Skip to content

[ImgBot] Optimize images#3

Merged
TeoSlayer merged 1 commit intomainfrom
imgbot
Feb 17, 2026
Merged

[ImgBot] Optimize images#3
TeoSlayer merged 1 commit intomainfrom
imgbot

Conversation

@imgbot
Copy link
Copy Markdown
Contributor

@imgbot imgbot Bot commented Feb 8, 2026

Beep boop. Your images are optimized!

Your image file size has been reduced 🎉

Details
File Before After Percent reduction
/docs/media/pilot.png 767.18kb 753.48kb 1.79%

📝 docs | :octocat: repo | 🙋🏾 issues | 🏪 marketplace

~Imgbot - Part of Optimole family

/docs/media/pilot.png -- 767.18kb -> 753.48kb (1.79%)

Signed-off-by: ImgBotApp <ImgBotHelp@gmail.com>
@TeoSlayer TeoSlayer self-requested a review February 17, 2026 15:50
Copy link
Copy Markdown
Owner

@TeoSlayer TeoSlayer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just an image

@TeoSlayer TeoSlayer merged commit 09c6b8d into main Feb 17, 2026
@imgbot imgbot Bot deleted the imgbot branch February 17, 2026 17:34
TeoSlayer pushed a commit that referenced this pull request Mar 24, 2026
- Add nameserver SetTTL/Reap + reapExpired test
- Add standalone tasksubmit.Server tests (accept + reject)
- Add CalculateTimeStaged/CalculateTimeCpu edge case tests
- Remove dead code: sendKeyExchangeAuto, sendAuthKeyExchange,
  sendKeyExchange, SetPeerPubKey, closeFull, unregisterAcceptCh
TeoSlayer pushed a commit that referenced this pull request Apr 30, 2026
/docs/media/pilot.png -- 767.18kb -> 753.48kb (1.79%)

Signed-off-by: ImgBotApp <ImgBotHelp@gmail.com>
Co-authored-by: ImgBotApp <ImgBotHelp@gmail.com>
TeoSlayer pushed a commit that referenced this pull request Apr 30, 2026
- Add nameserver SetTTL/Reap + reapExpired test
- Add standalone tasksubmit.Server tests (accept + reject)
- Add CalculateTimeStaged/CalculateTimeCpu edge case tests
- Remove dead code: sendKeyExchangeAuto, sendAuthKeyExchange,
  sendKeyExchange, SetPeerPubKey, closeFull, unregisterAcceptCh
TeoSlayer pushed a commit that referenced this pull request May 3, 2026
writeFrame's flip-to-relay heuristic now requires 3 consecutive
observations of "direct path silent for >8s" before latching onto
the relay path. Single transient ACK gaps from GC pauses, scheduler
jitter, brief packet reordering, or NAT-mapping refresh windows no
longer flap mid-flight.

Symptom this targets: bench bimodality observed in the autoscale
run — same 50 MB transfer ran 4.16 MB/s once, 0.09 MB/s next, 4.27
MB/s after that. The flap was the single-window flip firing on a
transient gap, then "auto-cleared on direct packet receipt" bouncing
back, repeatedly.

Symmetric hysteresis on the relay→direct auto-clear path: clearing
the relay flag now requires 3 consecutive direct-from-peer packets,
so a single direct burst during a relay stretch (one stray packet
from a brief NAT-path opening) doesn't bounce us back to direct only
to flap immediately.

Implementation:
  - blackholeMissCount[nodeID] (int)         — increments per stale-window
                                                writeFrame observation; 0
                                                resets on any direct receipt
  - directClearCount[nodeID] (int)           — increments per direct receipt
                                                while in relay; 0 resets on flip
  - blackholeMissesRequired = 3              — flip threshold
  - directClearsRequired   = 3               — clear threshold
  - clearRelayOnDirectLocked: also resets    — keeps flip counter accurate
    blackholeMissCount on every direct
    packet so writeFrame's window count
    tracks "consecutive misses since last
    confirmed direct," not lifetime misses

Tests:
  - TestWriteFrameFlipsToRelayOnTransientReceiveGap  — flipped:
    single 9s gap MUST NOT flip (was: asserted the flip happened).
  - TestWriteFrameFlipsToRelayAfterSustainedBlackhole — new positive
    case: 3 consecutive blackhole observations DO flip. Pins the
    legitimate use of the heuristic so a future "raise threshold"
    change can't accidentally disable it.
  - TestRelayClearRequiresConsecutiveDirectReceipts  — new test for
    the symmetric direction: single direct receipt does NOT clear
    relay; the Nth one does.
  - TestWriteFrameDoesNotFlipForUnknownPeer          — unchanged pin.

Race detector clean across full pkg/daemon -race suite (63s).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants