Zombies

[lnicola]

Looks like we still have a zombie process left:

And we're spawning ourselves?

Only works with proc macros (and maybe output dir loading) enabled.

[lnicola]

strace: Process 185136 attached
[pid 185136] execve("/home/me/.cargo/bin/rust-analyzer", ["/home/grayshade/.cargo/bin/rust-"..., "proc-macro"], 0x55d4f7ed4b00 /* 67 vars */) = 0
strace: Process 185137 attached
strace: Process 185138 attached
strace: Process 185139 attached
[pid 185139] execve("/usr/local/bin/cargo", ["cargo", "check", "--workspace", "--message-format=json", "--manifest-path", "/home/me/Projects/hello/C"..., "--all-targets"], 0x55d4f7ed4b00 /* 67 vars */) = -1 ENOENT (No such file or directory)
[pid 185139] execve("/usr/bin/cargo", ["cargo", "check", "--workspace", "--message-format=json", "--manifest-path", "/home/me/Projects/hello/C"..., "--all-targets"], 0x55d4f7ed4b00 /* 67 vars */) = 0
[pid 185139] execve("/home/me/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/bin/cargo", ["/home/me/.rustup/toolchai"..., "check", "--workspace", "--message-format=json", "--manifest-path", "/home/me/Projects/hello/C"..., "--all-targets"], 0x55e4516b9c50 /* 72 vars */) = 0
strace: Process 185140 attached
[pid 185140] +++ exited with 0 +++
[pid 185139] +++ exited with 0 +++
[pid 185097] --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=185139, si_uid=1000, si_status=0, si_utime=5, si_stime=1} ---
strace: Process 185141 attached
[pid 185141] execve("/home/me/.cargo/bin/rust-analyzer", ["/home/me/.cargo/bin/rust-"..., "proc-macro"], 0x55d4f7ed4b00 /* 67 vars */) = 0
strace: Process 185142 attached
[pid 185137] +++ exited with 0 +++
[pid 185136] +++ killed by SIGKILL +++

Look at 185136, the proc macro server: we start a new one (185141), then kill 185136 without reaping it.

[lnicola]

We don't wait for the child at https://github.com/rust-analyzer/rust-analyzer/blob/84c324b3dd84a191defcc490a6308a5f5c16a7f5/crates/proc_macro_api/src/process.rs#L122-L126, but I'm not sure why we run the server twice.

[lnicola]

CC @edwin0cheng I think we're calling switch_workspaces twice.

[edwin0cheng]

@lnicola that's intentional, but I forget we should only allow one proc-macro server. (#7514)