rmanoka

The downside of our current approach is: users accidentally add this feature while building the code (and the resultant link errors are a bit dense to parse); however, I don't...

Since the docs issue has been open for a while, let's go ahead with this approach. @spadarian can you fix the CI to not use your new feature during testing?...

Great idea @michaelkirk ; tried it locally and it works. See #281

> If I understand correctly, you are generating the code dynamically, so the code calling `GDALVersionInfo` doesn't even exist when `DOCS_RS` is not set... so there is no compilation error....

I also have the same problem in [`rasters`](https://github.com/AspecScire/rasters.rs/blob/master/Cargo.toml) crate. I have included the `[package.metadata.docs.rs]` section, but the build still errors out. The previous version of the crate (which used gdal...

I seem to have added `rustc-args = [...]`, which worked in previous versions of the `rasters` crate (that uses gdal 0.7.2). However, I tried locally (as explained using [docs.rs github](https://github.com/rust-lang/docs.rs#getting-started))...

The cargo issue above seems to be fixed now; should it work now?

The cargo issue is fixed but docs.rs may also need to fix things (ref. https://github.com/rust-lang/docs.rs/issues/1580 , https://github.com/rust-lang/docs.rs/issues/1591).

I noticed this fix in `proc-macro2` crate that has a similar problem. Would it work for us? https://github.com/dtolnay/proc-macro2/pull/310/files

@spadarian @pka can you check your gdal-dependent crates and verify the docs do build? I checked [rasters.rs](https://github.com/AspecScire/rasters.rs) locally and it builds; it doesn't even seem to need the `rustc-args` configuration...