Question about how conda pack works, why the pack step is necessary

[jgarvin]

Checklist

• I added a descriptive title
• I searched open reports and couldn't find a duplicate

What happened?

I would like to better understand what conda-pack does. My understanding is that conda build already makes binaries relocatable by patching RPATH and other measures. If this is the case, what extra work is pack/unpack doing? Why can't conda build do it?

Additional Context

No response

[ncoghlan]

conda-pack and conda build work at different levels.

conda-pack packs an entire conda environment into a relocatable archive, not just a single conda package.

[jgarvin]

@ncoghlan thanks for the clarification, but that still leaves me wondering why conda-pack has anything to do, if all the packages inside the environment are all already individually relocatable? It sounds like you're implying I should be able to move conda build artifacts around without them breaking due to path assumptions, and a conda environment is a bunch of conda build artifacts, so I'm curious where the pathing assumptions get baked in?

[ncoghlan]

Most of what conda-pack does is simply making the archive itself. The work that conda-unpack does is mainly fixing shebang lines.

There is also some special case handling for various conda metadata files.

It's certainly possible to write your own code to do the same things that conda-pack does, and most things will still work correctly if you simply tar or zip the entire directory with no customisation of the input. conda-pack just tidies up the rough edges that arise when using a more simplistic approach.

The clearest documentation of exactly what those rough edges are is the conda-pack source code.

[jgarvin]

Thanks, I think that's enough for me to go on 👍