<!doctype linuxdoc system><!-- fill-column: 78; -->
<linuxdoc>
<book>
<titlepag>
<title>distcc User Manual</title>
<author>Martin Pool</author>
<date>$Date: 2003/04/05 00:46:27 $, for distcc 2.1cvs</date>
</titlepag>
<toc>
<chapt>
<heading>Introduction</heading>
<p>
<em>"Speed, it seems to me, provides the one genuinely modern pleasure."</em>
--- Aldous Huxley
</p>
<sect>
<heading>Overview</heading>
<p>
<htmlurl url="http://distcc.samba.org/" name="distcc"> is a program
to distribute compilation of C or C++ code across several machines on a
network. distcc should always generate the same results as a
local compile, is simple to install and use, and is often
significantly faster than a local compile.
</p>
<p>
Unlike other distributed build systems, distcc does not
require all machines to share a filesystem, have
synchronized clocks, or to have the same libraries or header
files installed.
</p>
<p>
Compilation is centrally controlled by a client machine, which
is typically the developer's workstation or laptop. The
distcc client runs on this machine, as does <em>make</em>, the
preprocessor, the linker, and other stages of the build
process. Any number of "volunteer" machines help the client
to build the program, by running the compiler and assembler
as required. The volunteer machines run the <tt>distccd</tt>
daemon which listens on a network socket for requests.
</p>
<p>
distcc sends the complete preprocessed source code across the
network for each job, so all it requires of the volunteer
machines is that they be running the <tt>distccd</tt> daemon,
and that they have an appropriate compiler installed.
</p>
<p>
distcc is designed to be used with GNU make's parallel-build
feature (<tt>-j</tt>). Shipping files across the network takes
time, but few cycles on the client machine. Any files that can
be built remotely are essentially "for free" in terms of client
CPU.
</p>
<p>
distcc was written by Martin Pool.
</p>
<p>
distcc was inspired by Andrew Tridgell's <htmlurl
url="http://ccache.samba.org/" name="ccache"> program.
</p>
<p>
If you find distcc useful, please try to complete the survey
form in the distribution, or just send an email.
</p>
</sect>
<sect>
<heading>Licence</heading>
<p>
distcc and the <em>distcc User Manual</em> are copyright (C)
2002, 2003 by Martin Pool.
</p>
<p>
distcc is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
</p>
<p>
Permission is granted to copy, distribute and/or modify the
<em>distcc User Manual</em> under the terms of the GNU Free
Documentation License, Version 1.1 or any later version
published by the Free Software Foundation; with no Invariant
Sections, no Front-Cover Texts, and no Back-Cover Texts.
</p>
<p>
distcc 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 for more details.
</p>
<p>
You should have received a copy of the GNU General Public
License and GNU Free Documentation License along with
distcc. If not, write to the Free Software Foundation,
Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA,
or see <url url="http://www.gnu.org/licenses/">.
</p>
<p>
The author understands the GNU GPL to apply to distcc in the
following way: you are allowed to use distcc to compile a
non-free program, or to call it from a non-free Make, or to
call a non-free compiler. However, you may not distribute a
modified version of distcc unless you comply with the terms
of the GPL: in particular, giving your users access to the
source code and the right to redistribute it, and clearly
identifying your changes.
</p>
</sect>
<sect>
<heading>Security Considerations</heading>
<p>
<bf>
distcc should only be used on networks where all machines
and all users are trusted.
</bf>
</p>
<p>
The distcc daemon, <tt>distccd</tt>, allows other machines
on the network to run arbitrary commands on the volunteer
machine. Anyone that can make a connection to the volunteer
machine can run essentially any command as the user running
<tt>distccd</tt>.
</p>
<p>
distcc is suitable for use on a small to medium network of
friendly developers. It's certainly not suitable for use on
a machine connected to the Internet or a large
(e.g. university campus) network without firewalling in
place.
</p>
<p>
<tt>inetd</tt> or <tt>tcpwrappers</tt> can be used to impose
access control rules, but this should be done with an eye to
the possibility of address spoofing.
</p>
<p>
In summary, the security level is similar to that of
old-style network protocols like X11-over-TCP, NFS or RSH.
</p>
</sect>
<sect>
<heading>Getting Started</heading>
<p>
Four straightforward steps are required to install and use
distcc:
<enum>
<item>
Compile and install the <tt>distcc</tt> package on the
client and volunteer machines.
<item>
Start the <tt>distccd</tt> daemon on all volunteer
machines.
<item>
On the client, set the <tt>DISTCC_HOSTS</tt> environment
variable to indicate which volunteer machines to use.
