* test/e2e: replace sleeps with event-driven waits in chaos/group/store tests
Replace 21 time.Sleep calls with deterministic waiting using
WaitForOutput, WaitForTCPReady, and a new WaitForTCPUnreachable helper.
Add CountOutput method for snapshot-based incremental log matching.
* test/e2e: validate interval and cap dial/sleep to remaining deadline in WaitForTCPUnreachable
* test/e2e: replace RunProcesses client sleep with log-based proxy readiness detection
Replace the fixed 1500ms sleep in RunProcesses with event-driven proxy
registration detection by monitoring frpc log output for "start proxy
success" messages.
Key changes:
- Add thread-safe SafeBuffer to replace bytes.Buffer in Process, enabling
concurrent read/write of process output during execution
- Add Process.WaitForOutput() to poll process output for pattern matches
with timeout and early exit on process termination
- Add waitForClientProxyReady() that uses config.LoadClientConfig() to
extract proxy names, then waits for each proxy's success log
- For visitor-only clients (no deterministic readiness signal), fall back
to the original sleep with elapsed time deducted
* test/e2e: use shared deadline for proxy readiness and fix doc comment
- Use a single deadline in waitForClientProxyReady so total wait across
all proxies does not exceed the given timeout
- Fix WaitForOutput doc comment to accurately describe single pattern
with count semantics
* test/e2e: optimize RunFrps/RunFrpc with process exit detection
Refactor Process to track subprocess lifecycle via a done channel,
replacing direct cmd.Wait() in Stop() to avoid double-Wait races.
RunFrps/RunFrpc now use select on the done channel instead of fixed
sleeps, allowing short-lived processes (verify, startup failures) to
return immediately while preserving existing timeout behavior for
long-running daemons.
* test/e2e: guard Process against double-Start and Stop-before-Start
Add started flag to prevent double-Start panics and allow Stop to
return immediately when the process was never started. Use sync.Once
for closing the done channel as defense-in-depth against double close.
Replace the fixed 500ms sleep after each frps startup in RunProcesses
with a TCP dial-based readiness check that polls the server bind port.
This reduces the e2e suite wall time from ~97s to ~43s.
Also simplify the RunProcesses API to accept a single server template
string instead of a slice, matching how every call site uses it.
Remove 4 redundant pointer map write-backs in OpenConnection,
CloseConnection, AddTrafficIn, and AddTrafficOut since the map stores
pointers and mutations are already visible without reassignment.
Optimize GetProxiesByTypeAndName from O(n) full map scan to O(1) direct
map lookup by proxy name.
Extract two shared helpers to eliminate duplicated code across STCP,
SUDP, and XTCP visitors:
- dialRawVisitorConn: handles ConnectServer + NewVisitorConn handshake
(auth, sign key, 10s read deadline, error check)
- wrapVisitorConn: handles encryption + pooled compression wrapping,
returning a recycleFn for pool resource cleanup
SUDP is upgraded from WithCompression to WithCompressionFromPool,
aligning with the pooled compression used by STCP and XTCP.
Extract closeProxy() helper to eliminate duplicated 4-step cleanup
sequence (Close, PxyManager.Del, metrics, plugin notify) between
worker() and CloseProxy().
Extract loginUserInfo() helper to eliminate 4 repeated plugin.UserInfo
constructions using LoginMsg fields.
Optimize worker() to snapshot and clear the proxies map under lock,
then perform cleanup outside the lock to reduce lock hold time.
Replace 10 positional parameters in NewControl() with a single
SessionContext struct, matching the client-side pattern. This also
eliminates the post-construction mutation of clientRegistry and
removes two TODO comments.
- pkg/nathole: add RLock when reading clientCfgs map in PreCheck path
to prevent concurrent map read/write crash
- server/proxy: fix error variable shadowing in GetWorkConnFromPool
that could return a closed connection with nil error
- pkg/util/net: check ListenUDP error before spawning goroutines
and assign readConn to struct field so Close() works correctly
- Replace duplicate parseBasicAuth with existing httppkg.ParseBasicAuth
- Extract buildDomains helper in BaseProxy for HTTP/HTTPS/TCPMux proxies
- Extract toProxyStats helper to deduplicate ProxyStats construction
- Extract startVisitorListener helper in BaseProxy for STCP/SUDP proxies
- Extract acceptLoop helper in BaseVisitor for STCP/XTCP visitors
* client/proxy: extract wrapWorkConn to deduplicate UDP/SUDP connection wrapping
Move the repeated rate-limiting, encryption, and compression wrapping
logic from UDPProxy and SUDPProxy into a shared BaseProxy.wrapWorkConn
method, reducing ~18 lines of duplication in each proxy type.
