platform_build_soong/cmd/multiproduct_kati/main.go

526 lines
13 KiB
Go
Raw Normal View History

// Copyright 2017 Google Inc. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package main
import (
"context"
"flag"
"fmt"
"io"
"io/ioutil"
"os"
"path/filepath"
"runtime"
"strings"
"sync"
"syscall"
"time"
"android/soong/finder"
"android/soong/ui/build"
"android/soong/ui/logger"
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
"android/soong/ui/status"
"android/soong/ui/terminal"
"android/soong/ui/tracer"
"android/soong/zip"
)
var numJobs = flag.Int("j", 0, "number of parallel jobs [0=autodetect]")
var keepArtifacts = flag.Bool("keep", false, "keep archives of artifacts")
var incremental = flag.Bool("incremental", false, "run in incremental mode (saving intermediates)")
var outDir = flag.String("out", "", "path to store output directories (defaults to tmpdir under $OUT when empty)")
var alternateResultDir = flag.Bool("dist", false, "write select results to $DIST_DIR (or <out>/dist when empty)")
var onlyConfig = flag.Bool("only-config", false, "Only run product config (not Soong or Kati)")
var onlySoong = flag.Bool("only-soong", false, "Only run product config and Soong (not Kati)")
var buildVariant = flag.String("variant", "eng", "build variant to use")
var shardCount = flag.Int("shard-count", 1, "split the products into multiple shards (to spread the build onto multiple machines, etc)")
var shard = flag.Int("shard", 1, "1-indexed shard to execute")
var skipProducts multipleStringArg
var includeProducts multipleStringArg
func init() {
flag.Var(&skipProducts, "skip-products", "comma-separated list of products to skip (known failures, etc)")
flag.Var(&includeProducts, "products", "comma-separated list of products to build")
}
// multipleStringArg is a flag.Value that takes comma separated lists and converts them to a
// []string. The argument can be passed multiple times to append more values.
type multipleStringArg []string
func (m *multipleStringArg) String() string {
return strings.Join(*m, `, `)
}
func (m *multipleStringArg) Set(s string) error {
*m = append(*m, strings.Split(s, ",")...)
return nil
}
const errorLeadingLines = 20
const errorTrailingLines = 20
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
func errMsgFromLog(filename string) string {
if filename == "" {
return ""
}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
data, err := ioutil.ReadFile(filename)
if err != nil {
return ""
}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
lines := strings.Split(strings.TrimSpace(string(data)), "\n")
if len(lines) > errorLeadingLines+errorTrailingLines+1 {
lines[errorLeadingLines] = fmt.Sprintf("... skipping %d lines ...",
len(lines)-errorLeadingLines-errorTrailingLines)
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
lines = append(lines[:errorLeadingLines+1],
lines[len(lines)-errorTrailingLines:]...)
}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
var buf strings.Builder
for _, line := range lines {
buf.WriteString("> ")
buf.WriteString(line)
buf.WriteString("\n")
}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
return buf.String()
}
// TODO(b/70370883): This tool uses a lot of open files -- over the default
// soft limit of 1024 on some systems. So bump up to the hard limit until I fix
// the algorithm.
func setMaxFiles(log logger.Logger) {
var limits syscall.Rlimit
err := syscall.Getrlimit(syscall.RLIMIT_NOFILE, &limits)
if err != nil {
log.Println("Failed to get file limit:", err)
return
}
log.Verbosef("Current file limits: %d soft, %d hard", limits.Cur, limits.Max)
if limits.Cur == limits.Max {
return
}
limits.Cur = limits.Max
err = syscall.Setrlimit(syscall.RLIMIT_NOFILE, &limits)
if err != nil {
log.Println("Failed to increase file limit:", err)
}
}
func inList(str string, list []string) bool {
for _, other := range list {
if str == other {
return true
}
}
return false
}
func copyFile(from, to string) error {
fromFile, err := os.Open(from)
if err != nil {
return err
}
defer fromFile.Close()
toFile, err := os.Create(to)
if err != nil {
return err
}
defer toFile.Close()
_, err = io.Copy(toFile, fromFile)
return err
}
type mpContext struct {
Context context.Context
Logger logger.Logger
Status status.ToolStatus
Tracer tracer.Tracer
Finder *finder.Finder
Config build.Config
LogsDir string
}
func main() {
stdio := terminal.StdioImpl{}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
output := terminal.NewStatusOutput(stdio.Stdout(), "", false,
build.OsEnvironment().IsEnvTrue("ANDROID_QUIET_BUILD"))
log := logger.New(output)
defer log.Cleanup()
flag.Parse()
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
trace := tracer.New(log)
defer trace.Close()
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
stat := &status.Status{}
defer stat.Finish()
stat.AddOutput(output)
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
var failures failureCount
stat.AddOutput(&failures)
build.SetupSignals(log, cancel, func() {
trace.Close()
log.Cleanup()
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
stat.Finish()
})
buildCtx := build.Context{ContextImpl: &build.ContextImpl{
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
Context: ctx,
Logger: log,
Tracer: trace,
Writer: output,
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
Status: stat,
}}
args := ""
if *alternateResultDir {
args = "dist"
}
config := build.NewConfig(buildCtx, args)
if *outDir == "" {
name := "multiproduct"
if !*incremental {
name += "-" + time.Now().Format("20060102150405")
}
*outDir = filepath.Join(config.OutDir(), name)
// Ensure the empty files exist in the output directory
// containing our output directory too. This is mostly for
// safety, but also triggers the ninja_build file so that our
// build servers know that they can parse the output as if it
// was ninja output.