For example:
<tscreen><verb>DISTCC_HOSTS='angry toey:4202 localhost'</verb></tscreen>
<item>
Set the <tt>CC</tt> variable or edit Makefiles to prefix
distcc to calls to the C/C++ compiler. For example:
<tscreen><verb>distcc gcc -o hello.o -c hello.c</verb></tscreen>
</enum>
</p>
</sect>
<sect>
<heading>Reporting Bugs</heading>
<p>
If you think you have found a bug, please check the manual
and the <tt>TODO</tt> file to see if it is a known
restriction. If not, please send a clear and detailed
report to the mailing list <tt>distcc@lists.samba.org</tt>.
(For a clear
and detailed description of "clear and detailed", see Simon
Tatham's advice on reporting bugs, <url
url="http://www.chiark.greenend.org.uk/~sgtatham/bugs.html">.)
</p>
<p>
A good bug report for distcc should include:
<enum>
<item>
What you're trying to do. For example: "compile KDE",
"use gcc's <tt>-MD</tt> option".
</item>
<item>
What actually happens. For example: "distcc fails with
error 104", "the compilation never completes", "I get
error message XXX".
</item>
<item>
The version of distcc you're using (the output of
<tt>-</tt><tt>-version</tt> on both client and server.
If you got it from a distribution rather than building
it yourself, then mention that.
</item>
<item>
What other software you're using, in particular the
operating system and compiler. For the operating system
it's normally enough to give the overall version
(FreeBSD CURRENT, RedHat Linux 7.2, ...). For the
compiler, use "<tt>gcc -version</tt>".
</item>
<item>
The exact command you're using to run the compilation.
If you're using make, then include the line from its
output that runs the compiler.
</item>
<item>
The debug logs from the client and server. On the
client, you should set <tt>DISTCC_VERBOSE</tt> and
<tt>DISTCC_LOG</tt>. On the server, use
<tt>-</tt><tt>-verbose</tt> and
<tt>-</tt><tt>-log-file</tt>. If you can, trim the log
files to just the invocation that causes trouble.
Grepping for a process id can help with this. If the
problem is intermittent, then please leave logging
running until it recurs and then pull out a smaller
section of logs to send.
</item>
</enum>
</p>
<p>
Please do not obfuscate your logs. The name of a single
source file or machine is probably not confidential
information, but the confusion introduced by editing logs
can be significant.
</p>
<p>
Please send a problem description to the <tt>distcc</tt>
mailing list, on <tt>lists.samba.org</tt>. Please don't
send mail direct to the author: if you use the list, other
people may be able to help you, and the answers are publicly
archived.
</p>
</sect>
<sect>
<heading>
Test Suite
</heading>
<p>
distcc has a test suite written in Python using the
<em>ComfyChair</em> framework. It does not yet exercise all
functionality, but is improving.
</p>
<p>
To run the test suite, run <tt>make check</tt> from the distcc
source directoy.
</p>
</sect>
</chapt> <!-- End introduction chapter -->
<chapt>
<heading>Using distcc</heading>
<sect>
<heading>Invoking distcc</heading>
<p>
To setup distcc to be compatible with the widest range of
existing software, create a "masquerade dir" of compiler
links that will invoke distcc. When the distcc-in-disguise
gets invoked, it invokes the real compiler of the same name
either on the local client machine, or on a remote volunteer
host.
</p>
<p>
For instance, you could create the directory named
/usr/lib/distcc/bin and populate it links (symlinks are
slightly easier to maintain in the long run, but use hard
links if you prefer them):
<tscreen><verb>$ mkdir /usr/lib/distcc/bin
$ cd /usr/lib/distcc/bin
$ ln -s ../../../bin/distcc gcc
$ ln -s ../../../bin/distcc cc
$ ln -s ../../../bin/distcc g++
$ ln -s ../../../bin/distcc c++</verb></tscreen>
</p>
<p>
Then, to use distcc, a user just needs to put the directory
/usr/lib/distcc/bin early in the PATH (and have set the
DISTCC_HOSTS environment variable) and distcc will handle the
rest. Note that this masquerade dir must occur on the PATH
earlier than the directory that contains the actual compilers
of the same names, and that any auxiliary programs that these
compilers call (such as "as" or "ld") must also be found on the
PATH in a dir after the masquerade dir (since distcc calls out
to the real compiler with a PATH value that has all dirs up to
and including the masquerade dir trimmed off).
</p>
<p>
An alternate setup is to prefix the distcc command to compiler
command lines so that it is called explicitly. This allows you
to more easily control which things use distcc and which things
don't, but can be more problematical when trying to use distcc
with existing projects.