* client/proxy: unify work conn wrapping with pooled compression for all proxy types
Refactor wrapWorkConn to accept encKey parameter and return
(io.ReadWriteCloser, recycleFn, error), enabling HandleTCPWorkConnection
to reuse the same limiter/encryption/compression pipeline.
Switch all proxy types from WithCompression to WithCompressionFromPool.
TCP non-plugin path calls recycleFn via defer after Join; plugin and
UDP/SUDP paths skip recycle (objects are GC'd safely, per golib contract).
Consolidate the separate custom-domain loop and subdomain block into a
single unified loop, matching the pattern already applied to HTTPProxy
in PR #5207. No behavioral change.
The Run() method had two nearly identical loop blocks for registering
custom domains and subdomain, with the same group/non-group registration
logic copy-pasted (~30 lines of duplication).
Consolidate by collecting all domains into a single slice first, then
iterating once with the shared registration logic. Also fixes a minor
inconsistency where the custom domain block used routeConfig.Domain in
CanonicalAddr but the subdomain block used tmpRouteConfig.Domain.
In AddPrefix, the loop `for _, p := range l.prefixes` creates a copy
of each element. Assignments to p.Value and p.Priority only modify
the local copy, not the original slice element, causing updates to
existing prefixes to be silently lost.
This affects client/service.go where AddPrefix is called with
Name:"runID" on reconnection — the old runID value would persist
in log output instead of being updated to the new one.
Fix by using index-based access `l.prefixes[i]` to modify the
original slice element, and add break since prefix names are unique.
- pkg/transport/tls.go: check AppendCertsFromPEM return value and
return clear error when CA file contains no valid PEM certificates
- pkg/plugin/client/http2http.go: set ReadHeaderTimeout to 60s to
match other plugins and prevent slow header attacks
- pkg/plugin/client/http2https.go: same ReadHeaderTimeout fix
- pkg/util/net/websocket.go: store ln parameter in struct to prevent
nil pointer panic when Addr() is called
- pkg/auth/oidc.go: replace unsynchronized []string with map + RWMutex
for subjectsFromLogin to fix data race across concurrent connections
- pkg/nathole/nathole.go: add pool.PutBuf(buf) on ReadFromUDP error
and DecodeMessageInto error paths in waitDetectMessage
- pkg/proto/udp/udp.go: add defer pool.PutBuf(buf) in writerFn to
ensure buffer is returned when the goroutine exits
Fix connection leaks in multiple error paths across client and server:
- server/proxy/http: close tmpConn when WithEncryption fails
- client/proxy: close localConn when ProxyProtocol WriteTo fails
- client/visitor/sudp: close visitorConn on all error paths in getNewVisitorConn
- client/visitor/xtcp: close tunnelConn when WithEncryption fails
- client/visitor/xtcp: close lConn when NewKCPConnFromUDP fails
- pkg/plugin/client/unix_domain_socket: close localConn and connInfo.Conn when WriteTo fails, close connInfo.Conn when DialUnix fails
- pkg/plugin/client/tls2raw: close tlsConn when Handshake or Dial fails
Fix two bugs in TCPGroup.Listen():
- Release acquired port when net.Listen fails to prevent port leak
- Use realPort instead of port for net.Listen to ensure consistency
between port manager records and actual listening port
When frpc verify encounters config errors, the error messages now include
line/column information to help users locate problems:
- TOML syntax errors: report line and column from the TOML parser
(e.g., "toml: line 4, column 11: expected character ]")
- Type mismatch errors: report the field name and approximate line number
(e.g., "line 3: field \"proxies\": cannot unmarshal string into ...")
- File format detection: use file extension to determine format, preventing
silent fallthrough from TOML to YAML parser which produced confusing errors
Previously, a TOML file with syntax errors would silently fall through to the
YAML parser, which would misinterpret the content and produce unhelpful errors
like "json: cannot unmarshal string into Go value of type v1.ClientConfig".
https://claude.ai/code/session_017HWLfcXS3U2hLoy4dsg8Nv
Co-authored-by: Claude <noreply@anthropic.com>
- Fix SSH tunnel gateway incorrectly binding to proxyBindAddr instead of bindAddr
- This caused external connections to fail when proxyBindAddr was set to 127.0.0.1
- SSH tunnel gateway now correctly binds to bindAddr for external accessibility
- Update Release.md with bug fix description
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
- Update go.mod to use github.com/quic-go/quic-go v0.53.0
- Replace quic.Connection interface with *quic.Conn struct
- Replace quic.Stream interface with *quic.Stream struct
- Update all affected files to use new API:
- pkg/util/net/conn.go: Update QuicStreamToNetConn function and wrapQuicStream struct
- server/service.go: Update HandleQUICListener function parameter
- client/visitor/xtcp.go: Update QUICTunnelSession struct field
- client/connector.go: Update defaultConnectorImpl struct field
Fixes#4852
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
fatedierandGitHub
