#!/bin/sh

#

# Copyright (c) 2009, 2014, Oracle and/or its affiliates. All rights reserved.

# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

#

# This code is free software; you can redistribute it and/or modify it

# under the terms of the GNU General Public License version 2 only, as

# published by the Free Software Foundation.

#

# This code is distributed in the hope that it will be useful, but WITHOUT

# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

# version 2 for more details (a copy is included in the LICENSE file that

# accompanied this code).

#

# You should have received a copy of the GNU General Public License version

# 2 along with this work; if not, write to the Free Software Foundation,

# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

#

# Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

# or visit www.oracle.com if you need additional information or have any

# questions.

#

# Shell script for a fast parallel forest/trees command

usage() {

      echo "usage: $0 [-h|--help] [-q|--quiet] [-v|--verbose] [-s|--sequential] [--] <command> [commands...]" > ${status_output}

      echo "Environment variables which modify behaviour:"

      echo "   HGFOREST_QUIET      : (boolean) If 'true' then standard output is redirected to /dev/null"

      echo "   HGFOREST_VERBOSE    : (boolean) If 'true' then Mercurial asked to produce verbose output"

      echo "   HGFOREST_SEQUENTIAL : (boolean) If 'true' then repos are processed sequentially. Disables concurrency"

      echo "   HGFOREST_GLOBALOPTS : (string, must begin with space) Additional Mercurial global options"

      echo "   HGFOREST_REDIRECT   : (file path) Redirect standard output to specified file"

      echo "   HGFOREST_FIFOS      : (boolean) Default behaviour for FIFO detection. Does not override FIFOs disabled"

      echo "   HGFOREST_CONCURRENCY: (positive integer) Number of repos to process concurrently"

      echo "   HGFOREST_DEBUG      : (boolean) If 'true' then temp files are retained"

      exit 1

}

global_opts="${HGFOREST_GLOBALOPTS:-}"

status_output="${HGFOREST_REDIRECT:-/dev/stdout}"

qflag="${HGFOREST_QUIET:-false}"

vflag="${HGFOREST_VERBOSE:-false}"

sflag="${HGFOREST_SEQUENTIAL:-false}"

while [ $# -gt 0 ]

do

  case $1 in

    -h | --help )

      usage

      ;;

    -q | --quiet )

      qflag="true"

      ;;

    -v | --verbose )

      vflag="true"

      ;;

    -s | --sequential )

      sflag="true"

      ;;

    '--' ) # no more options

      shift; break

      ;;

    -*)  # bad option

      usage

      ;;

     * )  # non option

      break

      ;;

  esac

  shift

done

# debug mode

if [ "${HGFOREST_DEBUG:-false}" = "true" ] ; then

  global_opts="${global_opts} --debug"

fi

# silence standard output?

if [ ${qflag} = "true" ] ; then

  global_opts="${global_opts} -q"

  status_output="/dev/null"

fi

# verbose output?

if [ ${vflag} = "true" ] ; then

  global_opts="${global_opts} -v"

fi

# Make sure we have a command.

if [ ${#} -lt 1 -o -z "${1:-}" ] ; then

  echo "ERROR: No command to hg supplied!" > ${status_output}

  usage > ${status_output}

fi

# grab command

command="${1}"; shift

if [ ${vflag} = "true" ] ; then

  echo "# Mercurial command: ${command}" > ${status_output}

fi

# At this point all command options and args are in "$@".

# Always use "$@" (within double quotes) to avoid breaking

# args with spaces into separate args.

if [ ${vflag} = "true" ] ; then

  echo "# Mercurial command argument count: $#" > ${status_output}

  for cmdarg in "$@" ; do

    echo "# Mercurial command argument: ${cmdarg}" > ${status_output}

  done

fi

# Clean out the temporary directory that stores the pid files.

tmp=/tmp/forest.$$

rm -f -r ${tmp}

mkdir -p ${tmp}

if [ "${HGFOREST_DEBUG:-false}" = "true" ] ; then

  # ignores redirection.

  echo "DEBUG: temp files are in: ${tmp}" >&2

fi

# Check if we can use fifos for monitoring sub-process completion.

echo "1" > ${tmp}/read

while_subshell=1

while read line; do

  while_subshell=0

  break;

done < ${tmp}/read

rm ${tmp}/read

on_windows=`uname -s | egrep -ic -e 'cygwin|msys'`

if [ ${while_subshell} = "1" -o ${on_windows} = "1" ]; then

  # cygwin has (2014-04-18) broken (single writer only) FIFOs

  # msys has (2014-04-18) no FIFOs.

  # older shells create a sub-shell for redirect to while

  have_fifos="false"

else

  have_fifos="${HGFOREST_FIFOS:-true}"

fi

safe_interrupt () {

  if [ -d ${tmp} ]; then

    if [ "`ls ${tmp}/*.pid`" != "" ]; then

      echo "Waiting for processes ( `cat ${tmp}/.*.pid ${tmp}/*.pid 2> /dev/null | tr '\n' ' '`) to terminate nicely!" > ${status_output}

      sleep 1

      # Pipe stderr to dev/null to silence kill, that complains when trying to kill

      # a subprocess that has already exited.

      kill -TERM `cat ${tmp}/*.pid | tr '\n' ' '` 2> /dev/null

      wait

      echo "Interrupt complete!" > ${status_output}

    fi

    rm -f -r ${tmp}

  fi

  exit 130

}

nice_exit () {

  if [ -d ${tmp} ]; then

    if [ "`ls -A ${tmp} 2> /dev/null`" != "" ]; then

      wait

    fi

    if [ "${HGFOREST_DEBUG:-false}" != "true" ] ; then

      rm -f -r ${tmp}

    fi

  fi

}

trap 'safe_interrupt' INT QUIT

trap 'nice_exit' EXIT

subrepos="corba jaxp jaxws langtools jdk hotspot nashorn"

# Only look in specific locations for possible forests (avoids long searches)

pull_default=""

repos=""

repos_extra=""

if [ "${command}" = "clone" -o "${command}" = "fclone" -o "${command}" = "tclone" ] ; then

  # we must be a clone

  if [ ! -f .hg/hgrc ] ; then

    echo "ERROR: Need initial repository to use this script" > ${status_output}

    exit 1

  fi

  # the clone must know where it came from (have a default pull path).

  pull_default=`hg paths default`

  if [ "${pull_default}" = "" ] ; then

    echo "ERROR: Need initial clone with 'hg paths default' defined" > ${status_output}

    exit 1

  fi

  # determine which sub repos need to be cloned.

  for i in ${subrepos} ; do

    if [ ! -f ${i}/.hg/hgrc ] ; then

      repos="${repos} ${i}"

    fi

  done

  pull_default_tail=`echo ${pull_default} | sed -e 's@^.*://[^/]*/\(.*\)@\1@'`

  if [ $# -gt 0 ] ; then

    # if there is an "extra sources" path then reparent "extra" repos to that path

    if [ "x${pull_default}" = "x${pull_default_tail}" ] ; then

      echo "ERROR: Need initial clone from non-local source" > ${status_output}

      exit 1

    fi

    # assume that "extra sources" path is the first arg

    pull_extra="${1}/${pull_default_tail}"

    # determine which extra subrepos need to be cloned.

    for i in ${subrepos_extra} ; do

      if [ ! -f ${i}/.hg/hgrc ] ; then

        repos_extra="${repos_extra} ${i}"

      fi

    done

  else

    if [ "x${pull_default}" = "x${pull_default_tail}" ] ; then

      # local source repo. Clone the "extra" subrepos that exist there.

      for i in ${subrepos_extra} ; do

        if [ -f ${pull_default}/${i}/.hg/hgrc -a ! -f ${i}/.hg/hgrc ] ; then

          # sub-repo there in source but not here

          repos_extra="${repos_extra} ${i}"

        fi

      done

    fi

  fi

  # Any repos to deal with?

  if [ "${repos}" = "" -a "${repos_extra}" = "" ] ; then

    echo "No repositories to process." > ${status_output}

    exit

  fi

  # Repos to process concurrently. Clone does better with low concurrency.

  at_a_time="${HGFOREST_CONCURRENCY:-2}"

else

  # Process command for all of the present repos

  for i in . ${subrepos} ${subrepos_extra} ; do

    if [ -d ${i}/.hg ] ; then

      repos="${repos} ${i}"

    fi

  done

  # Any repos to deal with?

  if [ "${repos}" = "" ] ; then

    echo "No repositories to process." > ${status_output}

    exit

  fi

  # any of the repos locked?

  locked=""

  for i in ${repos} ; do

    if [ -h ${i}/.hg/store/lock -o -f ${i}/.hg/store/lock ] ; then

      locked="${i} ${locked}"

    fi

  done

  if [ "${locked}" != "" ] ; then

    echo "ERROR: These repositories are locked: ${locked}" > ${status_output}

    exit 1

  fi

  # Repos to process concurrently.

  at_a_time="${HGFOREST_CONCURRENCY:-8}"

fi

# Echo out what repositories we do a command on.

echo "# Repositories: ${repos} ${repos_extra}" > ${status_output}

if [ "${command}" = "serve" ] ; then

  # "serve" is run for all the repos as one command.

  (

    (

      cwd=`pwd`

      serving=`basename ${cwd}`

      (

        echo "[web]"

        echo "description = ${serving}"

        echo "allow_push = *"

        echo "push_ssl = False"

        echo "[paths]"

        for i in ${repos} ; do

          if [ "${i}" != "." ] ; then

            echo "/${serving}/${i} = ${i}"

          else

            echo "/${serving} = ${cwd}"

          fi

        done

      ) > ${tmp}/serve.web-conf

      echo "serving root repo ${serving}" > ${status_output}

      echo "hg${global_opts} serve" > ${status_output}

      (PYTHONUNBUFFERED=true hg${global_opts} serve -A ${status_output} -E ${status_output} --pid-file ${tmp}/serve.pid --web-conf ${tmp}/serve.web-conf; echo "$?" > ${tmp}/serve.pid.rc ) 2>&1 &

