Description
HILs with 'static not in the context of buffers:
- adc (Remove 'static from ADC HIL client #3455)
- ble_advertising.rs (hil: remove 'static from ble_advertising #1964)
- crc.rs (CRC HIL: make lifetime 'a, not 'static #1956)
- gpio.rs (hil: gpio: remove 'static #1965)
- i2c.rs (hil: i2c: remove 'static from client #3467)
- radio.rs (hil: remove 'static for client in radio #3459)
- screen (hil: screen: make client 'a #3461)
- sensors.rs (HIL: remove 'static from sensors #1959)
- spi.rs (hil: spi: remove 'static from client #3460)
Early in Tock we used 'static lifetimes fairly liberally. As interfaces matured, we shifted towards generic lifetimes ('a) for the flexibility it (could eventually) affords components. Unfortunately, many of the HILs currently specific static rather than generic lifetimes.
As noted in (2) of #1069, one example of a negative consequence is an artificial restriction of where and how clients can be set.
I believe that the use of static lifetimes is largely a historical artifact, however I wanted to open this as an RFC in case this is something folks feel we need to discuss further before adapting.
Current Status
From a highly scientific and imprecise study, we have more HILs with static lifetimes than generic, but I think that can be attributed to history. Loosely, the newer HILs seem to prefer generic lifetimes:
# Static
$ git grep static | grep -v '//!' | cut -d':' -f1 | sort | uniq -c
9 adc.rs
5 ble_advertising.rs
4 flash.rs
11 i2c.rs
4 led.rs
1 ninedof.rs
5 nonvolatile_storage.rs
12 radio.rs
4 sensors.rs
14 spi.rs
3 symmetric_encryption.rs
6 uart.rs
# Generic
$ git grep "'a" | grep -v '//!' | cut -d':' -f1 | sort | uniq -c
2 flash.rs
10 symmetric_encryption.rs