Skip to main content.

Web Based Programming Tutorials

Homepage | Forum - Join the forum to discuss anything related to programming! | Programming Resources

Perl 5 Unleashed

Chapter 24 -- Building and Installing the Perl 5 Interpreter

Chapter 24

Building and Installing the Perl 5 Interpreter


This chapter covers the gory details of installing Perl 5 on a given system. First, you find out where to get the distributions via the Internet. Then you learn the details of some of the things you need to do to get the Perl 5 package built and installed on to your machine. I cover building the sources for a UNIX system here in detail because this is what most of the sources are targeted for anyway. However, I also cover how to get Perl 5 on other platforms.

Getting Perl 5

There are several ways to get the Perl 5 package for your system. Some of the ways include buying a commercial package, getting it from the Internet via FTP or e-mail, or having someone else get it for you.

Getting Perl from the Internet

The latest Perl packages are usually available from any comp.sources.misc archive on a given news site. The latest source distribution will almost always be at Larry Wall's site. Here's the address:

Larry Wall maintains the official distribution site (for both perl4.036 and perl5) at his site. Always check this site first for the latest versions and, if you are the brave sort, for the beta releases being worked on.

A number of archives have materialized over the years, and these archives hold Perl source code and related items. Probably the largest archive is at the University of Florida. The Comprehensive Perl Archive Network (CPAN) is a collection of all Perl materials including source files, modules, and documentation. In order of probability, these other sites should also have the Perl source files:

By the time you get this book, this listing might be superseded by newer sites. The best course of action is to use archie or a Web search site such as Yahoo to search for the words Perl, CPAN, or Perl5 to get the list of the latest versions of Perl 5 packages.

Getting Perl via E-mail

If you do not have direct FTP access, all is not lost. You can get the packages via e-mail. Table 24.1 is a small list of known FTP mail sites. As a general rule, you should always attempt to use the site closest to you with the FTP archive closest to it. Many of these sites already have Perl on them.

For information on how to use one of these sites, send e-mail containing the word help to the address.

Table 24.1. Location of e-mail addresses that have Perl.
LocationE-mail Address

Send a message to containing this:

path your_email_address
send help
send PERL/perl5.0/INDEX

You can omit the path line if your message contains a normal From: line. You'll receive a help file and an index of the directory that contains the Perl materials.

Web Sites with Perl Information

The following Web sites contain information on Perl 5. You can get more listings by using the Net search facilities on the Web by searching for the keyword Perl5.

Naturally, your Web search tool will return a different list of directories. These sites will get you started at the very least.

The FAQ for Perl 5 is available at this site:


The best collection I've found to date is the InfoMagic release of the CD-ROM Mother of Perl. This two-disc set includes the archives, mailing list archives, and lots of mirrors of CPAN sites. My disc was up to 5.001m, but a more current disc set should be available at the time this book goes to print. For more information contact InfoMagic at this address:

Box 30370
Flagstaff, AZ 86003

Installing Perl on Most UNIX Systems

Installing Perl takes time, but it's easy. I cover the installation on a UNIX-based system because this covers the most general installations of Perl. Please check the section titled "Other Platforms" for more instructions if you are not on a UNIX platform.

Here are the basic steps to follow:

  1. Make sure you have all the files.
  2. Configure your Perl distribution for your system.
  3. Construct the dependencies.
  4. Make the executables in your local directory.
  5. Run a test version to see if everything is okay.
  6. If everything checks out, make the distribution available to the rest of the users on your machine.

First, unzip and untar the Perl 5 package in a directory not in the /usr directory tree. You'll install this distribution in the /usr or /usr/local later. For now, just use a subdirectory in your home directory.

Go to the directory where you downloaded your Perl distribution. Then unzip and untar the distribution files with these commands:

$ gunzip XXXX.tar.gz
$ tax -cvf XXXX.tar

where XXXX is the name of the Perl distribution you got from the Internet. At the time this book was written, this was perl5.002.tar.gz. The 5.002 version is installed and used for all examples in this book.

The INSTALL file is the master document you need to refer to if this chapter does not get you started. The text in the INSTALL file is in the plain old documentation (pod) format. Run pod2man to get the documentation in the man page format.

The MANIFEST file should contain the information you need on what files go where. The files after untarring the package should look like those shown in Listing 24.1.