    ) 2>&1 | sed -e "s@^@serve:   @" > ${status_output}

  ) &

else

  # Run the supplied command on all repos in parallel.

  # n is the number of subprocess started or which might still be running.

  n=0

  if [ ${have_fifos} = "true" ]; then

    # if we have fifos use them to detect command completion.

    mkfifo ${tmp}/fifo

    exec 3<>${tmp}/fifo

  fi

  # iterate over all of the subrepos.

  for i in ${repos} ${repos_extra} ; do

    n=`expr ${n} '+' 1`

    repopidfile=`echo ${i} | sed -e 's@./@@' -e 's@/@_@g'`

    reponame=`echo ${i} | sed -e :a -e 's/^.\{1,20\}$/ &/;ta'`

    pull_base="${pull_default}"

    # regular repo or "extra" repo?

    for j in ${repos_extra} ; do

      if [ "${i}" = "${j}" ] ; then

        # it's an "extra"

        pull_base="${pull_extra}"

      fi

    done

    # remove trailing slash

    pull_base="`echo ${pull_base} | sed -e 's@[/]*$@@'`"

    # execute the command on the subrepo

    (

      (

        if [ "${command}" = "clone" -o "${command}" = "fclone" -o "${command}" = "tclone" ] ; then

          # some form of clone

          clone_newrepo="${pull_base}/${i}"

          parent_path="`dirname ${i}`"

          if [ "${parent_path}" != "." ] ; then

            times=0

            while [ ! -d "${parent_path}" ] ; do  ## nested repo, ensure containing dir exists

              if [ "${sflag}" = "true" ] ; then

                # Missing parent is fatal during sequential operation.

                echo "ERROR: Missing parent path: ${parent_path}" > ${status_output}

                exit 1

              fi

              times=`expr ${times} '+' 1`

              if [ `expr ${times} '%' 10` -eq 0 ] ; then

                echo "${parent_path} still not created, waiting..." > ${status_output}

              fi

              sleep 5

            done

          fi

          # run the clone command.

          echo "hg${global_opts} clone ${clone_newrepo} ${i}" > ${status_output}

          (PYTHONUNBUFFERED=true hg${global_opts} clone ${clone_newrepo} ${i}; echo "$?" > ${tmp}/${repopidfile}.pid.rc ) 2>&1 &

        else

          # run the command.

          echo "cd ${i} && hg${global_opts} ${command} ${@}" > ${status_output}

          cd ${i} && (PYTHONUNBUFFERED=true hg${global_opts} ${command} "${@}"; echo "$?" > ${tmp}/${repopidfile}.pid.rc ) 2>&1 &

        fi

        echo $! > ${tmp}/${repopidfile}.pid

      ) 2>&1 | sed -e "s@^@${reponame}:   @" > ${status_output}

      # tell the fifo waiter that this subprocess is done.

      if [ ${have_fifos} = "true" ]; then

        echo "${i}" >&3

      fi

    ) &

    if [ "${sflag}" = "true" ] ; then

      # complete this task before starting another.

      wait

    else

      if [ "${have_fifos}" = "true" ]; then

        # check on count of running subprocesses and possibly wait for completion

        if [ ${n} -ge ${at_a_time} ] ; then

          # read will block until there are completed subprocesses

          while read repo_done; do

            n=`expr ${n} '-' 1`

            if [ ${n} -lt ${at_a_time} ] ; then

              # we should start more subprocesses

              break;

            fi

          done <&3

        fi

      else

        # Compare completions to starts

        completed="`(ls -a1 ${tmp}/*.pid.rc 2> /dev/null | wc -l) || echo 0`"

        while [ `expr ${n} '-' ${completed}` -ge ${at_a_time} ] ; do

          # sleep a short time to give time for something to complete

          sleep 1

          completed="`(ls -a1 ${tmp}/*.pid.rc 2> /dev/null | wc -l) || echo 0`"

        done

      fi

    fi

  done

  if [ ${have_fifos} = "true" ]; then

    # done with the fifo

    exec 3>&-

  fi

fi

# Wait for all subprocesses to complete

wait

# Terminate with exit 0 only if all subprocesses were successful

# Terminate with highest exit code of subprocesses

ec=0

if [ -d ${tmp} ]; then

  rcfiles="`(ls -a ${tmp}/*.pid.rc 2> /dev/null) || echo ''`"

  for rc in ${rcfiles} ; do

    exit_code=`cat ${rc} | tr -d ' \n\r'`

    if [ "${exit_code}" != "0" ] ; then

      if [ ${exit_code} -gt 1 ]; then

        # mercurial exit codes greater than "1" signal errors.

      repo="`echo ${rc} | sed -e 's@^'${tmp}'@@' -e 's@/*\([^/]*\)\.pid\.rc$@\1@' -e 's@_@/@g'`"

      echo "WARNING: ${repo} exited abnormally (${exit_code})" > ${status_output}

      fi

      if [ ${exit_code} -gt ${ec} ]; then

        # assume that larger exit codes are more significant

        ec=${exit_code}

      fi

    fi

  done

fi

exit ${ec}