Add CI for testing pyflask distributable

[CodyCBakerPhD]

Re-attempt #292 from a better base

[CodyCBakerPhD]

Unfortunately to see if the workflow works, it must be merged to main to trigger

[CodyCBakerPhD]

There was some confusion on my part about this - after double checking the flags to pyinstaller, I see that the default path of your tests is indeed the 'distributable' (./build/flask/nwb-guide/nwb-guide.exe) and not the 'build' (./build/nwb-guide/nwb-guide.exe)

I've simply added the tests on the build, as PyInstaller refers to them, to be exhaustive in case there ever could be any difference between the two

[garrettmflynn]

Oh sure. For some reason, I'd thought you meant that you'd target the dist/mac/NWB GUIDE.app/Contents/Resources/nwb-guide/nwb-guide file that is produced after copying the distributable file to the Electron app.

TBH I'm never touching the 'build' version, which is contained in a folder that primarily contains Pyinstaller logs, so we might just ignore it since it's unused.

[CodyCBakerPhD]

🎉Able to reproduce scipy issue in stable windows CI: https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5898618230/job/15999924669?pr=297

Mac indeed appears to be in some kind of infinite loop - your test should simply exit out if the server can launch successfully right? https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5898618230/job/15999924617?pr=297

[CodyCBakerPhD]

And then Linux appears to be missing the json schema paths: any idea how to fix? https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5898618230/job/15999924554?pr=297

[garrettmflynn]

Mac indeed appears to be in some kind of infinite loop - your test should simply exit out if the server can launch successfully right? https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5898618230/job/15999924617?pr=297

Actually, I didn't configure this properly for a test. Currently, there are errors thrown if the process fails—though the process remains open infinitely (as Flask does) if successful.

I'm listening for startup time in the proc.stdout.on("data", () => {}) function.Just inserted a call to process.exit() so the main process closes gracefully. Hopefully this will work?

[garrettmflynn]

And then Linux appears to be missing the json schema paths: any idea how to fix? https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5898618230/job/15999924554?pr=297

This is fixed in #284. We'd want to rebase to that branch if we want those changes here.

[CodyCBakerPhD]

This is fixed in #284. We'd want to rebase to that branch if we want those changes here.

In that case what I propose as the simplest approach is, despite the currently failures (they are not required to pass to merge) we go ahead and get this into main so every PR can benefit and test fixes specific to each branch

What do you think?

[garrettmflynn]

Yeah I agree. I'll definitely need to clean up the error catching (as the ones thrown by Mac aren't caught by my fix)—but at least this will let us test our current versions.

My local builds are stalling, otherwise I'd fix it now.

[CodyCBakerPhD]

My local builds are stalling, otherwise I'd fix it now.

Should we wait for those test function improvements before merging this then?

By stalling do you mean you're also seeing this infinite loop behavior? I think you mentioned it might have been because of those recent changes I made to the Flask version fetching so feel free to disable that if needed

[garrettmflynn]

Just pushed a preliminary version of those so I don't lose the changes.

By stalling do you mean you're also seeing this infinite loop behavior? I think you mentioned it might have been because of those recent changes I made to the Flask version fetching so feel free to disable that if needed

No, currently the build will run without logs as long as I wait (~5min). For some reason, my local build is being finicky.

[CodyCBakerPhD]

No, currently the build will run without logs as long as I wait (~5min). For some reason, my local build is being finicky.

Does checking out a point on main from a couple days ago work as expected?

[garrettmflynn]

Actually this is also happening with npm run dev:server. Something must be up...

[garrettmflynn]

Will yours run on Mac? Trying to add logs to determine the point of failure but it's as if nothing is actually run.

[CodyCBakerPhD]

Will yours run on Mac? Trying to add logs to determine the point of failure but it's as if nothing is actually run.

Checking out and pulling latest main and calling npm start or npm run dev:servercurrently works fine on my M1

I run into these kinds of things from time to time though - if you wipe everything (including git clone) and hard reinstall everything (including the npm ci or npm install stuff) it might go back to normal - or if not at least that's informative

[garrettmflynn]

Just restarted my computer and both work fine now.

[CodyCBakerPhD]

Good to hear - oh, hey, the mac test now fails with an actual errror! https://github.com/NeurodataWithoutBorders/nwb-guide/actions/runs/5904881769/job/16017892212?pr=297

[garrettmflynn]

Alright should be good to merge. Tests failed but my new code works to close the server if an error / success is detected.

[CodyCBakerPhD]

@garrettmflynn Cool, just need approval

[garrettmflynn]

Approved! Let's see if this helps on the active branches.