syncthing/lib/logger/logger.go

385 lines
9.2 KiB
Go
Raw Normal View History

2015-10-03 17:21:54 +02:00
// Copyright (C) 2014 Jakob Borg. All rights reserved. Use of this source code
// is governed by an MIT-style license that can be found in the LICENSE file.
//go:generate counterfeiter -o mocks/logger.go --fake-name Recorder . Recorder
2015-10-03 17:21:54 +02:00
// Package logger implements a standardized logger with callback functionality
package logger
import (
"fmt"
"io"
2015-10-03 17:21:54 +02:00
"io/ioutil"
"log"
"os"
"strings"
"sync"
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
"time"
2015-10-03 17:21:54 +02:00
)
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
// This package uses stdlib sync as it may be used to debug syncthing/lib/sync
// and that would cause an implosion of the universe.
2015-10-03 17:21:54 +02:00
type LogLevel int
const (
LevelDebug LogLevel = iota
LevelVerbose
LevelInfo
LevelWarn
NumLevels
)
const (
DefaultFlags = log.Ltime
DebugFlags = log.Ltime | log.Ldate | log.Lmicroseconds | log.Lshortfile
)
2015-10-03 17:21:54 +02:00
// A MessageHandler is called with the log level and message text.
type MessageHandler func(l LogLevel, msg string)
type Logger interface {
AddHandler(level LogLevel, h MessageHandler)
SetFlags(flag int)
SetPrefix(prefix string)
Debugln(vals ...interface{})
Debugf(format string, vals ...interface{})
Verboseln(vals ...interface{})
Verbosef(format string, vals ...interface{})
Infoln(vals ...interface{})
Infof(format string, vals ...interface{})
Warnln(vals ...interface{})
Warnf(format string, vals ...interface{})
ShouldDebug(facility string) bool
SetDebug(facility string, enabled bool)
IsTraced(facility string) bool
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
Facilities() map[string]string
FacilityDebugging() []string
NewFacility(facility, description string) Logger
2015-10-03 17:21:54 +02:00
}
type logger struct {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
logger *log.Logger
handlers [NumLevels][]MessageHandler
facilities map[string]string // facility name => description
debug map[string]struct{} // only facility names with debugging enabled
traces string
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
mut sync.Mutex
2015-10-03 17:21:54 +02:00
}
2015-10-31 11:15:49 +01:00
// DefaultLogger logs to standard output with a time prefix.
2015-10-03 17:21:54 +02:00
var DefaultLogger = New()
func New() Logger {
if os.Getenv("LOGGER_DISCARD") != "" {
// Hack to completely disable logging, for example when running
// benchmarks.
return newLogger(ioutil.Discard)
}
return newLogger(controlStripper{os.Stdout})
}
func newLogger(w io.Writer) Logger {
return &logger{
logger: log.New(w, "", DefaultFlags),
traces: os.Getenv("STTRACE"),
facilities: make(map[string]string),
debug: make(map[string]struct{}),
}
2015-10-03 17:21:54 +02:00
}
// AddHandler registers a new MessageHandler to receive messages with the
// specified log level or above.
func (l *logger) AddHandler(level LogLevel, h MessageHandler) {
l.mut.Lock()
defer l.mut.Unlock()
l.handlers[level] = append(l.handlers[level], h)
}
// See log.SetFlags
func (l *logger) SetFlags(flag int) {
l.logger.SetFlags(flag)
}
// See log.SetPrefix
func (l *logger) SetPrefix(prefix string) {
l.logger.SetPrefix(prefix)
}
func (l *logger) callHandlers(level LogLevel, s string) {
for ll := LevelDebug; ll <= level; ll++ {
for _, h := range l.handlers[ll] {
h(level, strings.TrimSpace(s))
}
}
}
// Debugln logs a line with a DEBUG prefix.
func (l *logger) Debugln(vals ...interface{}) {
l.debugln(3, vals...)
}
func (l *logger) debugln(level int, vals ...interface{}) {
s := fmt.Sprintln(vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(level, "DEBUG: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelDebug, s)
}
// Debugf logs a formatted line with a DEBUG prefix.
func (l *logger) Debugf(format string, vals ...interface{}) {
l.debugf(3, format, vals...)
}
func (l *logger) debugf(level int, format string, vals ...interface{}) {
s := fmt.Sprintf(format, vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(level, "DEBUG: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelDebug, s)
}
// Infoln logs a line with a VERBOSE prefix.
func (l *logger) Verboseln(vals ...interface{}) {
s := fmt.Sprintln(vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "VERBOSE: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelVerbose, s)
}
// Infof logs a formatted line with a VERBOSE prefix.
func (l *logger) Verbosef(format string, vals ...interface{}) {
s := fmt.Sprintf(format, vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "VERBOSE: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelVerbose, s)
}
// Infoln logs a line with an INFO prefix.
func (l *logger) Infoln(vals ...interface{}) {
s := fmt.Sprintln(vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "INFO: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelInfo, s)
}
// Infof logs a formatted line with an INFO prefix.
func (l *logger) Infof(format string, vals ...interface{}) {
s := fmt.Sprintf(format, vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "INFO: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelInfo, s)
}
// Warnln logs a formatted line with a WARNING prefix.
func (l *logger) Warnln(vals ...interface{}) {
s := fmt.Sprintln(vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "WARNING: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelWarn, s)
}
// Warnf logs a formatted line with a WARNING prefix.
func (l *logger) Warnf(format string, vals ...interface{}) {
s := fmt.Sprintf(format, vals...)
