Simon Guest

Perhaps a good solution here would be to introduce a new mode, `verbatim` say, for which every key by default is sent to the underlying program, after which the previous...

It's not unmaintained, but xiaq recently started a new job and has not had much time to work on Elvish recently. Development has indeed been a little slow recently. But...

I will try that, but not imminently sorry, am away from my computer. Thanks for posting!

I extracted the [PSI/J executor as a separate Python package](https://github.com/AgResearch/redun_psij), available on [PyPI](https://pypi.org/project/redun_psij/) or as a [Nix flake](https://github.com/AgResearch/redun_psij/blob/main/flake.nix). Perhaps that's enough at least until others start using it?

Actually I have switched my [Nix package fetcher](https://github.com/AgResearch/redun.nix/blob/package-0.32/flake.nix#L21) to fetch from PyPI rather than GitHub, and since that works fine provided there's a source tarball on PyPI, I can live...

Maybe related to #254

Yes, I am seeing the readonly `STEEL_HOME` problem on Nix. But according to my just-edited description above, I am currently a little disoriented about exactly when it fails. But I...

`add_search_directory` sounds good, thanks for the pointer! I do think it needs to be separate, as my application surely needs to be deployed as an application depending on vanilla Steel,...

The search path solves my original problem for sure. I need to further understand the cases where Steel fails to open its history file. I'll come back with more info.

I have looked into this some more, and my conclusion is that we would benefit from a slight enhancement to how Steel finds its cogs, perhaps using `STEEL_COGPATH` as above....