Your Perl directory should contain the following source files after you untar the package:

Listing 24.1. MANIFEST files content list.  
Artistic     configure   hints           op.h          proto.h
Changes      cop.h       hv.c            opcode.h      regcomp.c
Changes.Conf cv.h        hv.h       regcomp.h
Configure    deb.c       installman      os2           regexec.c
Copying      doio.c      installperl     patchlevel.h  regexp.h
EXTERN.h     doop.c      interp.sym      perl.c        run.c
INSTALL      dosish.h    keywords.h      perl.h        scope.c
INTERN.h     dump.c     perl_exp.SH   scope.h
MANIFEST     eg          lib             perlsh        sv.c
Makefile.SH  emacs       makeaperl.SH    perly.c       sv.h
README       embed.h     makedepend.SH   perly.c.diff  t
README.vms    makedir.SH      perly.fixer   taint.c
Todo         ext         malloc.c        perly.h       toke.c
XSUB.h       form.h      mg.c            perly.y       unixish.h
av.c         global.sym  mg.h            pod           util.c
av.h         globals.c      pp.c          util.h
cflags.SH    gv.c        miniperlmain.c  pp.h          utils
config_H     gv.h        mv-if-diff      pp_ctl.c      vms
config_h.SH  h2pl        myconfig        pp_hot.c      writemain.SH
configpm     handy.h     op.c            pp_sys.c      x2p

Configuring Your Perl Distribution

To configure your system, first run the Configure command in the perl5 directory where your source files were untarred. The Configure command will figure out various settings about your system. Configure will be able to default to many settings by itself; for other things it will ask you about. Figure 24.1 contains the output from the Configure command.

Figure 24.1 : The output from a configure command.

If the test scripts and programs run okay, the defaults are usually right. Configure will then create a set of defaults for itself and then proceed to make three files to store these sets of defaults: config.h,, and Makefile.

If your default shell is not ksh, bash, or the Bourne shell, sh, you may have to enter the following command:
sh Configure
This step will ensure that Configure is run under sh.

If you feel lucky, run Configure -d to take all the defaults, and edit yourself to patch up any flaws. I do not recommend this course of action unless you're a Perl wizard, in which case you probably would have skipped this entire chapter.

The Configure command supports a lot of options. The one option to know about is the -h option, which prints a list of the available options for Configure. The options are useful in overriding any predefined values. For example, you may have cc's and GNU's compilers both on your system with cc as the default. To force Configure to use gcc-elf, run the command Configure -Dcc=gcc-elf. Alternatively, in the interactive mode, you can answer gcc-elf at the cc prompt.

By default, Perl will be installed in the following directories:

You can specify a different directory for the default installation location as long as you have read and write access to the directory. The default directory pathname can be changed with the -D flag; for example, for specifying a pathname to a subdirectory in my home directory I would use the command line option -Dprefix='/home/khusain/perl5'. The replacement for /usr/local is changed when you respond to the question of prefix in Configure script.

A couple of other points about running Configure: Do not reuse your old config.h or config.SH file from a previous version. Too many things may have changed. This is true for patches to the Perl package as well. If you want to see some differences, save a copy of the old file, rerun the Configure command, and then run a diff command to see the differences.

Actually, if you have some overrides that you want to use regardless of those set up with Configure, you can put them in a file called config.over. This file can be used by Configure and overrides any values just before it writes the file. To get a head start, copy onto config.over, and then edit config.over. The details of this procedure are given in the INSTALL document.

Perl will use any dynamic extensions if your system supports them. If you want to force Perl to be compiled and linked statically, then you can choose this when Configure prompts you about it. Most of the later versions of Perl use dynamic loading; therefore, if your system does not support it, consider upgrading. See the section: "Linux: A Special Case for Perl 5," in this chapter to see what some of the issues are.

While you are running Configure, be patient. Answer each question after reading it carefully. Yes, the defaults are great, but ultimately you are the one who has to run Perl on your system. Make sure you know enough about your system to know if the defaults look right or not.

If your system is not recognized, you'll most certainly have to answer questions alertly. The Configure script tells you if it has recognized your system, as shown in Figure 24.2.

Figure 24.2 : Recognition of a computer system by the configure command.

