empty file list when asset directory is in nested folder

[yatesco]

Hi,

Given:

\ (the root of cargo workspace)
\Cargo.toml
\assets\ (a simple folder)
\component-a\ (a crate within the workspace)
\component-a\Cargo.toml
\component-a\assets
\static\ (a simple folder)
\static\assets

where each "\assets" directory is a copy with at least a.txt in, then #[folder = "assets/"] is the only one that works when used within \component-a.

Observations:

• #[folder = "component-a/assets"] compiles but lists no files
• #[folder = "static/assets"] compiles but lists no files

this is regardless of whether cargo run is run within \ or \component-a.

What I am trying to achieve is either component-a/assets/ or static/assets/.

Help :-)

[yatesco]

Actually, it is slightly more subtle: \assets\ must exist but it can be empty, the actual assets are loaded from \component-a\assets.

If \assets\ is missing then it won't compile. If \component-a\assets\ is missing then it will compile but not find anything.

[yatesco]

Nope...even more subtle ;-). For developing \assets\ must exist but the actual assets come from \component-a\assets but for a release build \assets\ is used :-(.

[yatesco]

Last comment (for now :-)).

Synchronicity (and thus sanity!) can be achieved by enabling the feature "debug-embed" which removes the need for \component-a\assets but does have the cost of needing a rebuild for each change to an asset.

[AzureMarker]

Can you provide a link to your code? Your explanation is hard to follow. Are you using a symlink?

[yatesco]

No symlinks (this is on Windows) and I’ll get a reproducible repo online tomorrow:-)

[yatesco]

Actually, the whole thing is simpler than I thought, and is related to #63

Regardless of whether using debug or release, the value in the folder directive is relative to the current directory, which is really confusing ;-). It would be nice if it could always be relative to the workspace root.

Repo is at https://github.com/yatesco/rust-embed-bug-with-workspaces

[yatesco]

Note to self: check if bbe6eea fixes this.

[AzureMarker]

@yatesco Is this fixed for you as of #101?

[AzureMarker]

Tested the reproduction code with the latest version (5.5.1), and the issue is fixed. The embedded path no longer depends on where you start cargo run.