Taiki Endo

> inline `asm!` won't ever be optimized _out_ either unless it's given the `pure` flag ~~IIUC, this is due to ["LLVM doesn't have attributes to represent ReadOnly/ReadNone + SideEffect"](https://github.com/rust-lang/rust/blob/a8314ef7d0ec7b75c336af2c9857bfaf43002bfc/compiler/rustc_codegen_llvm/src/asm.rs#L279-L290).~~ EDIT:...

> IIUC, this is due to "LLVM doesn't have attributes to represent ReadOnly/ReadNone + SideEffect" Oh, considering that LLVM has [InaccessibleMemOnly (ReadNone + SideEffect)](https://github.com/rust-lang/rust/blob/a8314ef7d0ec7b75c336af2c9857bfaf43002bfc/compiler/rustc_codegen_llvm/src/asm.rs#L286-L287), my understanding here is wrong. (Is...

> I don't see why `ReadOnly/ReadNone + SideEffect` would allow removing `asm` block offhand/can't think of example. But noted that maybe `pure` alone determining volatility is subject to change. volatile...

> _Even with your correction in the next comment_, I interpreted your comment to mean: > > "Right now, LLVM will treat inline asm as volatile as long as you...

> How are these custom compilation flags added currently? These are set by the build script: https://github.com/rust-embedded/riscv/blob/a9d3e33ff9d321bb0b19e260c66e1d5fa7a4221c/riscv/build.rs#L4-L12 However, `target.starts_with("riscv32")`/`target.starts_with("riscv64")` is not very robust and will not work if the target...

> With #205, the code `panic`s from the `unimplemented` functions. For some reason, the custom `cfg` arguments are not making it through to downstream users. Any thoughts? The cfg set...

> For reference I've been playing with removing `yaserde` in favor of `quick-xml` [here](https://github.com/luca-della-vedova/urdf-rs/tree/luca/quick_xml). > I reached the point where all the existing unit tests work (including serialization / deserialization...

One option would be to add a method that returns `Option` based on `JointType` like [this](https://github.com/openrr/openrr/blob/38b67711531ff339b4a037badd4e9f4150c6436b/arci/src/clients/joint_position_limiter.rs#L66). Given that it would complicate the parsing code, it would not necessarily be easy...

Thanks for the report! This is another regression introduced in https://github.com/openrr/urdf-rs/pull/64. I confirmed urdf-rs 0.6 (that uses serde-xml-rs) doesn't have this problem.

> I noticed that `urdf-rs` also fixes a patch version No, `"1.0.7"` means `">= 1.0.7, < 2.0.0"` (not `"=1.0.7"`) because the [cargo's default version requirement strategy is caret requirements](https://doc.rust-lang.org/cargo/reference/specifying-dependencies.html). https://github.com/openrr/urdf-rs/blob/e07b0ea013b47223213708a15430a2350c4ef02e/Cargo.toml#L16...