Also, do not hold down the Enter key or press it too many times. The returns may remain queued and provide a default reply when you really did not mean to.

There is no way to back up or re-enter all the information if you make a mistake. You have to start the entire process all over from the initial Configure command. Keep this in mind as you answer the questions. You can backspace on the current line, but you cannot go back to the previous prompt. The prompter will recognize $* as the entire default line should you want to append to the default values given. Arguments to responses can also be referred to as $1, $2, and so on to keep you from too much typing or typographical errors.

Use sh at least or some other shell instead that looks like the Bourne shell for the purpose of this installation. If you are bent on using your own version of csh, you'll have to copy the sample file config_H to config.h, and edit config.h to adapt to your shell.

Now would be a good time to peruse the config.h file to make sure that the system dependencies are correct. Most of them are taken care of by running the Configure script, but this will give you some insight on how the configuration sets everything up.

If you have any additional changes to make to the C definitions, make them in the cflags.SH file. For example, to turn off the optimizer on toke.c, find the line in the switch structure for toke.c and put the command optimize='-g' before the double semicolons. By default, the value is set to $optdebug.

To change the C flags for all the files, edit and change either $ccflags or $optimize. In order to ensure that the changes you have made to these header files propagate to all other programs, you must run the following command:

Configure -S ; make depend

Any changes you make to the configuration files will not be automatically propagated to the rest of the files-you must run Configure -S and make depend to ensure that all affected files are updated.

Making Dependencies

After you've configured your Perl package for your system, you must set up all the dependencies for the source files. This is done by typing the command make depend. The make file looks for all the affected include files and modifies the Makefile accordingly.

The Configure program will offer to do this for you; you just have to answer Yes, as shown in Figure 24.3. If you do not get this prompt, you can always run it from the command line.

Figure 24.3 : Running make depend from configure.

Now, attempt to make Perl in the current directory. Run the make command from the prompt without any arguments. The execution of this make command will take a while.

Generally speaking, you should encounter no errors. If you see errors and if these errors impede the development of the platform, the make program will terminate. If the errors are not critical or can be compensated for, make will continue and generate the correct files for you. Linux users should refer to the online documentation on ELF in Linux.

No errors during the compile phase does not equal a clean build in this case. Finish the entire procedure by testing the installation.

You may see some warnings, too. Again, this depends on the platform on which you are building Perl. In most cases, the warnings can be ignored. If you do see warnings about LITTLE_ENDIAN being redefined, ignore them. For your peace of mind, you might want to check the definitions of LITTLE_ENDIAN in the files that the make file warns you about. Both the declarations should be the same-that is, either 1234 or 4321, depending on the type of machine you happen to be building the version of Perl on.

The gcc compilers installed in your system should be able to handle the Perl distribution with only a few burps. For example, some function prototypes might not be recognized by the compiler. There is a way to counter some of these errors. For instance, try adding a -DCRIPPLED_cc flag. You might even consider turning off the -O2 optimization flag in in your compiler.

If you get duplicate function definitions (a Perl function has the same name as another function on your system), try using the -DEMBED switch to get rid of the error. These errors are rare, but not impossible to come by.

The worst error you could run into is an incomplete distribution. In this case, the only recourse you have is to abandon all efforts with your distribution and just start over with a fresh release from the Internet. Chances are that even if you actually paid someone for this crippled Perl release, they won't be interested in providing you with free support.

The file is the one you need to know about if you have to modify the compiler switches. The file is too long to list here, but I have listed some of the switches that you should definitely take a look at if you need to correct something:

bin='/usr/bin' & binexp='/usr/bin'
ccflags='-D__USE_BSD_SIGNAL -Dbool=char -DHAS_BOOL'
cppflags='-D__USE_BSD_SIGNAL -Dbool=char -DHAS_BOOL'
extensions='DB_File Fcntl GDBM_File ODBM_File POSIX SDBM_File Socket'
libs='-lgdbm -ldbm -ldb -ldld -lm -lc -lbsd'


Some compilers will not compile or optimize the larger files without some extra switches for using larger jump offsets or allocating larger internal tables.

Perl without any extensions is referred to as Miniperl. If you run into problems while building with extensions, you should try to build Miniperl. If you can successfully build Miniperl but the process crashes during the building of extensions, you need to get the later versions of the extensions. In such an event, get a completely new package and try to avoid doing piecemeal patches, unless you know what you are doing-in which case you should not even be reading this chapter ;-) .