build.SetupOutDir(buildCtx, config)
if err := os.MkdirAll(*outDir, 0777); err != nil {
log.Fatalf("Failed to create tempdir: %v", err)
}
}
config.Environment().Set("OUT_DIR", *outDir)
log.Println("Output directory:", *outDir)
logsDir := filepath.Join(config.OutDir(), "logs")
os.MkdirAll(logsDir, 0777)
build.SetupOutDir(buildCtx, config)
os.MkdirAll(config.LogsDir(), 0777)
log.SetOutput(filepath.Join(config.LogsDir(), "soong.log"))
trace.SetOutput(filepath.Join(config.LogsDir(), "build.trace"))
var jobs = *numJobs
if jobs < 1 {
jobs = runtime.NumCPU() / 4
ramGb := int(config.TotalRAM() / 1024 / 1024 / 1024)
if ramJobs := ramGb / 25; ramGb > 0 && jobs > ramJobs {
jobs = ramJobs
}
if jobs < 1 {
jobs = 1
}
}
log.Verbosef("Using %d parallel jobs", jobs)
setMaxFiles(log)
finder := build.NewSourceFinder(buildCtx, config)
defer finder.Shutdown()
build.FindSources(buildCtx, config, finder)
vars, err := build.DumpMakeVars(buildCtx, config, nil, []string{"all_named_products"})
if err != nil {
log.Fatal(err)
}
var productsList []string
allProducts := strings.Fields(vars["all_named_products"])
if len(includeProducts) > 0 {
var missingProducts []string
for _, product := range includeProducts {
if inList(product, allProducts) {
productsList = append(productsList, product)
} else {
missingProducts = append(missingProducts, product)
}
}
if len(missingProducts) > 0 {
log.Fatalf("Products don't exist: %s\n", missingProducts)
}
} else {
productsList = allProducts
}
finalProductsList := make([]string, 0, len(productsList))
skipProduct := func(p string) bool {
for _, s := range skipProducts {
if p == s {
return true
}
}
return false
}
for _, product := range productsList {
if !skipProduct(product) {
finalProductsList = append(finalProductsList, product)
} else {
log.Verbose("Skipping: ", product)
}
}
if *shard < 1 {
log.Fatalf("--shard value must be >= 1, not %d\n", *shard)
} else if *shardCount < 1 {
log.Fatalf("--shard-count value must be >= 1, not %d\n", *shardCount)
} else if *shard > *shardCount {
log.Fatalf("--shard (%d) must not be greater than --shard-count (%d)\n", *shard,
*shardCount)
} else if *shardCount > 1 {
finalProductsList = splitList(finalProductsList, *shardCount)[*shard-1]
}
log.Verbose("Got product list: ", finalProductsList)
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
s := buildCtx.Status.StartTool()
s.SetTotalActions(len(finalProductsList))
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
mpCtx := &mpContext{
Context: ctx,
Logger: log,
Status: s,
Tracer: trace,
Finder: finder,
Config: config,
LogsDir: logsDir,
}
products := make(chan string, len(productsList))
go func() {
defer close(products)
for _, product := range finalProductsList {
products <- product
}
}()
var wg sync.WaitGroup
for i := 0; i < jobs; i++ {
wg.Add(1)
go func() {
defer wg.Done()
for {
select {
case product := <-products:
if product == "" {
return
}
buildProduct(mpCtx, product)
}
}
}()
}
wg.Wait()
if *alternateResultDir {
args := zip.ZipArgs{
FileArgs: []zip.FileArg{
{GlobDir: logsDir, SourcePrefixToStrip: logsDir},
},
OutputFilePath: filepath.Join(config.RealDistDir(), "logs.zip"),
NumParallelJobs: runtime.NumCPU(),
CompressionLevel: 5,
}
if err := zip.Zip(args); err != nil {
log.Fatalf("Error zipping logs: %v", err)
}
}
Add a unified status reporting UI This adds a new status package that merges the running of "actions" (ninja calls them edges) of multiple tools into one view of the current state, and gives that to a number of different outputs. For inputs: Kati's output parser has been rewritten (and moved) to map onto the StartAction/FinishAction API. A byproduct of this is that the build servers should be able to extract errors from Kati better, since they look like the errors that Ninja used to write. Ninja is no longer directly connected to the terminal, but its output is read via the protobuf frontend API, so it's just another tool whose output becomes merged together. multiproduct_kati loses its custom status routines, and uses the common one instead. For outputs: The primary output is the ui/terminal.Status type, which along with ui/terminal.Writer now controls everything about the terminal output. Today, this doesn't really change any behaviors, but having all terminal output going through here allows a more complicated (multi-line / full window) status display in the future. The tracer acts