cat
42cea1e7c6
internal/pkg: streaming archive reader/writer
...
Test / Create distribution (push) Successful in 1m5s
Test / Sandbox (push) Successful in 2m49s
Test / Hakurei (push) Successful in 3m53s
Test / ShareFS (push) Successful in 3m51s
Test / Sandbox (race detector) (push) Successful in 5m35s
Test / Hakurei (race detector) (push) Successful in 6m32s
Test / Flake checks (push) Successful in 1m15s
This is much more robust and efficient than the simple buffering implementation for larger files. Allocations happen almost exclusively in WalkDir.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-06-03 16:00:36 +09:00
cat
f398f71fa9
internal/pkg: input iterator via IR cache
...
Test / Create distribution (push) Successful in 1m5s
Test / Sandbox (push) Successful in 2m51s
Test / ShareFS (push) Successful in 3m42s
Test / Hakurei (push) Successful in 3m58s
Test / Sandbox (race detector) (push) Successful in 5m31s
Test / Hakurei (race detector) (push) Successful in 6m32s
Test / Flake checks (push) Successful in 1m22s
Primarily useful for garbage collection.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-05-31 14:19:27 +09:00
cat
6e113b8836
internal/pkg: content-based dependency substitution
...
Test / Create distribution (push) Successful in 1m4s
Test / Sandbox (push) Successful in 2m54s
Test / ShareFS (push) Successful in 3m51s
Test / Hakurei (push) Successful in 4m6s
Test / Sandbox (race detector) (push) Successful in 5m31s
Test / Hakurei (race detector) (push) Successful in 6m55s
Test / Flake checks (push) Successful in 1m28s
This change introduces a new fast path for FloodArtifact. It is taken when a curing artifact has identical-by-content controlled relevant inputs and are otherwise identical to an already-cured artifact.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-05-12 00:19:42 +09:00
cat
ceb4d26087
internal/pkg: record cache variant on-disk
...
Test / Create distribution (push) Successful in 1m4s
Test / Sandbox (push) Successful in 2m47s
Test / ShareFS (push) Successful in 3m51s
Test / Sandbox (race detector) (push) Successful in 5m16s
Test / Hakurei (race detector) (push) Successful in 6m23s
Test / Hakurei (push) Successful in 2m39s
Test / Flake checks (push) Successful in 1m25s
This makes custom artifacts much less error-prone to use.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-04-23 00:53:21 +09:00
cat
0ff7ab915b
internal/pkg: move IR primitives out of cache
...
Test / Create distribution (push) Successful in 1m1s
Test / Sandbox (push) Successful in 51s
Test / Hakurei (push) Successful in 2m43s
Test / ShareFS (push) Successful in 2m39s
Test / Sandbox (race detector) (push) Successful in 5m17s
Test / Hakurei (race detector) (push) Successful in 6m22s
Test / Flake checks (push) Successful in 1m24s
These are memory management and caching primitives. Having them as part of Cache is cumbersome and requires a temporary directory that is never used. This change isolates them from Cache to enable independent use.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-04-17 01:02:13 +09:00
cat
485db515f7
internal/pkg/ir: raise string limit to 16 MiB
...
Test / Create distribution (push) Successful in 1m1s
Test / Sandbox (push) Successful in 2m40s
Test / Hakurei (push) Successful in 4m6s
Test / ShareFS (push) Successful in 4m8s
Test / Hpkg (push) Successful in 4m39s
Test / Sandbox (race detector) (push) Successful in 5m9s
Test / Hakurei (race detector) (push) Successful in 5m59s
Test / Flake checks (push) Successful in 1m45s
A string holds "current" hakurei source code. For now the compressed tarball is 4.9 MiB long.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-02-23 15:12:19 +09:00
cat
a40527dcb2
internal/pkg/ir: document reason for avoiding ident cache
...
Test / Create distribution (push) Successful in 34s
Test / Sandbox (push) Successful in 1m49s
Test / ShareFS (push) Successful in 2m55s
Test / Hakurei (push) Successful in 3m7s
Test / Hpkg (push) Successful in 3m36s
Test / Sandbox (race detector) (push) Successful in 5m3s
Test / Hakurei (race detector) (push) Successful in 6m19s
Test / Flake checks (push) Successful in 1m41s
This got brought up earlier today as a potential optimisation. This change documents why it is not viable, and hopefully clears up some performance implications of using IRDecoder, namely that its decoding costs do not amortise.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-02-17 14:38:42 +09:00
cat
ffd2f979fb
internal/pkg: skip duplicate early
...
Test / Create distribution (push) Successful in 1m0s
Test / Sandbox (push) Successful in 2m43s
Test / Hakurei (push) Successful in 3m50s
Test / ShareFS (push) Successful in 4m1s
Test / Hpkg (push) Successful in 4m44s
Test / Sandbox (race detector) (push) Successful in 5m9s
Test / Hakurei (race detector) (push) Successful in 5m55s
Test / Flake checks (push) Successful in 1m44s
This significantly increases IR generation performance.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-02-07 17:11:41 +09:00
cat
e0c720681b
internal/pkg: standardise artifact IR
...
Test / Create distribution (push) Successful in 1m0s
Test / Sandbox (push) Successful in 2m41s
Test / Hakurei (push) Successful in 4m1s
Test / ShareFS (push) Successful in 4m1s
Test / Hpkg (push) Successful in 4m35s
Test / Sandbox (race detector) (push) Successful in 5m4s
Test / Hakurei (race detector) (push) Successful in 6m0s
Test / Flake checks (push) Successful in 1m46s
This should hopefully provide good separation between the artifact curing backend implementation and the (still work in progress) language. Making the IR parseable also guarantees uniqueness of the representation.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2026-02-05 08:24:09 +09:00
cat