Machines with incomplete dbm routines (specifically the dbm_open() function) will need to set the #undef I_ODBM preprocessor directive. For duplicate reference error messages when linking files that call malloc, try the -DHIDEMYMALLOC switch. Both of these problems did not disappear on my Linux machine until I upgraded the entire system and kernel to ELF. I did not have to use I_ODBM or the -DHIDEMYMALLOC flag after the upgrade.

What Are hints Files?

In the Perl distribution, you'll find a ./hints subdirectory. The ./hints subdirectory is where platform-specific declarations are kept. The files here are used by Configure to set things that Configure either cannot or does not guess properly. You may have to edit these files for special versions of your computer operating system.

Generally, you should not have to use the hints file if you are using a UNIX-like operating system. Check the contents of this directory to see whether your system is listed and whether Perl works correctly with it. Check the README.hints file for a list of those systems for which Perl has been tested.

A sample listing of the file is shown in Listing 24.2.

Listing 24.2. The hints file.
  1 # hints/
  2 # Original version by rsanders
  3 # Additional support by Kenneth Albanowski <>
  4 #
  5 # ELF support by H.J. Lu <>
  6 # Additional info from Nigel Head <nhead@ESOC.bitnet>
  7 # and Kenneth Albanowski <>
  8 #
  9 # Consolidated by Andy Dougherty <>
 10 #
 11 # Last updated Tue May 30 14:25:02 EDT 1995
 12 #
 13 # If you wish to use something other than 'gcc' for your compiler,
 14 # you should specify it on the Configure command line.  To use
 15 # gcc-elf, for exmample, type
 16 # ./Configure -Dcc=gcc-elf
 18 # perl goes into the /usr tree.  See the Filesystem Standard
 19 # available via anonymous FTP at in
 20 # /pub/linux/docs/linux-standards/fsstnd.
 21 # Allow a command line override, e.g. Configure -Dprefix=/foo/bar
 22 case "$prefix" in
 23 '') prefix='/usr' ;;
 24 esac
 26 # Perl expects BSD style signal handling.
 27 # gcc-2.6.3 defines _G_HAVE_BOOL to 1, but doesn't actually supply bool.
 28 ccflags="-D__USE_BSD_SIGNAL -Dbool=char -DHAS_BOOL $ccflags"
 30 # The following functions are gcc built-ins, but the Configure tests
 31 # may fail because they don't supply proper prototypes.
 32 # This should be fixed as of 5.001f.  I'd appreciate reports.
 33 d_memcmp=define
 34 d_memcpy=define
 36 # Configure may fail to find lstat() because it's a static/inline
 37 # function in <sys/stat.h>.
 38 d_lstat=define
 40 # Explanation?
 41 d_dosuid='define'
 43 # I think Configure gets this right now, but I'd appreciate reports.
 44 malloctype='void *'
 46 # Explanation?
 47 usemymalloc='n'
 49 case "$optimize" in
 50 '') optimize='-O2' ;;
 51 esac
 53 # Are we using ELF?  Thanks to Kenneth Albanowski <>
 54 # for this test.
 55 cat >try.c <<'EOM'
 56 /* Test for whether ELF binaries are produced */
 57 #include <fcntl.h>
 58 #include <stdlib.h>
 59 main() {
 60             char buffer[4];
 61             int i=open("a.out",O_RDONLY);
 62             if(i==-1)
 63                         exit(1); /* fail */
 64             if(read(i,&buffer[0],4)<4)
 65                         exit(1); /* fail */
 66             if(buffer[0] != 127 || buffer[1] != 'E' ||
 67             buffer[2] != 'L' || buffer[3] != 'F')
 68                         exit(1); /* fail */
 69             exit(0); /* succeed (yes, it's ELF) */
 70 }
 71 EOM
 72 if ${cc:-gcc} try.c >/dev/null 2>&1 && ./a.out; then
 73     cat <<'EOM'
 75 You appear to have ELF support.  I'll try to use it for dynamic loading.
 76 EOM
 77 else
 78     cat <<'EOM'
 80 You don't have an ELF gcc.  I will use dld if possible.  If you are
 81 using a version of DLD earlier than 3.2.6, or don't have it at all, you
 82 should probably upgrade. If you are forced to use 3.2.4, you should
 83 uncomment a couple of lines in hints/ and restart Configure so
 84 that shared libraries will be disallowed.
 86 EOM
 87     lddlflags="-r $lddlflags"
 88     # These empty values are so that Configure doesn't put in the
 89     # Linux ELF values.
 90     ccdlflags=' '
 91     cccdlflags=' '
 92     so='sa'
 93     dlext='o'
 94     ## If you are using DLD 3.2.4 which does not support shared libs,
 95     ## uncomment the next two lines:
 96     #ldflags="-static"
 97     #so='none'
 98 fi
100 rm -f try.c a.out
102 if /bin/bash -c exit; then
103   echo You appear to have a working bash. Good.
104 else
105   cat << 'EOM'
106 Warning: it would appear you have a defective bash shell installed. This is
107 likely to give you a failure of op/exec test #5 during the test phase of the
108 build, Upgrading to a recent version (1.14.4 or later) should fix the
109 problem.
111 EOM
113 fi
115 # In addition, on some systems there is a problem with perl and NDBM, which
116 # causes AnyDBM and NDBM_File to lock up. This is evidenced in the tests as
117 # AnyDBM just freezing.  Currently we disable NDBM for all linux systems.
118 # If someone can suggest a more robust test, that would be appreciated.
119 # This will generate a harmless message:
120 # Hmm...You had some extra variables I don't know about...I'll try to keep 'em.
121 #          Propagating recommended variable d_dbm_open
122 d_dbm_open=undef

