Skip to content

0259-gar-784-785-deps-lopdf-randchacha-compat.md

Latest commit

 

History

History
92 lines (68 loc) · 4.15 KB

0259-gar-784-785-deps-lopdf-randchacha-compat.md

File metadata and controls

92 lines (68 loc) · 4.15 KB
Raw

Plan 0259 — GAR-784 + GAR-785: lopdf 0.34→0.40 + rand_chacha 0.9→0.10 compat fixes

Status: Done — merged 2026-06-03 via PR #634 (fa8e3c0) Linear: GAR-784 + GAR-785 Branch: routine/202506031615-deps-lopdf-randchacha-compat Date: 2026-06-03 (America/New_York)

Goal

Fix two blocked Dependabot PRs (#628 + #629) by adapting call sites in garraia-media and garraia-workspace to the new APIs, closing both in a single clean PR.

Root causes

lopdf 0.34 → 0.40 (garraia-media, PR #629, GAR-785)

lopdf 0.40 renamed Object::as_string() (returns Result<&str, _>) to Object::as_str() (returns Result<&[u8], _>). Call sites in pdf.rs used as_string().ok().map(|s| s.to_string()). After the bump, as_str() returns &[u8] so the to_string() call fails to compile.

Fix: replace .and_then(|v| v.as_str().ok()).map(|s| s.to_string()) with .and_then(|v| v.as_str().ok()).map(|b| String::from_utf8_lossy(b).into_owned()) across all 7 metadata fields. from_utf8_lossy handles both UTF-8 and Latin-1 PDF strings gracefully.

rand_chacha 0.9 → 0.10 (garraia-workspace, PR #628, GAR-784)

rand_chacha 0.10 uses rand_core 0.10 while rand 0.9 uses rand_core 0.9. These are incompatible trait versions. The migration_smoke.rs test's unit_vector helper used use rand::{Rng, SeedableRng} (rand_core 0.9) but then called ChaCha8Rng::seed_from_u64(seed) and rng.random_range(...) which require rand_core 0.10 traits. Multiple rand_core versions in the dep graph caused a type mismatch.

Fix:

  • Remove rand = "0.9" dev-dep (was only used in unit_vector).
  • Bump rand_chacha = "0.9" → "0.10".
  • In unit_vector, switch to use rand_chacha::rand_core::{Rng, SeedableRng} and generate floats via rng.next_u64() (no dependency on rand::Rng).

Design invariants

  1. No production code changed — both fixes are in lib/test code only.
  2. from_utf8_lossy ensures metadata extraction never panics on non-UTF-8 PDFs.
  3. unit_vector output changes numerically (different generation method) but the smoke test does not check specific vector values — only that the DB insert/query round-trips correctly.

Out of scope

  • Full lopdf 0.40 API audit (other lopdf methods in pdf.rs).
  • Migrating rand to 0.10 across the workspace.
  • Closing PR #628 / #629 manually (superseded by this PR).

File Structure

File Change
crates/garraia-media/Cargo.toml lopdf = "0.34" → "0.40"
crates/garraia-media/src/pdf.rs as_str().ok().map(to_string) → as_str().ok().map(from_utf8_lossy) (7 sites)
crates/garraia-workspace/Cargo.toml rand = "0.9" removed; rand_chacha = "0.9" → "0.10"
crates/garraia-workspace/tests/migration_smoke.rs use rand::{Rng, SeedableRng} → rand_chacha::rand_core::{Rng, SeedableRng}; random_range → next_u64
plans/0258-gar-783-invite-decline.md Status updated to Done (PR #632 merged)
plans/README.md Rows 0258 updated + 0259 added
ROADMAP.md Plan 0258 / GAR-783 marked ✅

M1 Tasks

  • T1: lopdf version bump + as_str + from_utf8_lossy in pdf.rs
  • T2: rand_chacha bump + rand removal + unit_vector fix in migration_smoke.rs
  • T3: Docs — bookkeeping for plan 0258 (Done) + plan 0259 row

Acceptance criteria

  • cargo check -p garraia-media -p garraia-workspace clean.
  • cargo clippy -p garraia-media -p garraia-workspace --tests --no-deps -- -D warnings clean.
  • CI green on this PR.

Risk register

Risk Likelihood Mitigation
as_str() returns bytes-only (not lossy-decoded) for non-ASCII PDFs Low from_utf8_lossy replaces invalid UTF-8 sequences with � — acceptable for metadata
unit_vector output changes break snapshot tests Low No snapshot assertions on vector values; smoke test checks round-trip, not specific values

Cross-references

  • Dependabot PR #628 (rand_chacha): superseded by this PR
  • Dependabot PR #629 (lopdf): superseded by this PR
  • GAR-784, GAR-785 — the two Linear issues tracking these fixes