fs.cache.enabled=false prevents server start up if other required fields are set about trino

[mosabua]

So you are saying it should just send a warning rather than produce an error.

The problem with just sending a warning would be that you basically have properties in your config that might be silently ignored. That would probably be worse.

So I think the behavior in the spilling is actually the bug.

What you can do for now is just comment out the lines ...

[mosabua]

Just had a discussion with @martint and @dain -- essentially the behavior with fs.cache is correct, and the behavior with spill-enabled is a bug.

[Sartan4455]

I do no see what is wrong with conditional settings that are ignored? To me, that's not an uncommon thing for config files. Then again, neither is commented out options. :)

I did indeed comment them out because the server errored. But what's the point of = false if = false and comment out are the same?

To me these 4 are equivalent
cache.enabled = false
# cache.enabled = false
# cache.enabled = true
<nothing>

The only difference is the first and last ones causes the server to not start if other settings are set.
I was thinking no warnings or errors.
cache.enabled is there and true, check for required fields.
If not, ignore those required fields.
By commenting them out, that's a sign they are not needed. Then if they are removed potentially valuable information(the directory/size and other settings) could be lost. However I can see if they were old bad settings they'd might silently just work.

I am fine with either way really. It just seemed odd to me, thus the bug.

[wendigo]

Setting spill configuration when spill is disabled is a bug, not the other way around

[mosabua]

There is really no right answer with this .. and the Trino convention is as implemented for caching. I will therefore close this bug. If anyone wants to send a PR to fix the behavior for spilling .. that would be good but can be tracked separately.