fix: #76 accumulate stderr

[KaiPrince]

I believe the cause of CI failing is that error messages are being overwritten by subsequent writes to stderr. This PR fixes the failing CI by concatenating the stderr buffer rather than replacing it.

Related Issue

Resolves #76

Summary of Changes

1. Use += to buffer the stderr output

[thescientist13]

Not sure if this is related to CI, but I have been noticing intermittent issues in Greenwood test cases that seem related to this

• https://github.com/ProjectEvergreen/greenwood/actions/runs/18547542952/job/52868380578
• https://github.com/ProjectEvergreen/greenwood/actions/runs/18548446876/attempts/1

Where one of the new reject based test cases keeps failing on the error output
https://github.com/ProjectEvergreen/greenwood/blob/master/packages/cli/test/cases/build.plugins.error-name/build.plugins.error-name.spec.js#L49

1) Build Greenwood With: 
     Custom Configuration with a bad name value for a plugin
       should throw an error that plugin.name is not a string:

    AssertionError: expected promise to be rejected with an error including 'Configuration error: plugins must have a name. got undefined instead.' but got '\r\nNode.js v22.20.0\r\n'
    + expected - actual

    -
    -Node.js v22.20.0
    +Configuration error: plugins must have a name. got undefined instead.

I wonder if this will help?

[thescientist13]

Lol, literally the same thing just happened on my new PR here 🙃 - #78
https://github.com/thescientist13/gallinago/actions/runs/18894551028/job/53929173625?pr=78

  1) CLI Error Handling
       handling (and bubbling) an exception from the child process
         should throw an error that this is child throwing (a Promise.reject):

      AssertionError: expected promise to be rejected with an error including 'Error: Child process throwing a Promi…' but got '\r\nNode.js v24.10.0\r\n'
      + expected - actual

      -
      -Node.js v24.10.0
      +Error: Child process throwing a Promise.reject to the parent.

[thescientist13]

Do you mind writing up an issue for this to detail the specifics (to superceed #76 ), for the release notes? I will also test this out in Greenwood tomorrow too, but I think this may be something we need for sure. 👍

[thescientist13]

curious, would this be the same thing though as we had before? I'm wondering if we need to concact the entire output before doing console.error? Or maybe I'm just misunderstanding, but this does feel like there is a there there, so to speak. 🤔

[thescientist13]

Got a PR up in Greenwood with this change and nothing is breaking! :)

Will run it a couple more times but I think this one will be good to go once we clean up the title / details in the source issue.

[thescientist13]

OK, I think I understand what is happening now. Thanks for this fix!