You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Compile request, stream, and daemon tiers in bounded parallel processes, dividing Binaryen's CPU budget among active jobs while preserving configured optimization levels and assertions.
Reuse unchanged server tiers after checking actual dependency contents, failed import lookups, library listings, compiler implementation, options, and output hashes. Keep dependency manifests in .toil/compiler without duplicating compiled artifacts.
Stage and validate tier outputs before publishing, isolate WAT, maps, bindings, and editor declarations, drain workers after failures or cancellation, and serialize complete builds of the same project across processes.
Skip SSR/SSG module evaluation for routes with no matching exports and skip generic email rendering when every template is an auth override. Preserve re-exported and dynamic rendering opt-ins.
Add TOILJS_BUILD_CACHE=0 for forced recompilation and TOILJS_BUILD_JOBS to limit concurrent compiler jobs. Custom transforms and extended compiler configs keep serial, uncached CLI behavior.
On /root/nerva-ui, full recompilation improved from 28.6 seconds to 13.7 seconds and builds with unchanged server inputs took 2.6 seconds. All three Wasm artifacts were identical across full, cached, and forced-recompile runs. Validation includes 796 passing tests, framework build, typecheck, and lint.