#development
Title
h
hundreds-father-404
02/12/2021, 3:07 AM@witty-crayon-22786 I get core dumps deterministically when trying to test remote execution locally in my devbox. are you still hoping for those?
w
witty-crayon-22786
02/12/2021, 6:03 PMthe backtrace is actually an interesting one
đź‘€ 1
Copy code
#0 0x00007fb9dadb7588 in __rust_probestack () from /data/home/eric/pants/src/python/pants/engine/internals/native_engine.so
#1 0x00007fb9d9eb4b8a in task_executor::Executor::future_with_correct_context (future=<error reading variable: value of type `core::future::from_generator::GenFuture<generator-0>' requires 120728 bytes, which is more than max-value-size>)
at /data/home/eric/pants/src/rust/engine/task_executor/src/lib.rs:158
#2 0x00007fb9d9eb501e in task_executor::Executor::spawn (self=0x55a8d29a69b8, future=...) at /data/home/eric/pants/src/rust/engine/task_executor/src/lib.rs:104
#3 0x00007fb9d9f847dd in <process_execution::remote_cache::CommandRunner as process_execution::CommandRunner>::run::__run::{{closure}} () at process_execution/src/remote_cache.rs:509
...đź‘€ 1
h
hundreds-father-404
02/12/2021, 6:04 PMIs it insightful beyond my workunits change?
w
witty-crayon-22786
02/12/2021, 6:04 PMmaybe not. it’s panicing in the call to spawn in
future_with_correct_contextđź‘Ť 1
…oh. which is … maybe a stack overflow?
h
hundreds-father-404
02/12/2021, 6:07 PMiiuc, the Future is too large? We do have several layers of nested async futures being passed around
w
witty-crayon-22786
02/12/2021, 6:08 PMyea. this is still not a solved problem. blugh.
i’m not sure if this calls for a convention (Box whenever you X), an investigation (is there a main culprit), etc
đź‘Ť 1
f
fast-nail-55400
02/12/2021, 7:24 PMlast time I had to deal with such a thing, I just put in log statements until I narrowed down to which future was actually blowing up the stack and put a box there
w
witty-crayon-22786
02/12/2021, 7:26 PM…i wonder if there is a way to “sizeof” them …?
f
fast-nail-55400
02/12/2021, 8:45 PMI had tried to do that as well, but didn’t get anywhere with it. mainly because it doesn’t tell you what else is on the stack at the same time that together blow the stack up
it would be nice if Rust had an annotation similar to the type name limits to force an error if a generator/future goes over a certain size
đź‘Ť 1
w
witty-crayon-22786
02/12/2021, 8:46 PMi was thinking that you could pepper size calls at various boundaries, and then add/remove boxes until the sizes were reasonable
yea, this is awkward. two or three rust warts this week
h
hundreds-father-404
02/13/2021, 12:41 AMWoot that fixed it! Thanks!
đź‘Ť 1
w
witty-crayon-22786
03/19/2021, 3:14 AMrelated: https://rust-lang.github.io/wg-async-foundations/vision/status_quo/alan_runs_into_stack_trouble.html … and there are actually already some tips in there.
Copy code
RUSTFLAGS = "-Zprint-type-sizes"