</p>
<p>
For example, to compile the standard application program:
<tscreen><verb>distcc gcc -o hello.o -c hello.c</verb></tscreen>
</p>
<p>
Standard Makefiles, including those using the GNU
autoconf/automake system use the <bf>$CC</bf> variable as
the name of the C compiler to run and the <bf>$CXX</bf> variable
as the name of the C++ compiler to run. In many cases, it is
sufficient to just override one or both of these variables,
either from the command line, or perhaps from your login script
(if you wish to use distcc for all compilations). The following
example sets both variables and takes advantage of the fact that
distcc defaults to calling "gcc" if no other compiler name is
provided:
<tscreen><verb>make CC=distcc CXX='distcc g++'</verb></tscreen>
</p>
<p>
Unfortunately, this setup sometimes leads to incompatibilities
with packages that don't expect the compiler name to contain
spaces or to projects that don't honor the above variables when
deciding what compiler to use. For instance, the KDE package
will fail to compile using the above CXX-munging idiom, but will
compile just fine if you use a masquerade dir (which causes all
executions of "g++" to really run distcc).
</p>
</sect>
<sect>
<heading>Options</heading>
<p>
Options to distcc must precede the compiler name. Any
arguments or options following the name of the compiler are
passed through to the compiler.
</p>
<p>
<descrip>
<tag><tt>-</tt><tt>-help</tt></tag>
<p>
Print a detailed usage message and exit.
</p>
<tag><tt>-</tt><tt>-version</tt></tag>
<p>
Show distcc version and exit.
</p>
</descrip>
</p>
</sect>
<sect>
<heading>Environment Variables</heading>
<p>
The way in which distcc runs the compiler is controlled by a
few environment variables.
</p>
<p>
<bf>NOTE:</bf> Some versions of make do not export Make
variables as environment variables by default. Also,
assignments to variables within the Makefile may override
their definitions in the environment that calls make. The
most reliable method seems to be to set <tt>DISTCC_*</tt>
variables in the environment of Make, and to set <tt>CC</tt>
on the right-hand-side of the Make command line. For
example:
<tscreen><verb>$ DISTCC_HOSTS='localhost wistful toey'
$ PATH="/usr/lib/distcc/bin:$PATH"
$ export DISTCC_HOSTS PATH
$ ./configure
$ make all</verb></tscreen>
<p>
or:
<tscreen><verb>$ DISTCC_HOSTS='localhost wistful toey'
$ export DISTCC_HOSTS
$ CC='distcc' ./configure
$ make CC='distcc' all</verb></tscreen>
</p>
<p>
Some Makefiles may, contrary to convention, explicitly call
<tt>gcc</tt> or some other compiler, in which case
overriding <tt>$CC</tt> will not be enough to call distcc.
While this is harmless (but suboptimal), using a masquerade
dir of distcc links will avoid this.
</p>
<p>
Remember that you should not use both methods for calling
distcc at the same time. If you are using a masquerade dir,
don't munge CC and/or CXX (just put the dir early on your
PATH). If you're not using a masquerade dir, you'll need to
either change CC and/or CXX, or modify the Makefile(s) to
call distcc explicitly.
</p>
<p>
<descrip>
<tag><tt>DISTCC_HOSTS</tt></tag>
<p>
Space-separated list of volunteer host specifications.
</p>
<tag>
<tt>DISTCC_VERBOSE</tt>
</tag>
<p>
If set to <tt>1</tt>, distcc produces explanatory messages on the
standard error stream. This can be helpful in debugging
problems. Bug reports should include verbose output.
</p>
<tag>
<tt>DISTCC_LOG</tt>
<p>
Log file to receive messages from distcc itself, rather
than stderr.
</p>
<tag><tt>DISTCC_SAVE_TEMPS</tt>
<p>
If set to <tt>1</tt>, temporary files are not deleted
after use. Good for debugging, or if your disks are too
empty.
</p>
<tag><tt>DISTCC_TCP_CORK</tt>
<p>
If set to <tt>0</tt>,
disable use of "TCP corks", even if they're present on
this system. Using corks normally helps pack requests into
fewer packets and aids performance.
</p>
</descrip>
</p>
</sect>
<sect>
<heading>Which Jobs are Distributed?</heading>
<p>
Building a C or C++ program on Unix involves several phases:
<itemize>
<item>
Preprocessing source (<tt/.c/) and headers (<tt/.h/) to
a preprocessed file (<tt/.i/)
</item>
<item>
Compiling preprocessed source (<tt/.i/) to assembly
instructions (<tt/.s/)
</item>
<item>
Assembling to an object file (<tt/.o/)
</item>
<item>
Linking object files and libraries to form an
executable, library, or shared library.
</item>
</itemize>
</p>
<p>
distcc only ever runs the compiler and assembler remotely.
The preprocessor must always run locally because it needs to
access various header files on the local machine which may
not be present, or may not be the same, on the volunteer.