A hint is simply a shell script run by Configure that sets things it needs for your system. Usually the hint files are documented well enough to know why certain variables are defined.

And sometimes they're not. It's usually not that hard to figure out what the define is about, though. Case in point: the lines from the file. Turning malloc off turns off Perl's version of malloc. There is really no need for an explanation here:

# Explanation?

I use here as the example because it's the system I happen to run. If you are on an SCO system, you have some choices to make, depending on the type of system you have. Here are the hints files included with 5.002:                  sp;
3b1cc           powerunix    .sh
README.hints        sco_2_3_0    .sh        sco_2_3_1    .sh       sco_2_3_2    .sh         sco_2_3_3    .sh          sco_2_3_4    .sh        sco_3        .sh

Running Some Tests

Just because everything compiled doesn't mean you are out of the woods. You have to run the regression tests on the perl you just made. You should see the message All tests successful. (See Figure 24.4.)

Figure 24.4 : Results from a test run.

If it doesn't say All tests successful, something went wrong. Figure 24.4 shows a sample run without any error messages.

You can't run the make test command in the background because running that command disables the opening of the device /dev/tty.

In the unfortunate event that you do not see the successful message, you have to figure out what went wrong. Go to the t directory. (I just love these directory names in UNIX.) Then run the TEST command to see the results.

If an individual test has cratered, you can run that test manually with this command:

./perl op/groups.t

Note that in the t directory, ./perl is linked to ../perl. This is a fully working version of the real enchilada!

Finishing It Up

You've now run the test, and the doubts in your mind are a faded memory. Let's discuss how to install Perl so that other users on your system are able to use it.

It would be a good idea to run as root now if you are writing in the /usr tree. Of course, you could have specified a totally different location over which you have write privileges-in which case being root is a moot point. Keep in mind that if you are not root, you must own the directories where you are writing and ignore any messages about chown not working.

Now, take a deep breath and run the make install command:

make install

This command puts Perl into a public directory. If you used the defaults, the location is /usr/local/bin. The make install command will also try to put the man pages in a reasonable place, possibly under /usr/man. You may have to nroff the man page yourself, though. (In newer systems, the first time you run the man(1) command, it will automatically run nroff and update the cat-files for you.)

The make install command creates the following files for you (I'm using /usr/local/bin as the default location):

The library files go in $privlib and $archlib as defined in the file. (This is defaulted to /usr/local/lib/perl5/.) The Perl documentation pages go in /usr/lib/perl5/pod/. Perl's *.h header files and the libperl.a library are also installed under $archlib so that you may later build new extensions even if the Perl source is no longer available. The man pages go in
/usr/local/man/man1 and /usr/local/lib/perl5/man/man3.

