLLs from newer boost versions does not work.
This is connected to this rdkit issue.
rdkit/rdkit#1957
The text was updated successfully, but these errors were encountered: