bug: crashed while running a subgraph check

[kyle-banner]

Crash Report

name = 'rover'
operating_system = 'unix:Debian'
crate_version = '0.1.8'
explanation = '''
Panic occurred in file 'library/std/src/io/stdio.rs' at line 935
'''
cause = 'failed printing to stderr: Resource temporarily unavailable (os error 11)'
method = 'Panic'
backtrace = '''

  0: 0x56518f5c6a21 - std::io::stdio::_eprint::h0abfddebba7f1456
  1: 0x56518e338e5e - <unresolved>
  2: 0x56518e33039e - rover::command::subgraph::Subgraph::run::h4b03b805ad05df21
  3: 0x56518e2f538f - rover::cli::Rover::run::h840963b9e834130b
  4: 0x56518e2ceb1d - <unresolved>
  5: 0x56518e2cb513 - <unresolved>
  6: 0x56518e2cb529 - <unresolved>
  7: 0x56518f5d17da - std::rt::lang_start_internal::h4461fc58637f04f8
  8: 0x56518e2cf352 - <unresolved>
  9: 0x7f300e87f09b - __libc_start_main
 10: 0x56518e2c9699 - <unresolved>
 11:        0x0 - <unresolved>'''

$ rover subgraph check charter-sspp-graph@qa-cnet --schema build/schema.graphql --name=graphql-account-scs-service --query-count-threshold=100 --validation-period=1w
Checked the proposed subgraph against charter-sspp-graph@qa-cnet
Compared 112 schema changes against 21 operations
┌────────┬──────────────────────────────────────────┬────────────────────────────────────────────────────────────────────────────────────────┐
│ Change │                   Code                   │                                      Description                                       │
├────────┼──────────────────────────────────────────┼────────────────────────────────────────────────────────────────────────────────────────┤
│ PASS   │ FIELD_ADDED                              │ type `Account`: field `address` added                                                  │
├────────┼──────────────────────────────────────────┼────────────────────────────────────────────────────────────────────────────────────────┤
│ PASS   │ FIELD_ADDED                              │ type `Account`: field `newQueryField` added

// many other fields added...

│ PASS   │ FIELD_ADDED                              │ type `Mutation`: field `newMutationField` added                                    │
└────────┴──────────────────────────────────────────┴─────────────────────
+ result=101
+ set +x
error Command failed with exit code 101.

(edited by @EverlastingBugstopper just to fix up some formatting)

[EverlastingBugstopper]

@kyle-banner thanks for the report, I'm sorry you ran into this bug, we'd love to iron it out for you.

It looks like you're getting an OS Error 11 which is... definitely unexpected! Can I ask for the output of the rover info command, and also what environment you're running this on? This error is occurring in a call to Rust's eprintln! macro, which makes me think that something weird is going on with the system that's running Rover. Are you calling Rover from a Makefile or some other orchestrator?

Could you also try to reproduce this error by appending --log debug and copy + paste the entirety of what's printed?

[danpayne17]

Hi @EverlastingBugstopper . Kyle is on vacation so I'll chime in here with what we know. Here is the output from rover info:

Rover Info:
Version: 0.1.8
Install Location: /root/.rover/bin/rover
OS: Debian 10 (buster) [64-bit]
Shell: Unknown

Rover is being called from a bash shell script that I wrote. The script is being called by a GitLab pipeline runner. Attached is the debug output from the command.

Thanks for taking a look at this!

rover-debug-console-output.txt

[EverlastingBugstopper]

That is super strange! It looks like the panic occurs rrriiigghhhttt as it's about to finish printing. At the top it says there were 112 schema changes, and the table seems to include 112 passes, and then panics right around the time it's about to finish printing the table.

I've tried to reproduce this issue by printing even larger tables, but can't seem to reproduce!

Truly think this has something to do with the environment that it's running in - are you able to run the bash script you have locally and cause the same panic?

[EverlastingBugstopper]

Ah! I think I may have found the culprit. Seems like node does some funky business to set stdout/stderr streams to non-blocking mode, which doesn't play nicely with Rust.

I've found this comment describing the issue which also links to workarounds by Mozilla.

[danpayne17]

When I run locally, it works fine, but obviously I'm on a different system:

Rover Info:
Version: 0.1.8
Install Location: /Users/dan.payne/.rover/bin/rover
OS: Mac OS 10.14.6 [64-bit]
Shell: /usr/local/bin/bash

[danpayne17]

Oh nice! I will look at that comment.

[danpayne17]

That fixed it! Buried in there was that the issue goes away in later versions of node12. I upgraded and now pipeline completes successfully. Thanks for the help!

[EverlastingBugstopper]

Awesome!! Glad I could help 😄