Bohdan Ivashko

@zuperspasm I'll need a full crash log to be able to determine what's causing it, can you launch the app with something like `db1000n 1> logs.txt 2>&1` and attaching the...

@sharkdp is this issue still relevant? it seems like receiver thread is already using buffered writer (see [code](https://github.com/sharkdp/fd/blob/5039d2db99149dfe44380dbb8e3cc8f4ddd3e25f/src/walk.rs#L389)). It even locks for the full duration it to avoid contention and...

Sure, here's the script to generate the mock data ``` use std::fs::File; use std::fs; use std::io::prelude::*; use rand::{distributions::Alphanumeric, Rng}; fn main() -> std::io::Result { for i in 0..10 { let...

rg without changes: - linux ``` 0.80user 1.34system 0:02.17elapsed 98%CPU (0avgtext+0avgdata 11540maxresident)k 248424inputs+0outputs (0major+1871minor)pagefaults 0swaps ``` - nixpkgs ``` 0.49user 0.53system 0:00.12elapsed 838%CPU (0avgtext+0avgdata 13260maxresident)k 240inputs+0outputs (1major+2253minor)pagefaults 0swaps ``` -...

This PR allows us to run some workers in consumer-only mode (we need at least one such worker or there's a minor chance of getting a deadlock in some cases)....

@ldez not sure about minimal, but here's a gh-actions run on small OSS codebase (about 5k lines of go code in total) that triggers this https://github.com/arriven/db1000n/runs/7636407897?check_suite_focus=true

I've been hitting similar issue and decided to investigate a bit. I was only able to reproduce it with `cranelift-isle` (due to [this](https://github.com/bytecodealliance/wasmtime/blob/0c615365c645776c1abf42be89edc1d5292163c8/cranelift/isle/isle/build.rs#L5)), `grpcio-sys` (due to [this](https://github.com/tikv/grpc-rs/blob/d1d8dff393281bc5166be2ac88b9bc2dff5d83a9/grpc-sys/build.rs#L496)), and `librocksdb-sys` (due...

@weihanglo thanks for the feedback, I already thought that `b)` seems more plausible but having detailed breakthrough of how cargo works definitely helps I have a feeling that another way...

A small update on this: I've started hacking around these approaches, changing [mtime_recursive](https://github.com/rust-lang/cargo/blob/dd9900585a6c2baeda170435d8641a52788af8af/crates/cargo-util/src/paths.rs#L224) to ignore directories leads to an issue where cargo no longer detects if a file gets removed...

At the same time, I can definitely see a scenario where someone wants to debug a registry crate by modifying the source in `$CARGO_HOME` (i.e. by add some debug prints...