The linker similarly needs to examine libraries and object
files, and so must run locally.
</p>
<p>
The compiler and assembler take only a single input file,
the preprocessed source, produce a single output, the object
file. distcc ships these two files across the network and
can therefore run the compiler/assembler remotely.
</p>
<p>
Fortunately, for most programs running the preprocessor is
relatively cheap, and the linker is called relatively
infrequent, so most of the work can be distributed.
</p>
<p>
distcc examines its command line to determine which of these
phases are being invoked, and whether the job can be
distributed. Here is an example of a typical command that
can be preprocessed locally and compiled remotely:
<tscreen><verb>distcc gcc -o hello.o -DGREETING="hello" -c hello.c</verb></tscreen>
</p>
<p>
The command-line scanner is intended to behave in the same
way as gcc. In case of doubt, distcc runs the job locally.
</p>
<p>
In particular, this means that commands that compile and
link in one go cannot be distributed. These are quite rare
in realistic projects. Here is one example of a command
that could not be distributed, because it calls the compiler
and linker
<tscreen><verb>distcc gcc -o hello hello.c</verb></tscreen>
</p>
</sect>
<sect>
<heading>Running Jobs in Parallel</heading>
<p>
Moving source across the network is less efficient to
compiling it locally. If you have access to a machine much
faster than your workstation, the performance gain may
overwhelm the cost of transferring the source code and it
may be quicker to ship all your source across the network to
compile it there.
<p>
In general, it is even better to compile on two or machines
in parallel. Any number of invocations of distcc can run at
the same time, and they will distribute their work across
the available hosts.
<p>
distcc does not manage parallelization, but relies on Make
or some other build system to invoke compiles in parallel.
<p>
With GNU Make, you should use the <tt/-j/ option to specify
a number of parallel tasks slightly higher than the number
of available hosts. For example:
<tscreen><verb>$ export DISTCC_HOSTS='angry toey wistful localhost'
$ make -j5</verb></tscreen>
</p>
<sect>
<heading>Choosing a Host</heading>
<p>
The <tt>$DISTCC_HOSTS</tt> variable tells distcc which
volunteer machines are available to run jobs. This is a
space-separated list of host specifications, each of which
has the syntax:
<tscreen><verb>HOSTNAME[/MAX_JOBS][:PORT]</verb></tscreen>
<p>
You can specify the maximum number of jobs that the host
should receive by affixing a number after a slash (e.g.
"localhost/2").
<p>
A numeric TCP port may optionally be specified after a
colon. If no port is specified, it uses the default, which
is currently 3632.
<p>
If only one invocation of distcc runs at a time, it will
always execute on the first host in the list. (This
behaviour is not absolutely guaranteed, however, and may
change in future versions.)
<p>
The name <tt>localhost</tt> is handled specially by running
the compiler in place.
<p>
The daemon may be tested on localhost by setting
<tscreen><verb>DISTCC_HOSTS=127.0.0.1</verb></tscreen>
Although <tt>localhost</tt> causes distcc to execute the job
directly, using an IP address will cause it to make a TCP
connection to a daemon on localhost. This is slower, but
useful for testing.
</sect>
<sect>
<heading>Load Distribution Algorithm</heading>
<p>
When distcc is invoked, it needs to decide which of the
volunteers in <tt>DISTCC_HOSTS</tt> should be used to
compile a job. It uses a simple heuristic to try to spread
load across machines appropriately.
</p>
<p>
You can imagine all of the compile machines as being leaky
buckets, some with larger holes (faster CPUs) than others.
The distcc client tries to keep water at the same level on
each one (the same number of jobs running), preferring hosts
occurring earlier in DISTCC_HOSTS. Over the course of a
build, the faster machines will complete jobs more quickly,
and therefore be topped up more quickly and do more work
overall, but without the client ever actually needing to
know which one is fastest.
</p>
<p>
This design has the advantage of not requiring the client to
know in advance the speeds of the volunteers, and being
quite simple to implement. It copes quite well with
machines that are temporarily slowed down: they are just
topped-up more slowly in the future.
</p>
<p>
Scheduling is coordinated between different invocations of
the <tt>distcc</tt> client by lockfiles in the temporary
directory. There is no coordination between clients running
as different users, on different hosts, or with different
<tt>TMPDIR</tt> paths.
</p>
<p>
On Linux, scheduling slightly too many jobs on any machine
is quite harmless, as long as the number is not so high that
the machine begins thrashing. So it's OK to provide a
<tt>-j</tt> number substantially higher than the number of
available processors.
</p>
<p>
The biggest problem with this design is that it handles
multiprocessor machines poorly: they probably ought to have
jobs scheduled proportional to the number of processors. At
the moment, the best thing is to run with a <tt>-j</tt>
factor equal to the product of the maximum number of CPUs in
any machine (<tt>MAX_CPUS</tt>) and the number of machines.
This should make sure that roughly <tt>MAX_CPUS</tt> tasks
run on every machine at all times, and will therefore keep
all CPUs loaded, but will cause excessive task-switching on
machines with fewer CPUs. Task switching is not very
expensive on Linux so it is not a big problem, but it does
lose a few percentage points of speed. This should be fixed
in a future release.
</p>
</sect>
<sect>
<heading>Diagnostic Messages</heading>
<p>
Error messages or warnings from local or remote compilers
are passed through to diagnostic output on the client. The
compiler takes all file names and line numbers from pragmas
in the preprocessed output, so error messages will always
have the correct pathnames for files on the client.
<p>
distcc can supply extensive debugging information when the
verbose option is used. This is controlled by the
<tt>$DISTCC_VERBOSE</tt> environment variable on the client,
and the <tt>-</tt><tt>-verbose</tt> option on the server.
</p>
<p>
By default, distcc prints diagnostic messages to stderr.
Sometimes these are too intrusive into the output of the
regular compiler, and so they may be selectively redirected
by setting the <tt>$DISTCC_LOG</tt> environment variable to a
filename.
</p>
<p>
Error messages from the daemon are sent to both the log file
on the volunteer, and also back to the client's diagnostic
output. (By default, it uses the syslog <tt>daemon</tt>
channel.) If compilation is failing, please examine the log
file on the relevant volunteer machine.
</sect>
<sect>
<heading>distcc Exit Codes<label id="exit"></heading>
<p>
The exit code of distcc is normally that of the compiler:
zero for successful compilation and non-zero otherwise.
</p>
<p>
If distcc fails to distribute a job to a selected volunteer machine,
it will try to run the compiler locally on the client. distcc only
tries a single remote machine for each job.
</p>
<p>
distcc tries to distinguish between a failure to distribute
the job, and a "genuine" failure of the compiler on the
remote machine, for example because of a syntax error in the
program. In the second case, distcc does not re-run the
compiler locally, and returns the same exit code as the
remote compiler.
</p>
<p>
If the compiler exits with a signal, distcc returns an exit
code of 128 plus the signal number, following Unix
convention.
</p>
<p>
If distcc fails to run the compiler, it may return one one
of the following error codes. These are also used by
distccd.
</p>
<p>
<descrip>
<tag>100 <tt/EXIT_DISTCC_FAILED/</tag>
<p>
Generic or unspecified failure in distcc.
</p>
<tag>102 <tt>EXIT_BIND_FAILED</tt></tag>
<p>
Failed to bind and listen on network socket. Port may
already be in use.
</p>
<tag>103 <tt/EXIT_CONNECT_FAILED/</tag>
<p>
Failed to establish network connection or listen on
socket. The host may be invalid or unreachable, or
there may be no daemon listening.
</p>
<tag>104 <tt>EXIT_COMPILER_CRASHED</tt>
<p>
The underlying compiler exited because of a signal.
This probably indicates a compiler bug, or a problem
with the hardware or OS on the server. (Obsolete in 0.13.)
</p>
<tag>105 <tt>EXIT_OUT_OF_MEMORY</tt>
<p>
Obvious.
</p>
<tag>106 <tt>EXIT_BAD_HOSTSPEC</tt>
<p>
<tt>$DISTCC_HOSTS</tt> was undefined, empty, or
syntactically invalid. (At the moment, you should never
see this code because distcc will fall back to building
locally. Let me know if you would prefer a hard error.)
</p>
<tag>107 <tt>EXIT_IO_ERROR</tt>
<p>
There was a disk or network IO error while distributing
the job. For example, the network may be failing, or a
disk on either end may be out of space.
</p>
<tag>108 <tt>EXIT_TRUNCATED</tt>
<p>
The network socket was closed unexpectedly. This
probably indicates a network problem or distcc bug.
</p>
<tag>109 <tt>EXIT_PROTOCOL_ERROR</tt></tag>
<p>
The distcc internal network protocol was not followed by
the remote program. This probably indicates a network
problem or distcc bug.
</p>
<tag>110 <tt>EXIT_COMPILER_MISSING</tt></tag>
<p>
The specified compiler was not found in the path of the
server.
</p>
<tag>111 <tt>EXIT_RECURSION</tt></tag>
<p>
distcc ended up invoking itself recursively. This is
caught as an error to prevent unbounded recursion or
inefficiency. For example, this error may occur if
distcc is called by the <tt>cc</tt> command in the
daemon's path.
