aya: use montonic SinceBoot for loaded_at

[ajwerner]

This commit was written to deflake test_loaded_at. The flakiness is due
to the lack of monotonicity in SystemTime clock calculations. See
https://github.com/aya-rs/aya/actions/runs/6340369670/job/17221591723.

This PR addresses that problem by introducing a wrapper type to
encapsulate the monotonically increasing duration since boot,
SinceBoot. This type has a method into_system_time() to convert it
into a SystemTime.

[netlify]

✅ Deploy Preview for aya-rs-docs ready!

Built without sensitive environment variables

Name	Link
🔨 Latest commit	02272cc
🔍 Latest deploy log	https://app.netlify.com/sites/aya-rs-docs/deploys/6515a0e882bc000008f72a42
😎 Deploy Preview	https://deploy-preview-799--aya-rs-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[mergify]

Hey @alessandrod, this pull request changes the Aya Public API and requires your review.

[tamird]

Reviewed 5 of 5 files at r1, 1 of 1 files at r2, all commit messages.
Reviewable status: all files reviewed, 9 unresolved discussions (waiting on @ajwerner and @alessandrod)

---

-- commits line 5 at r1:
"calculations" is wrong - this doesn't have to do with calculations, it has to do with reading wall time more than once.

---

-- commits line 8 at r2:
s/PR/commit/ or just drop the word.

consider doing the same in the previous paragraph - drop the noun

---

aya/src/time.rs line 7 at r2 (raw file):

/// A timestamp relative to the system boot time.
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
pub struct SinceBoot(Duration);

should the nomenclature be something like Instant?

---

aya/src/time.rs line 10 at r2 (raw file):

impl SinceBoot {
    /// The current SinceBoot.

mirror the comment on https://doc.rust-lang.org/stable/std/time/struct.SystemTime.html#method.now?

---

aya/src/time.rs line 12 at r2 (raw file):

    /// The current SinceBoot.
    ///
    /// Note that these calls will be monotonic.

not appropriate to document CLOCK_BOOTTIME, and "these calls" is not the right subject? also, this is used in testing only -- can we perhaps move it there?

---

aya/src/time.rs line 19 at r2 (raw file):

    /// Converts the timestamp to a `SystemTime`.
    ///
    /// Note that this will not be robust to changes in the system clock, and thus these

I don't think we should document the semantics of ST here. it's all discussed at length: https://doc.rust-lang.org/stable/std/time/struct.SystemTime.html#

---

aya/src/time.rs line 26 at r2 (raw file):

    }

    pub(crate) fn from_nanos(nanos: u64) -> Self {

maybe just new and take a duration? no reason to make the reader read "nanos" 3 times.

---

aya/src/time.rs line 34 at r2 (raw file):

    let since_boot = get_time(libc::CLOCK_BOOTTIME);
    let since_epoch = get_time(libc::CLOCK_REALTIME);
    UNIX_EPOCH + since_epoch - since_boot

this has only one caller, and when you combine it with the caller this is just SystemTime::now() + (self - Self::now()), I think?

---

aya/src/time.rs line 55 at r2 (raw file):

    #[test]
    fn test_since_boot_now_into_system_near_system_time() {
        let since_boot = SinceBoot::now().into_system();

let's avoid reading the real clock more than once -- as that is the source of flakes

[tamird]

Reviewable status: all files reviewed, 9 unresolved discussions (waiting on @ajwerner and @alessandrod)

---

aya/src/time.rs line 34 at r2 (raw file):

Previously, tamird (Tamir Duberstein) wrote…

this has only one caller, and when you combine it with the caller this is just SystemTime::now() + (self - Self::now()), I think?

edit: i think this should be SystemTime::now() - (Self::now() - self) to avoid a negative duration.

[ajwerner]

This doesn't seem worth the API churn. Will deflake with retries