After the installation is complete, you should see the message Installation complete, as shown in Figure 24.5. If you do not see this message, you have to run this script again to see what went wrong. Chances are that either you do not have permission to write to the directories you requested or that the directory in which you chose to install Perl doesn't have the space, or both.

Figure 24.5 : Installation and verification.

This is optional. If you want to make the documentation available in man, HTML, or TeX format, use the following commands for HTML files:

cd pod; make html

For the documentation in TeX format, from the Perl source directory type the following command:

cd pod; make tex

Man pages are available for individual pod files via this command:

pod2man myfile.pod | nroff -man | less

Actually, if you're given a .pod file, running the pod2html command produces HTML files for you to peruse. I find that it's easier to search and read these HTML files using a Web browser than it is to mess around with TeX and man page files.

Confirming the Installation

To confirm that you've indeed installed Perl on your machine, type the perl -v command. You should see the output shown here:

$ which perl
$ perl -v

This is perl, version 5.002

Copyright 1987-1996, Larry Wall
Perl may be copied only under the terms of either the Artistic License or the GNU General Public License, which may be found in the Perl 5.0 source kit.

The DOS Platform

A beta version of Perl based on the Perl 5.000 version is available in binary form only, from this site in Canada:

The perl5a.README file with this version of Perl has a very ominous warning message with it. The message reads verbatim:

-r--r--r--   1 bin      bin      1170891 Oct 28 09:37
-r--r--r--   1 bin      bin      1198985 Oct 28 09:37
-r--r--r--   1 bin      bin       575899 Oct 28 09:37

This is, basically, a snapshot of my sources and executables, and is intended for use by people who want a "preview" of what I'm doing. It is not intended for production work, although I imagine that some people may want to use it for that. You must get all three .zip files. The sources and executables are all mixed together with these files, and I have not yet made an attempt at creating a minimal "binaries + support files" package.

This was enough to warn me not to install them on my machine. I'm still happy with perl4 with DOS, thank you very much. If you're a hotshot, be my guest. For more information about this port, please contact the author, Darryl Okahata, at The versions under development include those for DOS and Microsoft windows. If you feel that you must work with Perl 5 features and have to work under DOS, you're stuck. Perhaps you should consider installing Windows 95 or NT on your machine and running the version of Perl 5 for Windows NT. Please see Chapter 9, "Portability Issues with Windows and Perl."

Other Platforms

Perl 5 is being ported to other platforms and is at various stages of development. Some common platforms are listed here. You should check out these locations. In general though, you'll find only binary files and not the sources.

I found the following ports in various stages of development at the FTP site

Check the directories in /pub/languages/perl/CPAN/ports. There may be other levels of subdirectories below these directories. Some have documentation, some source only, and some complete packages. For example, the winNT subdirectory has a Perl 5.001m port in ./winNT/perl5/perl5.001m, whereas the amiga directory just has one file in it. As you can see, there is considerable effort being put into porting this powerful language to other platforms.

Please see Chapter 9, "Portability Issues with Windows and Perl," for complete installation instructions and discussion of available features of Perl for Windows.

Linux: A Special Case for Perl 5

Notably missing from the previous list is the Linux operating system. With the Linux operating system, you'll need the Executable and Linking Format (ELF) binaries to use the dynamic loading features in Perl. All major releases of the Linux kernel and libraries will support the ELF format in the future. If you do not have ELF binaries, you might be left behind. The motivation for going to ELF is simple: The sheer size of static libraries makes it very prohibitive to construct small applications anymore. Also, the current a.out format for libraries does not permit the use of the dynamic loading function dlopen(). Using dynamic loading is critical for the newer programs that are coming out for newer UNIX and UNIX clone systems.

The way to tell whether you are already running ELF binaries is to compile a short program and then run the file command on it. The ensuing output from the file command should be the following if you are running the ELF binaries:

$ file `which gcc`
/usr/bin/gcc: ELF 32-bit LSB executable I386 (386 and up) Version 1.