2015-10-03 17:21:54 +02:00
l.mut.Lock()
defer l.mut.Unlock()
2019-02-02 12:16:27 +01:00
l.logger.Output(2, "WARNING: "+s)
2015-10-03 17:21:54 +02:00
l.callHandlers(LevelWarn, s)
}
// ShouldDebug returns true if the given facility has debugging enabled.
func (l *logger) ShouldDebug(facility string) bool {
l.mut.Lock()
_, res := l.debug[facility]
2015-10-03 17:21:54 +02:00
l.mut.Unlock()
return res
}
// SetDebug enabled or disables debugging for the given facility name.
func (l *logger) SetDebug(facility string, enabled bool) {
l.mut.Lock()
defer l.mut.Unlock()
if _, ok := l.debug[facility]; enabled && !ok {
l.SetFlags(DebugFlags)
l.debug[facility] = struct{}{}
} else if !enabled && ok {
delete(l.debug, facility)
if len(l.debug) == 0 {
l.SetFlags(DefaultFlags)
}
}
2015-10-03 17:21:54 +02:00
}
// IsTraced returns whether the facility name is contained in STTRACE.
func (l *logger) IsTraced(facility string) bool {
return strings.Contains(l.traces, facility) || l.traces == "all"
}
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
// FacilityDebugging returns the set of facilities that have debugging
// enabled.
func (l *logger) FacilityDebugging() []string {
enabled := make([]string, 0, len(l.debug))
2015-10-03 17:21:54 +02:00
l.mut.Lock()
for facility := range l.debug {
enabled = append(enabled, facility)
2015-10-03 17:21:54 +02:00
}
l.mut.Unlock()
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
return enabled
}
// Facilities returns the currently known set of facilities and their
// descriptions.
func (l *logger) Facilities() map[string]string {
l.mut.Lock()
res := make(map[string]string, len(l.facilities))
for facility, descr := range l.facilities {
res[facility] = descr
}
l.mut.Unlock()
return res
2015-10-03 17:21:54 +02:00
}
// NewFacility returns a new logger bound to the named facility.
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
func (l *logger) NewFacility(facility, description string) Logger {
l.SetDebug(facility, l.IsTraced(facility))
2015-10-03 17:21:54 +02:00
l.mut.Lock()
l.facilities[facility] = description
2015-10-03 17:21:54 +02:00
l.mut.Unlock()
return &facilityLogger{
logger: l,
facility: facility,
}
}
// A facilityLogger is a regular logger but bound to a facility name. The
// Debugln and Debugf methods are no-ops unless debugging has been enabled for
// this facility on the parent logger.
type facilityLogger struct {
*logger
facility string
}
// Debugln logs a line with a DEBUG prefix.
func (l *facilityLogger) Debugln(vals ...interface{}) {
if !l.ShouldDebug(l.facility) {
return
}
l.logger.debugln(3, vals...)
2015-10-03 17:21:54 +02:00
}
// Debugf logs a formatted line with a DEBUG prefix.
func (l *facilityLogger) Debugf(format string, vals ...interface{}) {
if !l.ShouldDebug(l.facility) {
return
}
l.logger.debugf(3, format, vals...)
2015-10-03 17:21:54 +02:00
}
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
// A Recorder keeps a size limited record of log events.
type Recorder interface {
Since(t time.Time) []Line
Clear()
}
type recorder struct {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
lines []Line
initial int
mut sync.Mutex
}
// A Line represents a single log entry.
type Line struct {
When time.Time `json:"when"`
Message string `json:"message"`
Level LogLevel `json:"level"`
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
}
func NewRecorder(l Logger, level LogLevel, size, initial int) Recorder {
r := &recorder{
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
lines: make([]Line, 0, size),
initial: initial,
}
l.AddHandler(level, r.append)
return r
}
func (r *recorder) Since(t time.Time) []Line {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
r.mut.Lock()
defer r.mut.Unlock()
res := r.lines
for i := 0; i < len(res); i++ {
if res[i].When.After(t) {
// We must copy the result as r.lines can be mutated as soon as the lock
// is released.
res = res[i:]
cp := make([]Line, len(res))
copy(cp, res)
return cp
}
}
return nil
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
}
func (r *recorder) Clear() {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
r.mut.Lock()
r.lines = r.lines[:0]
r.mut.Unlock()
}
func (r *recorder) append(l LogLevel, msg string) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
line := Line{
When: time.Now(),
Message: msg,
Level: l,
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
}
r.mut.Lock()
defer r.mut.Unlock()
if len(r.lines) == cap(r.lines) {
if r.initial > 0 {
// Shift all lines one step to the left, keeping the "initial" first intact.
copy(r.lines[r.initial+1:], r.lines[r.initial+2:])
} else {
copy(r.lines, r.lines[1:])
}
// Add the new one at the end
r.lines[len(r.lines)-1] = line
return
}
r.lines = append(r.lines, line)
if len(r.lines) == r.initial {
r.lines = append(r.lines, Line{time.Now(), "...", l})
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 17:25:21 +02:00
}
}
// controlStripper is a Writer that replaces control characters
// with spaces.
type controlStripper struct {
io.Writer
}
func (s controlStripper) Write(data []byte) (int, error) {
for i, b := range data {
if b == '\n' || b == '\r' {
// Newlines are OK
continue
}
if b < 32 {
// Characters below 32 are control characters
data[i] = ' '
}
}
return s.Writer.Write(data)
}