Files
zeekayandHanzo Dev d678c2d213 transform: publish upstream simplex as luxfi/bft on luxfi/log
Reapplies the fork's whole delta on top of the merge, as one reviewable diff
rather than 43 conflict resolutions. Nothing here changes consensus behaviour.

- module github.com/luxfi/bft, Go 1.26.4; root package simplex -> bft, which
  would otherwise collide with the simplex/ subpackage under this module name.
- go.uber.org/zap -> github.com/luxfi/log, so common.Logger takes ...log.Field.
  zap.Error maps to log.Err: log.Error is luxfi/log's package-level logging
  function, not a field constructor, and the compiler is the only thing that
  catches that. Needs luxfi/log v1.6.1 for the Stringers constructor.
- avalanchego/ -> host/. The package holds NodeID, VMBlock, Bitmask and size
  limits — the host node's types, vendored to avoid importing a whole node — so
  it is named for what it holds. Renaming it luxd/ would have swapped one brand
  for another and said just as little.
- testutil/logger.go reimplemented without zapcore. Upstream layers a
  keyword-filtering Core, a debug-swallowing Encoder and a Tee because that is
  zap's extension model; luxfi/log has no such seam, so every policy is applied
  in one emit. That also fixes two defects the layering caused:
  SilenceExceptKeywords rebuilt the logger and discarded an installed Intercept
  hook, and the two sinks read different env vars (LOG_LEVEL vs log_level) so
  they disagreed about suppressing debug. Hooks run outside the lock, so a hook
  that logs fails the test instead of deadlocking it. Fatal panics rather than
  exiting, because exiting takes the suite down with no attribution.
- simplex/monitor_test.go keeps a local logger: testutil imports simplex, so an
  internal test cannot import testutil back without closing a cycle.
- README keeps the name Simplex throughout — it is the protocol's own name and
  the subject of reference [1]. Only the deployment target changed. The HyperSDK
  sentence still says Avalanche because that sentence is true.
- Drops the fork's Makefile: it duplicated unit-tests.sh and fuzz.sh, and ran the
  suite with -timeout 30s when simplex alone needs over 100s.

Co-authored-by: Hanzo Dev <[email protected]>
2026-08-06 04:21:17 -07:00

95 lines
1.6 KiB
Go

// Copyright (C) 2019-2025, Ava Labs, Inc. All rights reserved.
// See the file LICENSE for licensing terms.
package long_running
import (
"context"
"sync"
"testing"
"github.com/luxfi/bft/common"
"github.com/luxfi/bft/testutil"
)
var _ common.BlockBuilder = (*LongRunningBlockBuilder)(nil)
type LongRunningBlockBuilder struct {
t *testing.T
mu sync.Mutex
blockPending bool
changed chan struct{} // signals blockPending state was changed
}
func NewNetworkBlockBuilder(t *testing.T) *LongRunningBlockBuilder {
return &LongRunningBlockBuilder{
t: t,
blockPending: true,
changed: make(chan struct{}, 1),
}
}
func (b *LongRunningBlockBuilder) notifyChanged() {
select {
case b.changed <- struct{}{}:
default:
}
}
func (b *LongRunningBlockBuilder) BuildBlock(
ctx context.Context,
metadata common.ProtocolMetadata,
blacklist common.Blacklist,
) (common.VerifiedBlock, bool) {
for {
b.mu.Lock()
pending := b.blockPending
b.mu.Unlock()
if pending {
return testutil.NewTestBlock(metadata, blacklist), true
}
select {
case <-ctx.Done():
return nil, false
case <-b.changed:
}
}
}
func (b *LongRunningBlockBuilder) WaitForPendingBlock(ctx context.Context) {
for {
b.mu.Lock()
pending := b.blockPending
b.mu.Unlock()
if pending {
return
}
select {
case <-ctx.Done():
return
case <-b.changed:
}
}
}
func (b *LongRunningBlockBuilder) TriggerBlockShouldBeBuilt() {
b.mu.Lock()
defer b.mu.Unlock()
b.blockPending = true
b.notifyChanged()
}
func (b *LongRunningBlockBuilder) TriggerBlockShouldNotBeBuilt() {
b.mu.Lock()
defer b.mu.Unlock()
b.blockPending = false
b.notifyChanged()
}