name paths for purity, preventing double-hash store paths #346

Closed
kiara wants to merge 3 commits from kiara/Fediversity:purge-double-hash into main
Owner
see https://github.com/NixOS/nix/issues/10627
kiara added 1 commit 2025-05-30 16:59:33 +02:00
name paths for purity, preventing double-hash store paths
Some checks failed
/ check-pre-commit (pull_request) Failing after 9s
/ check-peertube (pull_request) Successful in 17s
/ check-panel (pull_request) Failing after 10s
/ check-deployment-basic (pull_request) Failing after 39s
/ check-deployment-cli (pull_request) Failing after 37s
196d4e1540
https://github.com/NixOS/nix/issues/10627

Is this solving some issue we had?

Is this solving some issue we had?
Author
Owner

@fricklerhandwerk this should be noticeable only after #92; the impurity makes it less cacheable as the hash would depend on the root directory name.
so no more urgent than that.

@fricklerhandwerk this should be noticeable only after #92; the impurity makes it less cacheable as the hash would depend on the root directory name. so no more urgent than that.
kiara added 1 commit 2025-06-07 17:06:36 +02:00
restore path for place that needed path over string
Some checks failed
/ check-pre-commit (pull_request) Successful in 13s
/ check-peertube (pull_request) Successful in 18s
/ check-panel (pull_request) Successful in 1m7s
/ check-deployment-basic (pull_request) Successful in 29m39s
/ check-deployment-cli (pull_request) Failing after 29m37s
0930a8b75b
fricklerhandwerk force-pushed purge-double-hash from d4cbc915cd to 0c30914acb 2025-06-11 18:27:55 +02:00 Compare
fricklerhandwerk force-pushed purge-double-hash from 0c30914acb to 43c64debfc 2025-06-11 18:28:16 +02:00 Compare
fricklerhandwerk force-pushed purge-double-hash from 43c64debfc to 5e887094c3 2025-06-11 18:28:48 +02:00 Compare

I really don't know whether the decrease in readability is worth the questionable increase in caching. I see that for shell.nix it would slightly improve the situation if a contributor happened to change the containing directory's name, but otherwise? @Niols Is the double-hashed-storepath actually a thing for the deployment test? Or could the explicit store path name be isolated to just the test code? Would it maybe get solved by moving the test flakes to the VM test rather than re-using the main flake?

I really don't know whether the decrease in readability is worth the questionable increase in caching. I see that for `shell.nix` it would slightly improve the situation if a contributor happened to change the containing directory's name, but otherwise? @Niols Is the double-hashed-storepath actually a thing for the deployment test? Or could the explicit store path name be isolated to just the test code? Would it maybe get solved by moving the test flakes to the VM test rather than re-using the main flake?
Author
Owner

i do suppose even if top-level caching is beaten by this, caching of dependency derivations should be fine still, so i'd be okay with closing if preferred

i do suppose even if top-level caching is beaten by this, caching of dependency derivations should be fine still, so i'd be okay with closing if preferred
fricklerhandwerk closed this pull request 2025-06-11 21:30:03 +02:00
Some checks are pending
/ check-pre-commit (pull_request) Failing after 7s
Required
Details
/ check-peertube (pull_request) Successful in 16s
Required
Details
/ check-deployment-cli (pull_request) Failing after 11m5s
Required
Details
/ check-deployment-basic (pull_request) Failing after 11m7s
Required
Details
/ check-panel (pull_request) Failing after 11m9s
Required
Details
/ check-data-model (pull_request)
Required
/ check-resources (pull_request)
Required
/ check-mastodon (pull_request)
Required
/ check-deployment-panel (pull_request)
Required

Pull request closed

Sign in to join this conversation.
No description provided.