templates check: Add --stabilise flag to make renders reproducible
#5214
Conversation
Deploying matrix-authentication-service-docs with
|
| Latest commit: |
f678334
|
| Status: | 🚫 Build failed. |
reivilibre
force-pushed
the
rei/templatecheck_stabilise
branch
from
be3ceff to
4793f53
Compare
crates/spa/src/vite.rs
Outdated
| for name in &[ | ||
| "src/shared.css", | ||
| "src/templates.css", | ||
| "src/main.tsx", | ||
| "src/swagger.ts", | ||
| ] { |
There was a problem hiding this comment.
I don't particularly like that this list is hard-coded like that, especially since you might have a fork of the templates & assets that include extra assets handled by the vite build pipeline?
Instead, if feasible I would have a 'fake' version of include_assets that doesn't use the manifest at all?
crates/cli/src/commands/templates.rs
Outdated
| // XXX: we should disallow SeedableRng::from_entropy | ||
| let mut rng = rand_chacha::ChaChaRng::from_entropy(); | ||
| let now = if stabilise { | ||
| DateTime::from_timestamp_secs(0).unwrap() |
There was a problem hiding this comment.
Could we use a more believable timestamp so that the IDs look like they are not all zeros?
|
|
||
| impl<T: TemplateContext> TemplateContext for WithCaptcha<T> { | ||
| fn sample( | ||
| fn sample<R: Rng + Clone>( |
There was a problem hiding this comment.
hmmmm a cloneable Rng is a bit suspicious at this stage? 🤔
There was a problem hiding this comment.
It's only samples :D
| /// | ||
| /// Returns an error if any template fails to render with any of the sample. | ||
| pub(crate) fn all(templates: &Templates, now: chrono::DateTime<chrono::Utc>, rng: &mut impl rand::Rng) -> anyhow::Result<::std::collections::BTreeMap<(&'static str, SampleIdentifier), String>> { | ||
| pub(crate) fn all<R: Rng + Clone>(templates: &Templates, now: chrono::DateTime<chrono::Utc>, rng: &R) -> anyhow::Result<::std::collections::BTreeMap<(&'static str, SampleIdentifier), String>> { |
There was a problem hiding this comment.
Instead of requiring a cloneable RNG here, we could seed a ChaChaRng out of the rng passed as parameter and clone it within the function. That would be a lot less suspicious :p
There was a problem hiding this comment.
I mean, either one is suspicious, but at least this one is a bit less prolific so sure.
No description provided.