mirrors/arti.git/crates/tor-netdir/src/lib.rs, branch arti-v2.5.0mirror of https://gitlab.torproject.org/tpo/core/arti
http://git.dilluti0n.com/mirrors/arti.git/atom?h=arti-v2.5.02026-06-22T17:00:31ZRemove renaming aliases for MicrodescAndHash2026-06-22T17:00:31ZIan Jacksonijackson@chiark.greenend.org.uk2026-06-16T19:36:08Zurn:sha1:0d697eccea93dc45d781b2f5ece917aa057000f2tor-netdoc: Break Microdesc up into Microdesc and MicrodescAndHash2026-06-22T16:59:20ZIan Jacksonijackson@chiark.greenend.org.uk2026-06-16T19:32:04Zurn:sha1:479acbf246db677d4139bca317cb0e3d1fd9340a
The Microdesc type having a copy of the document hash was always
weird, and it's weirder now that there's all these parse2 types that
don't have it and parse2 derives on Microdesc that don't touch it.
Make a new type for the descriptor and its hash.
Use deref to arrange that the new type works almost like the old one.
Adjust the use lines in the dependencies to temporarily import
MicrodescAndHash as Microdesc.
pick_n_weighted: change variable names for clarity2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-17T21:48:31Zurn:sha1:b6ab42b604f326b86090fa627168b4198c6eeb95
"iter" wasn't an iterator per se. Changed it to "sampled_items", and
renamed other variables to call the items more generically "items" than
"relay", since the mixed usage was a little confusing, and the function
doesn't really know anything about relays.
NetDir::pick_n_relays and pick_n_weighted: clarify/fix docs re zero-weight items2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-16T21:22:30Zurn:sha1:d1a2f905d50b10c849ba39f984f5436b05e37341NetDir::pick_n_weighted: fix doc-comment to not refer to2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-16T21:12:06Zurn:sha1:17bb7b1b7ce80ffcc73a31e3a2a586a722f047daRename NetDir::pick_n_filtered_weighted -> pick_n_weighted2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-16T21:08:19Zurn:sha1:495138dbcee039d0c539bf0b26c4bbc9e269eb17pick_n_filtered_weighted: move warnings logs to caller2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-15T23:34:36Zurn:sha1:06233fce3168706d96874a28b72147e6424fe5cb
We have more useful context in the caller, `pick_n_relays`.
pick_n_filtered_weighted: test and fix for rand 0.10 behavior2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-15T23:03:25Zurn:sha1:00730bb616beddb68d4d40092e0f36d2e1dfbbdc
This tests our implementation of sampling to ensure the behavior is
as-expected. It also updates the implementation so that the tests pass
(It was broken when updating to rand 0.10; see
https://gitlab.torproject.org/tpo/core/arti/-/work_items/1907#note_3426398).
This removes the direct testing of
`rand::IndexedRandom::sample_weighted`, since this is now an
impplementation detail of `pick_n_filtered_weighted`, which is itself
tested.
Verified that this fixes the case of sometimes not being able to select
a Guard in integration tests.
<https://gitlab.torproject.org/tpo/core/chutney/-/work_items/40057#note_3426299>
<https://gitlab.torproject.org/tpo/core/arti/-/work_items/1907#note_3408157>
<https://gitlab.torproject.org/tpo/core/arti/-/work_items/2209>
NetDir::pick_n_relays: move implementation to a testable helper2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-15T22:12:45Zurn:sha1:3e48aa7327ad5336bae0aacc26020c6078498641cargo fmt2026-06-17T21:54:20ZJim Newsomejnewsome@torproject.org2026-06-15T21:59:47Zurn:sha1:5e8bf486b768280947bc02f5095d6333e9f925c8