Jeff Garzik

My guess would be forcing the dialect via `-std=c++98` or `-std=c++0x` in a special Travis compile matrix

Hitting this issue also. Test case demonstrating problem: ``` use lmdb::{Cursor, Transaction}; use std::path::Path; fn main() { let cfg_builder = lmdb::Environment::new(); let path = Path::new("./test.db"); let env = cfg_builder.open(path).unwrap(); let...

ACK, for my small part

For the moment, users must assume the risk and model where an entire payload for a single http transaction must fit in memory. Adding strikethrough, above.

Pull request looks good to me. More comprehensive than #3771. Maybe a `main`-branch-first approach is ideal?

@petertodd, what do you think? We could merge your pythonize branch before it grows further mold.

Until merged, the branch is up at https://github.com/jgarzik/kvdbd/tree/rustls Unfortunately, a problem was found that stops work on this issue: https://github.com/jgarzik/rustls-pemfile-issue

sled test results: 1G keys are ok. Presumption: key and value sizes are 64-bit limit?

@jakeschurch Thanks for advancing the issue. I think you indeed found the bug (multi-bin related). The protoc issue is unrelated; apparently it wants updating to 3.10.x or so.

Agreed. I'm happy to merge changes like this into upstream pynode.