Johannes Löthberg

Is there anything more than the CMake part of this that's missing?

I just did some quick test runs and without `llvm-cov` running our test suite takes ~40s both with and without `nextest` (Most of the test needs to talk to a...

Personally the reason I just kind of assumed that it only supported hard-coded elements was reading the following two paragraphs from the book: > The view macro expands an HTML-like...

I just ran into the same thing. I could fix the git errors pretty easily by deleting the corrupted commit object, but I'm not really sure how to fix the...

We've also just started running into this after implementing some new functionality which depends on restarting threads for error recovery. We can confirm the same behavior of `CURRENT` being set...

Okay, we believe that we've found the root cause which appears to be the Rust `std` depending on semantics that POSIX does not guarantee. It seems that `std` is ([here](https://github.com/esp-rs/rust/blob/5944fc5cd231b5aaf09b471def70ecefefa4737d/library/std/src/sys/thread_local/guard/key.rs#L33-L60))...

> > On the other hand, the Rust std explicitly does not guarantee that thread cleanup happens, and so from a correctness standpoint this memory leak is okay but suboptimal....

> Yes but what you quote is different: they basically say "there are caveats with running TLS destructors when the program is about to exit". That is what you're _inferring_...

> > At thread exit, if a key value has a non-NULL destructor pointer, and the thread has a non-NULL value associated with that key, the value of the key...