GeyserPluginService: Use common exit flag.

[ilya-bobyr]

Problem

Geyser plugin thread would never shutdown correctly, as it is using an exit flag that is never set.

Summary of Changes

Use common exit flag.

And a minor refactoring in the use statement.

[codecov]

Codecov Report

Merging #31915 (a872c0a) into master (1d6b033) will decrease coverage by 0.1%.
The diff coverage is 0.0%.

@@            Coverage Diff            @@
##           master   #31915     +/-   ##
=========================================
- Coverage    81.9%    81.9%   -0.1%     
=========================================
  Files         757      757             
  Lines      207054   207053      -1     
=========================================
- Hits       169648   169645      -3     
- Misses      37406    37408      +2     

[CriesofCarrots]

Just to save someone else the verification I just did, #30352 added an earlier exit on L492. I was concerned the duplication was deliberate as part of hot-reloading geyser plugins. However, that PR also removed this later exit. This line was re-added in #30645 in what I believe was a bad rebase (touches adjacent code, merged one day later).

Anyway, thank you for fixing this @ilya-bobyr. I think we should backport to v1.16

[steviez]

LGTM, good find!

And thanks for tracking down the specific PR's @CriesofCarrots, agreed with your assessment as to why we had two exits. And agreed on backporting.

[ilya-bobyr]

Just to save someone else the verification I just did, #30352 added an earlier exit on L492. I was concerned the duplication was deliberate as part of hot-reloading geyser plugins. However, that PR also removed this later exit. This line was re-added in #30645 in what I believe was a bad rebase (touches adjacent code, merged one day later).

Anyway, thank you for fixing this @ilya-bobyr. I think we should backport to v1.16

Thank you for checking!
I did not look at the history, but I did trace the flag usage before making the change.
Geyser never changes the flag value. So, it will always stay false, effectively preventing the shutdown.

[ilya-bobyr]

Thinking a bit more about this.
The fact that an invalid merge did not trigger any tests means there are no tests that check valid shutdown for the individual validator services.