Move environment-variable checking into fs-mistrust
Compare changes
+ 17
− 26
@@ -165,7 +165,7 @@ pub use cfg::{
@@ -223,18 +223,14 @@ pub async fn run<R: Runtime>(
@@ -403,22 +399,30 @@ pub fn main_main() -> Result<()> {
@@ -428,25 +432,13 @@ pub fn main_main() -> Result<()> {
@@ -486,7 +478,6 @@ pub fn main_main() -> Result<()> {
This recapitulation of the mistrust construction bothered me somewhat. I'm not sure, but I wonder if it woudl be better to move building of
cfg_sources
and the associated manipulation of the mistrust, into arti-client. The function in arti-client would takefs_mistrust_disabled
. It would have to be generic over the relevant traits fromtor_config::load
.What do you think ?
In any case, this would be a followup to this MR. But we might want to do it before the next release to avoid API churn.
IMO this would be fine to do later on. I don't think it would necessarily break existing APIs: it would just be a new way to construct the configuration sources, right?
(Also I have no compunctions about API churn before 1.0.)
I think it would be an api break for the
arti
library crate which we don't care about at all. Filed it as #524. Assigned it to me but feel free to pick it up...