</p>
</descrip>
</p>
</sect> <!-- end of "distcc Exit Codes" -->
<sect>
<heading>File Metadata</heading>
<p>
distcc transfers only the binary contents of source, error,
and object files, without any concern for metadata,
attributes, character sets or end-of-line conventions.
</p>
<p>
distcc never transmits file times across the network or
modifies them, and so should not care whether the clocks on
the client and volunteer machines are synchronized or not.
When an object file is received onto the client, its
modification time will be the current time on the client
machine.
</p>
</sect>
</chapt> <!-- End of "using distcc" -->
<chapt>
<heading>distcc Compatibility</heading>
<sect>
<heading>distcc with ccache</heading>
<p>
distcc works well with the <htmlurl
url="http://ccache.samba.org/" name="ccache"> tool for
caching compilation results. The best way to use the two of
them together is to create a masquerade dir for ccache and a
masquerade dir for distcc, and update your path to have the
ccache-in-disguise links be found first, followed by the
distcc-in-disguise links, and finally the real compiler.
For instance:
<tscreen><verb>PATH="/usr/lib/ccache/bin:/usr/lib/distcc/bin:$PATH"
export PATH</verb></tscreen>
</p>
<p>
Another alternative is to munge CC and/or CXX to mention both
commands:
<tscreen><verb>CC='ccache distcc'
CXX='ccache distcc g++'</verb></tscreen>
</p>
</sect>
<sect>
<heading>distcc with autoconf</heading>
<p>
distcc works quite well with autoconf.
</p>
<p>
<tt>DISTCC_VERBOSE</tt> can give autoconf trouble because
autoconf tries to parse error messages from the compiler.
If you redirect distcc's diagnostics using
<tt>DISTCC_LOG</tt> then it seems to be fine.
</p>
<p>
Some autoconf-based systems "freeze" the compiler name
used for configure into their Makefiles. To make them use
distcc, you should either set the PATH to include the distcc
masquerade dir or set <tt>$CC</tt> prior to running
<tt>./configure</tt>, and/or override <tt>$CC</tt> on the
right-hand-side of the Make command line.
</p>
<p>
Some poorly-written shell scripts may assume that
<tt>$CC</tt> is a single word. Using a masquerade dir
avoids this problem.
</p>
</sect>
<sect>
<heading>distcc with libtool</heading>
<p>
Some versions of libtool seem not to cope well when CC is
set to more than one word, such as <tt>"distcc gcc"</tt>.
Setting <tt>CC=distcc</tt>, which is supported in 0.10 and
later, seems to work well. Using a masquerade dir, which
was first supported in 2.0, is even better (since it avoids
problems with C++ programs too).
</p>
</sect>
<sect>
<heading>distcc with Gentoo Linux</heading>
<p>
Gentoo is a "ports"-based free software distribution, in
which packages are always built from source on
installation. distcc works well with Gentoo to speed
installation.
</p>
<p>
You can install distcc either using the upstream tarball
from <tt>distcc.samba.org</tt> (which may be newer), or
using <tt>emerge distcc</tt> to get the Gentoo port, which
may be better integrated. You can also get ccache with
<tt>emerge ccache</tt>.
</p>
<p>
Using distcc (and ccache) to speed your "emerge" commands is
simplicity itself since full support for these packages is
already built into portage. The two items you need to customize
for your system are the MAKEOPTS value and the DISTCC_HOSTS
value. You can either set these in the /etc/make.conf file, or
you can export them from your local environment. Then, simply
/etc/make.conf to add the "distcc" (and "ccache") features to the
"FEATURE" setting (uncommenting the line, if needed), start up
any missing remote distccd servers, and portage will take care
of munging the PATH that the ebuild uses in an appropriate
manner (note that the Gentoo version of distcc comes with a
masquerade dir as a part of the standard installation, and has
had this since distcc-1.1-r8 (when it was first patched into
the ebuild). For example, if you wanted 9 parallel compiles
and you had 4 remote systems, you might run something like this:
<tscreen><verb>export MAKEOPTS=-j9
export DISTCC_HOSTS='local larry moe curly shemp'
emerge </verb><em>packagename</em></tscreen>
</p>
<p>
To use distcc for compilations outside of an emerge/ebuild
command, include the /usr/lib/distcc/bin dir early on your
PATH before configuring/compiling the package.
</p>
</sect>
<sect>
<heading>distcc with gcc dependency computation</heading>
<p>
gcc has the ability to produce information about header
dependencies as a side-effect of preprocessing. These can
be included in Makefiles in various ways to make sure that
files are up-to-date.
</p>
<p>
This feature is enabled using <tt>-MD</tt>, <tt>-M</tt>
and related options.
