@zackees/soldr 0.8.19 → 0.8.20
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +16 -4
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -18,7 +18,7 @@ Third-party comparison (soldr vs Swatinem/rust-cache vs ...): published from `za
|
|
|
18
18
|
|
|
19
19
|
*[performance details](https://zackees.github.io/soldr/)*
|
|
20
20
|
|
|
21
|
-
Cold builds are a wash; warm and cross-worktree
|
|
21
|
+
Cold builds are a wash; clean-target reconstruction from a warm compiler cache and cross-worktree sharing are where soldr's wrapper architecture pays off. The README chart's clean-target row intentionally deletes `target/`; it does not claim Cargo's intact-target freshness fast path. Full historical trend + interactive view: [zackees.github.io/soldr](https://zackees.github.io/soldr/). For the swatinem/rust-cache comparison (GHA target-dir caching, a different layer) see [PERF.md](PERF.md#readme-comparison-bars-issue-785).
|
|
22
22
|
|
|
23
23
|
**Instant tools. Instant builds. One command.**
|
|
24
24
|
|
|
@@ -240,11 +240,23 @@ files touched. What you get over `build-backend = "maturin"`:
|
|
|
240
240
|
- **Compilation caching** — rustc invocations run under soldr's
|
|
241
241
|
`RUSTC_WRAPPER`, so repeat builds hit the cache.
|
|
242
242
|
|
|
243
|
-
Local PEP 517 builds use
|
|
244
|
-
debug information, no LTO, and
|
|
245
|
-
|
|
243
|
+
Local PEP 517 builds use an explicit fast `dev` profile by default:
|
|
244
|
+
`opt-level = 0`, 256 codegen units, line-table debug information, no LTO, and
|
|
245
|
+
incremental compilation. Explicit Maturin/Cargo settings and
|
|
246
|
+
`SOLDR_PEP517_PROFILE` remain authoritative per setting; release pipelines
|
|
246
247
|
should select their release profile explicitly.
|
|
247
248
|
|
|
249
|
+
Successful wheel and editable builds print a one-line cache and timing summary
|
|
250
|
+
to stderr. Set `SOLDR_PEP517_STATS=off` to silence it; `SOLDR_PEP517_STATS=full`
|
|
251
|
+
(and detected verbose `pip`/`uv` frontends) also prints the complete session
|
|
252
|
+
statistics payload.
|
|
253
|
+
|
|
254
|
+
Soldr also caches the last successful wheel for each project/build mode under
|
|
255
|
+
`~/.soldr/pep517/wheels/`. Before packaging it scans source and staged-artifact
|
|
256
|
+
metadata (relative path, size, and modification time); an unchanged tree
|
|
257
|
+
hardlinks the cached wheel into pip's requested output directory and skips
|
|
258
|
+
wheel rebuilding/compression. Set `SOLDR_PEP517_WHEEL_CACHE=off` to opt out.
|
|
259
|
+
|
|
248
260
|
Projects whose packaging backend is not maturin can use soldr as a managed
|
|
249
261
|
wrapper by selecting a delegate in `pyproject.toml`:
|
|
250
262
|
|