[BUG] Timestamp::now() exposed via hdk, but always panics about holochain

[mattyg]

i found that chrono::prelude::Utc::now() also causes a runtime error. (hc 0.2.3-rc.1)

Yeah we can't get chrono timestamps in the wasm guest (need to ask the host for timestamp) -- that's also why Timestamp::now() panics (it's just calling Utc::now() under the hood). We have a function sys_time() for use within the wasm guest.

I think the only issue here is that Timestamp::now() either shouldn't be exposed via the hdk, or should have a feature flag to switch its implementation to use sys_time() when compiled for wasm targets

[mattyg]

I see, then yeah it seems like there is just something funky with the feature flags being passed into kitsune_p2p_timestamp when running holochain feature test_utils (although that shouldn't affect the flags on the hdk used within the dna, right?)

Also the docs for Timestamp::now() reference a different implementation in holochain_types that doesn't actually exist.

[mattyg]

It looks like the feature full from kitsune_p2p_timestamp needs to get bubbled up to the holochain crate and set as part of it's own test_utils feature.

Edit: maybe not, I think I'm confused -- is there supposed to be a Timestamp::now() implementation that works in wasm? Or are we supposed to use sys_time() -- if thats the case maybe it just needs to be removed from hdk / hdi

[jost-s]

Hi @mattyg, yeah, I've noticed this doesn't work too. Thanks for the reproduction, it's reproducible.

According to the Timestamp docs it's not supposed to work in WASMs https://github.com/holochain/holochain/blob/develop/crates/kitsune_p2p/timestamp/src/lib.rs#L46-L47

The docs contradict the fact there is a Timestamp::now(), so needs updating too.

I don't know what's the recommended way here, but you can use sys_time as well as chrono::prelude::Utc::now().timestamp().

[nphias]

i found that chrono::prelude::Utc::now() also causes a runtime error. (hc 0.2.3-rc.1)

[ThetaSinner]

I've got a fix for this. It is feature indeed an unhelpful combination of features through dependency crates. I found a reasonably nice solution in holochain_zome_types that resolves the issue, with a few other tidy-ups.

I even managed to make a test which ensures this stays unavailable which I think is pretty cool, because I get excited by testing tools / libraries that do that kind of thing :D

[ThetaSinner]

Back-ported in #3121