as an output of the status package, tracing all the action start / finish events. This replaces reading the .ninja_log file, so it now properly handles multiple output files from a single action. A new rotated log file (out/error.log, or out/dist/logs/error.log) just contains a description of all of the errors that happened during the current build. Another new compressed and rotated log file (out/verbose.log.gz, or out/dist/logs/verbose.log.gz) contains the full verbose (showcommands) log of every execution run by the build. Since this is now written on every build, the showcommands argument is now ignored -- if you want to get the commands run, look at the log file after the build. Test: m Test: <built-in tests> Test: NINJA_ARGS="-t list" m Test: check the build.trace.gz Test: check the new log files Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
2018-05-18 01:37:09 +02:00
s.Finish()
if failures == 1 {
log.Fatal("1 failure")
} else if failures > 1 {
log.Fatalf("%d failures", failures)
} else {
fmt.Fprintln(output, "Success")
}
}
func buildProduct(mpctx *mpContext, product string) {
var stdLog string
outDir := filepath.Join(mpctx.Config.OutDir(), product)
logsDir := filepath.Join(mpctx.LogsDir, product)
if err := os.MkdirAll(outDir, 0777); err != nil {
mpctx.Logger.Fatalf("Error creating out directory: %v", err)
}
if err := os.MkdirAll(logsDir, 0777); err != nil {
mpctx.Logger.Fatalf("Error creating log directory: %v", err)
}
stdLog = filepath.Join(logsDir, "std.log")
f, err := os.Create(stdLog)
if err != nil {
mpctx.Logger.Fatalf("Error creating std.log: %v", err)
}
defer f.Close()
log := logger.New(f)
defer log.Cleanup()
log.SetOutput(filepath.Join(logsDir, "soong.log"))
action := &status.Action{
Description: product,
Outputs: []string{product},
}
mpctx.Status.StartAction(action)
defer logger.Recover(func(err error) {
mpctx.Status.FinishAction(status.ActionResult{
Action: action,
Error: err,
Output: errMsgFromLog(stdLog),
})
})
ctx := build.Context{ContextImpl: &build.ContextImpl{
Context: mpctx.Context,
Logger: log,
Tracer: mpctx.Tracer,
Writer: f,
Thread: mpctx.Tracer.NewThread(product),
Status: &status.Status{},
}}
ctx.Status.AddOutput(terminal.NewStatusOutput(ctx.Writer, "", false,
build.OsEnvironment().IsEnvTrue("ANDROID_QUIET_BUILD")))
args := append([]string(nil), flag.Args()...)
args = append(args, "--skip-soong-tests")
config := build.NewConfig(ctx, args...)
config.Environment().Set("OUT_DIR", outDir)
if !*keepArtifacts {
config.Environment().Set("EMPTY_NINJA_FILE", "true")
}
build.FindSources(ctx, config, mpctx.Finder)
config.Lunch(ctx, product, *buildVariant)
defer func() {
if *keepArtifacts {
args := zip.ZipArgs{
FileArgs: []zip.FileArg{
{
GlobDir: outDir,
SourcePrefixToStrip: outDir,
},
},
OutputFilePath: filepath.Join(mpctx.Config.OutDir(), product+".zip"),
NumParallelJobs: runtime.NumCPU(),
CompressionLevel: 5,
}
if err := zip.Zip(args); err != nil {
log.Fatalf("Error zipping artifacts: %v", err)
}
}
if !*incremental {
os.RemoveAll(outDir)
}
}()
buildWhat := build.BuildProductConfig
if !*onlyConfig {
buildWhat |= build.BuildSoong
if !*onlySoong {
buildWhat |= build.BuildKati
}
}
before := time.Now()
build.Build(ctx, config, buildWhat)
// Save std_full.log if Kati re-read the makefiles
if buildWhat&build.BuildKati != 0 {
if after, err := os.Stat(config.KatiBuildNinjaFile()); err == nil && after.ModTime().After(before) {
err := copyFile(stdLog, filepath.Join(filepath.Dir(stdLog), "std_full.log"))
if err != nil {
log.Fatalf("Error copying log file: %s", err)
}
}
}
mpctx.Status.FinishAction(status.ActionResult{
Action: action,
})
}
type failureCount int
func (f *failureCount) StartAction(action *status.Action, counts status.Counts) {}
func (f *failureCount) FinishAction(result status.ActionResult, counts status.Counts) {
if result.Error != nil {
*f += 1
}
}
func (f *failureCount) Message(level status.MsgLevel, message string) {
if level >= status.ErrorLvl {
*f += 1
}
}
func (f *failureCount) Flush() {}
func (f *failureCount) Write(p []byte) (int, error) {
// discard writes
return len(p), nil
}
func splitList(list []string, shardCount int) (ret [][]string) {
each := len(list) / shardCount
extra := len(list) % shardCount
for i := 0; i < shardCount; i++ {
count := each
if extra > 0 {
count += 1
extra -= 1
}
ret = append(ret, list[:count])
list = list[count:]
}
return
}