</p>
<p>
Unfortunately, gcc changed the behaviour of this feature
between gcc 2.95 and 3.x in such a way that it seems
properly for distcc to generally support it. The
difficulty is that the filename to which dependencies are
written depends in a very complicated way on the gcc
command line. distcc needs to change the command line to
run the preprocessor locally and the compiler remotely,
and this can sometimes cause problems. (This also causes
problems for Makefiles that are supposed to work with both
versions of the compiler.)
<p>
<tt>-M</tt> causes gcc to produce dependency information
instead of compiling. distcc understands this and passes
the option straight through to gcc. It should work correctly.
<p>
With gcc 2.95, <tt>-MD</tt> always writes dependencies
into the preprocessor's working directory. distcc should
work fine.
<p>
With gcc 3.2, <tt>-MD</tt> writes the output into either
the source directory or output directory, depending on the
presence of the <tt>-o</tt> option. However, gcc 3.2
also has a <tt>-MF</tt> option that can be used to
explicitly set the dependency output file, and this works
well with distcc.
<p>
<p>
In summary: for gcc 2.95, no changes are required. For
gcc 3.2, <tt>-MF</tt> should be used to specify the file
to write dependencies to.
</p>
</sect>
</chapt>
<chapt>
<heading>The distccd Server</heading>
<p><label id="server">
The distccd server may be started either from a super-server
such as <tt>inetd</tt>, or as a stand-alone daemon.
</p>
<p>
distccd does not need to run as root and should not.
</p>
<p>
distccd does not have a configuration file; it's behaviour is
controlled only by command-line options and requests from
clients.
</p>
<sect>
<heading>Invoking distccd</heading>
<p>
These options may be used for either inetd or standalone
mode.
</p>
<p>
If you want to see if the daemon started properly, look in the
log file. By default this is something like
<tt>/var/log/daemon</tt> or <tt>/var/log/messages</tt>,
depending on your system.
</p>
<p>
<descrip>
<tag><tt>-</tt><tt>-help</tt></tag>
<p>
Explains usage of the daemon and exits.
</p>
<tag><tt>-</tt><tt>-version</tt></tag>
<p>
Shows the daemon version and exits.
</p>
<tag><tt>-N</tt>, <tt>-</tt><tt>-nice NICENESS</tt>
<p>
Makes the daemon more nice about giving up the CPU to
other tasks on the machine. <em>NICENESS</em> is a
value from 0 (regular priority) to 20 (lowest
priority). This option is good if you want to run
distccd in the background on a machine used for other
purposes.
</p>
<tag><tt>-p, -</tt><tt>-port PORT</tt></tag>
<p>
Set the TCP port to listen on. (Standalone mode only.)
</p>
<tag>
<tt>-P, -</tt><tt>-pid-file FILE</tt>
</tag>
<p>
Save daemon process id to file.
</p>
<tag>
<tt>-</tt><tt>-verbose</tt>
</tag>
<p>
Include debug messages in log.
</p>
<tag>
<tt>-</tt><tt>-no-detach</tt>
</tag>
<p>
Do not detach from the shell that started the daemon.
This may be useful when running distccd from a system
such as <em>daemontools</em> that manages daemons after
they start.
</p>
<tag>
<tt>-</tt><tt>-no-fork</tt>
</tag>
<p>
Don't fork children for each connection, to allow
attaching <tt>gdb</tt>. <bf>Don't use this if you don't
understand it!</bf>
</p>
<tag>
<tt>--no-fifo</tt>
</tag>
<p>
Send input to the compiler by writing to a temporary
file, rather than using a pipe. This is required when
the server's temporary directory is on NFS, on at least
some machines. It may be faster in some circumstances,
but probably is not.
</p>
<tag>
<tt>-</tt><tt>-log-file=FILE</tt>
</tag>
<p>Send messages here instead of syslog.</p>
<tag><tt>-</tt><tt>-log-stderr</tt></tag>
<p>
Send log messages to stderr, rather than to a file or
syslog. This is mainly intended for use in debugging.
</p>
<tag><tt>-</tt><tt>-inetd</tt>
<p>
Serve a client connected to stdin/stdout. As the name
suggests, this option should be used when distccd is run
from within a super-server like <tt>inetd</tt>. distccd
assumes inetd mode when stdin is a socket.
</p>
<tag><tt>-</tt><tt>-daemon</tt>
<p>
Bind and listen on a socket, rather than running from
inetd. This is used for standalone mode. distccd
assumes daemon mode at startup if stdin is a tty, so
<tt>--daemon</tt> should be explicitly specified when
starting distccd from a script or in a non-interactive
ssh connection.
</p>
</descrip>
</p>
</sect>
<sect>
<heading>distccd Exit Codes</heading>
<p>
As for distcc <ref id="exit">.
