Problems running a "development" diracx container

[sfayer]

Hi,

We're trying to run a diracx services container using a source tree mapped in as a bind mount in place of the container's default diracx install (i.e. so we can run a locally patched/changed version of the code). We've mounted the main diracx git tree into the /diracx_sources path inside the container:

(base) mambauser@mycontainer-$ ls /diracx_sources
CHANGELOG.md  LICENSE  MANIFEST.in  README.md  codecov.yml  containers  diracx  diracx-api  diracx-cli  diracx-client  diracx-core  diracx-db  diracx-logic  diracx-routers  diracx-testing  docs  extensions  mkdocs.yml  pixi.toml  pyproject.toml  run_local.sh  tests

We then started up the standard ghcr.io/diracgrid/diracx/services image with the environment variable DIRACX_CUSTOM_SOURCE_PREFIXES="/diracx_sources" set, but that made no difference, it still started the built-in code.

After looking at the entrypoint.sh script, we realised it's looking through the prefix for directories ending in "-${image}" for each of the "core", "client", etc... So we set DIRACX_CUSTOM_SOURCE_PREFIXES="/diracx_sources/diracx", but that then fails with:

find: ‘/diracx_sources/diracx’: No such file or directory

To work around this we then created an empty "diracx" directory in the top of the source tree, and voilà: It starts correctly with the updated code, but this all seems a bit fragile.... How is this meant to work? Are we really meant to be setting SOURCE_PREFIXES to point to a non-existent directory in the tree, which we then have to create to make it work?

We tried to unpick how run_demo.sh does this, but it was too complicated to follow through exactly how that works.

Regards,
Simon

Mentioning @marianne013