This PR here brings back example scraping. Unfortunately the rustdoc book is also stale so I had to go deep into the tracking issue to figure out how it changed. The feature can now be disabled on a per example basis, but I enabled it for all of them like was intended before.
I also enabled all the features for docs.rs builds. We used to disable svg and image due to duplicate artifacts. The duplicate artifacts tracking issue is still open, however rustdoc no longer seems to complain for our builds. Perhaps the code changed or the issue is more limited now? Not sure, but I think we can keep the features enabled for now.
I tested this on my own machine with the nightly toolchain and everything seems to work fine.
Critically, I also updated our CI tests. The docs are now built with the nightly toolchain, as that matches docs.rs. I also updated the flags to be a closer match to what docs.rs is configured to use. This way we can hopefully prevent another event where a published build fails on docs.rs. On top of that I made sure we enable our various features in all sorts of CI tests and removed libx11-dev as a seemingly stale Linux dependency.
Docs.rs failed to build Druid v0.8.0 docs because we were specifying a stale unstable nightly rustdoc option in our
Cargo.toml.I temporarily disabled this option in #2348.
This PR here brings back example scraping. Unfortunately the rustdoc book is also stale so I had to go deep into the tracking issue to figure out how it changed. The feature can now be disabled on a per example basis, but I enabled it for all of them like was intended before.
I also enabled all the features for docs.rs builds. We used to disable
svgandimagedue to duplicate artifacts. The duplicate artifacts tracking issue is still open, howeverrustdocno longer seems to complain for our builds. Perhaps the code changed or the issue is more limited now? Not sure, but I think we can keep the features enabled for now.I tested this on my own machine with the nightly toolchain and everything seems to work fine.
Critically, I also updated our CI tests. The docs are now built with the nightly toolchain, as that matches docs.rs. I also updated the flags to be a closer match to what docs.rs is configured to use. This way we can hopefully prevent another event where a published build fails on docs.rs. On top of that I made sure we enable our various features in all sorts of CI tests and removed
libx11-devas a seemingly stale Linux dependency.