</p>
</sect>
<sect>
<heading>distccd Environment Variables</heading>
<p>
<descrip>
<tag><tt>DISTCC_SAVE_TEMPS</tt>
<p>
If set to <tt>1</tt>, temporary files are not deleted
after use. Good for debugging or if your disks are too
empty.
</p>
</descrip>
</sect>
</chapt> <!-- End of "The distccd Server" -->
<chapt>
<heading>Cross compiling</heading>
<p>
Cross compilation means building programs to run on a
machine with a different processor, architecture, or
operating system to where they were compiled. distcc
supports cross compilation, including teams of
mixed-architecture machines, although some changes to the
compilation commands may be required.
</p>
<p>
The compilation command passed to distcc must be one that
will execute properly on every volunteer machine to produce
an object file of the appropriate type. If the machines
have different processors, then simply using <tt>distcc
cc</tt> will probably <bf>not</bf> work, because that will
normally invoke the volunteer's native compiler.
</p>
<p>
Machines with the same instruction set but different
operating systems may not necessarily generate compatible .o
files. Empirically it seems that the native FreeBSD
compiler generates object files compatible with Linux for C
programs, but not for C++. It may be a good idea to install
a Linux cross compiler on BSD volunteers.
</p>
<p>
Different versions of the compiler may generate incompatible
object files. This seems to be much more of a problem with
C++ than with C, because the C++ ABI (application binary
interface) has changed in recent years. If you will be
building C++ programs, it may be a good idea to install the
same version of <tt>g++</tt> on all machines.
</p>
<p>
Several different gcc configurations can be installed
side-by-side on any machine. If you build gcc from source,
you should use the <tt>--program-suffix</tt> configuration
options to cause it to be installed with a name that encodes
the gcc version and the target platform.
</p>
<p>
The recommended convention for the gcc name is
<em>target</em><tt>-gcc-</tt><em>version</em>, such as
<tt>i686-linux-gcc-3.2</tt>. GCC 3.3 will install itself
under this name, in addition to <em>target</em><tt>-gcc</tt>
and, if it's native, <tt>gcc-</tt><em>version</em> and
<tt>gcc</tt>.
</p>
<p>
The compiler must be installed under the same name on the
client and on every volunteer machine.
</p>
<p>
gcc also has <tt>-b</tt> and <tt>-V</tt> options to specify
a target and version, but at the moment the gcc team
recommend using a qualified compiler name instead.
</p>
<p>
For more information on cross-compiling, see <em>Specifying
Target Machine and Compiler Version</em> in the gcc manual,
and the gcc installation guide.
</p>
</chapt> <!-- end of "Cross compiling" -->
<chapt>
<heading>distcc Internals</heading>
<sect>
<heading>
Protocol
</heading>
<p>
distcc uses a simple, application-specific protocol running
directly over a TCP socket. A new request socket is opened
for each job.
</p>
<p>
The request and response begin with a magic number and
version number, allowing incompatible versions or
misconfigurations to be identified. At the moment there is
only one deployed protocol version, and no attempt to
support backward or forward compatibility, though this could
be added in the future.
</p>
<p>
The request and response consist of tagged, length-preceded
elements. Each element of the request contains a
four-character ASCII token, an eight-digit ASCII hexadecimal
length or value, and, depending on the tag, a byte stream
whose length is determined by the hexadecimal field.
</p>
<p>
The complete request is sent to the server before the reply
begins. Opening the TCP socket is performed concurrently
with execution of the preprocessor on the client.
</p>
<p>
The request from the client contains
<enum>
<item>Magic number and version</item>
<item>Compiler command line</item>
<item>Preprocessed source code</item>
</enum>
</p>
<p>
The response from the server contains
<enum>
<item>Magic number and version
<item>Compiler exit code & status
<item>Compiler error messages
<item>Compiler stdout
<item>Object file (if any)
</enum>
</p>
<p>
Consult the source for more information.
</p>
</sect>
<sect>
<heading>
Working files
</heading>
<p>
distcc stores working files in a subdirectory of
<tt>/tmp</tt>. These include synchronization files, and
compiler input/output temporary files.
<p>
Temporary files should normally be cleaned up when the
program exits. If distcc misbehaves, these files may be
useful in tracking down the cause. Any that remain can be
removed by the system's temporary file reaper, or by hand.
</p>
</sect>
<sect>
<heading>
Lock files
</heading>
<p>
distcc uses lock files to allow each client to balance its
jobs across available volunteer machines. For each
volunteer host, a zero-length file is created. Clients
using that volunteer hold a <tt>flock</tt> lock on the file
while running.
</p>
</chapt>
</book>
</linuxdoc>