Home → Problems →

Windows miniconda install of conda-forge packages fails with picosat: out of memory in 'resize'

• Edit

Kris Thielemans

17 Mar, 2025 06:44 PM

Hi

since about 1 month all our appveyor builds fail when trying to install some packages via (mini)conda. This happens on both our jobs - (using the Visual Studio 2022and Visual Studio 2019 image resp.).
Relevant lines in our appveyor.yml are https://github.com/UCL/STIR/blob/master/.appveyor.yml#L34-L39

Example log is https://ci.appveyor.com/project/KrisThielemans/stir/builds/51707240..., log is attached, but here is the relevant bit

for /D %%d in (C:\Miniconda*-x64) do set MINICONDA=%%d
echo Using Miniconda %MINICONDA%
Using Miniconda C:\Miniconda38-x64
set PATH=%MINICONDA%;%MINICONDA%\Scripts;%MINICONDA%\Library\bin;%PATH%
conda install -c conda-forge  -yq libparallelproj swig numpy pytest matplotlib
Collecting package metadata (current_repodata.json): ...working... done
Solving environment: ...working... failed with initial frozen solve. Retrying with flexible solve.
Solving environment: ...working... failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): ...working... done
Solving environment: ...working... *** picosat: out of memory in 'resize'

Any suggestions?

• log_(1).txt 1.13 KB

1. Support Staff 1 Posted by Feodor Fitsner on 17 Mar, 2025 07:13 PM

   

   Hi Kris,

   I'm not an expert in Conda, but I've asked ChatGPT about that error and got this reply: https://chatgpt.com/share/67d8728d-c4c0-8000-9c65-4587842541c0

   There is a number of recommendations you may try and one of them is increasing memory. I hardly believe dependency resolution in any tool/program should normally consume all VM memory (6 GB, I've seen the worst case where Java process takes 2 GB while running Gradle build) and take 50 minutes to resolve, but you can check if it's memory by connecting to VM via RDP and observing the process: https://www.appveyor.com/docs/how-to/rdp-to-build-worker/ Check mamba as GPT suggests. Hope that helps.

   • Edit

2. 2 Posted by Kris Thielemans on 18 Mar, 2025 06:30 PM

   

   Many thanks Feodor.

   Indeed that helped. I made the mistake of not creating an new conda environment. There was apparently some conflict between the base version on what I tried to install.

   I'm currently still struggling with getting this to work, but my struggles hav nothing to do with appveyor, so we're all good here.

   Thanks again

   Kris

   PS: here are my struggles! https://github.com/UCL/STIR/pull/1578

   • Edit

3. Support Staff 3 Posted by Feodor Fitsner on 18 Mar, 2025 08:12 PM

   

   Nice, thanks for the update!

   • Edit