LSP login issues

[tresf]

Reported by @regulus79 on Discord:

A user on the forums mentioned that the sharing platform logs them out immediately, so they can't comment. I tried, and it happens for me too. I can log in, but once I click on a link, it logs me out again

... and @Monospace-V reminded of this solution that was posted on Discord:

Fixed. Not sure why the cache got blown away.

sudo mkdir -p ~/lmms.io/var/cache/lsp && sudo chown www-data:www-data ~/lmms.io/var/cache/lsp

If this happens again, there's a chance our git deploy plugin might need that directory whitelisted.

I can confirm that ~/lmms.io/var/cache/lsp is missing... but what's weird is it's explicitly added in our hooks/post_deploy script... Quoting:

lmms.io/hooks/post_deploy

Line 9 in b6e7afc

mkdir -p var/cache/lsp/

... which leads me to believe that this script isn't running.... I read the rest of the script and found this line:

lmms.io/hooks/post_deploy

Line 4 in b6e7afc

export APP_ENV=prod

... which seems to match the symptom described in:

• Why do deployments need APP_ENV=prod? #377

... so I have reason to believe that post_deploy isn't always running after a deployment.

[tresf]

Hmm... the webhook says it's never been run, which is a bit fishy. Did anyone recently delete and re-add it? Recent deliveries is empty, but we've had a PR in the last week, so I'd expect some events in there.

I ran the webhook manually from my browser and can see that post_deploy worked just fine, so I'm still scratching my head as to the cause. Perhaps something with post_deploy is failing intermittently however I would expect the deployment to fail in this scenario. Thoughts welcome.

At time of writing this, the lsp cache directory has been properly created by the post_deploy, LSP is functioning again.

[tresf]

Reported again via Discord... running the script again. Why the heck is post_deploy so intermittent?