Also, check the version numbers on your Linux kernel. Type the following command to check to see what your version number is. The non-ELF binaries will be located in the 1.2.13 version. Most versions after 1.3.18 are ELF compatible:

$ file `which perl`
/usr/bin/perl: ELF 32-bit LSB executable i386 (386 and up) Version 1

On ordering a Linux distribution and after going through the normal installation process, I found that my system defaulted to 1.2.13 and no option on the installation screen allowed me the choice to switch to an ELF system. Newer releases of Linux (March 1996 and later) give you options to install ELF binaries and source files.

You can build the ELF kernel using the existing kernel and gc binaries. In order to build the binaries, you have to have the following items:

The binary ELF compiler, gcc, is available from The files are located in the Linux /packages/Gcc directory on most Internet sites. The latest version of the gcc compiler is 2.7.2. The binary files can be downloaded from the Internet and unzipped in the root directory. (The ELF-HOWTO file is correct in this regard.) Just get the gcc binary files and read the instructions on how to unzip, untar, and install the binary files.

To unbundle all the source files, you need the file linux-1.3.xx.tar.gz. The xx in my case is 20; however, by the time this book goes to print, a newer release might be out. (It's probably in your best interest to get the latest release from the Internet archives.) Next, go over to the /usr/src directory and, as root, unzip and untar this source file you have just downloaded. After untarring the files, your directories should look very close to this:

total 113
drwxr-xr-x   6 root     root         1024 Feb 10 02:22 .
drwxr-xr-x  18 root     root         1024 Feb 10 03:38 ..
-rw-r--r--   1 root     root       107229 Aug  3  1995 ftape-2.03b.tar.gz
drwxr-xr-x   2 root     root         1024 Feb 10 02:11 ghostscript-2.6.2
lrwxrwxrwx   1 root     root           12 Feb 10 02:20 linux -> linux-1.3.20
drwxr-xr-x  14 root     root         1024 Feb 10 13:21 linux-1.3.20
drwxr-xr-x   3 root     root         1024 Feb 10 02:23 sendmail
drwxr-xr-x   6 root     users        2048 Feb 10 02:12 term-2.3.5

Now confirm that your gcc is installed correctly. Type the following command to see whether the version number is correct. The ensuing version from the following command should be 2.7.0 or higher. If you downloaded the binaries from, the version will be 2.7.2:

$ gcc -v
Reading specs from /usr/lib/gcc-lib/i486-linux/2.7.0/specs
gcc version 2.7.0

Now create the kernel using the following commands, in this order:

cd /usr/src/linux
make depend
make config
make clean
make zlilo

Depending on the type of pc you have, each command will take some time-so be prepared to wait. After you have made the kernel, copy it to the root. Then run setup to reinstall LILO and point to this new kernel as the boot kernel. Next, reboot.

After rebooting, you can always use the version number check with the gcc -v command. Then check version numbers in the /lib directory. The version numbers in your system may be higher, but you should have at least the version numbers listed below for the libc and libdl files:

-rwxr-xr-x   2 24580 Feb 22 22:44
lrwxrwxrwx   1 13 Feb 10 02:15 ->

-rwxr-xr-x   1 634880 Apr 29  1995
-rwxr-xr-x   2     558615 Feb 22 22:46
-rwxr-xr-x   1 562683 May 18  1995
-rwxr-xr-x   2 558615 Feb 22 22:46
lrwxrwxrwx   1 18 Feb 10 02:15 ->
-rwxr-xr-x   1 49152 Dec 13  1994
lrwxrwxrwx   1 18 Feb 10 02:09 ->
-rwxr-xr-x   1 52641 May 18  1995
lrwxrwxrwx   1 15 Feb 22 22:44 -> [icc]
lrwxrwxrwx   1 15 Feb 22 22:44 ->
-rwxr-xr-x   1 6983 Feb 22 22:44
-rwxr-xr-x   1 6843 Aug 15  1995
lrwxrwxrwx   1 14 Feb 10 02:15 ->
-rwxr-xr-x   1 110592 Dec 13  1994
lrwxrwxrwx   1 13 Feb 26 07:59 ->
-rwxr-xr-x   1 35942 May 18  1995
-rwxr-xr-x   1 34449 Feb 22 22:46[ PARTIAL LIST ONLY]

In this list, notice how the links are created from minor version numbers, to version numbers, and then finally to the actual version number. It's important to keep the *.so files present for those files that require the .so extension.

Also note the types of files now in the /lib directory. You'll notice both the ELF and a.out files in there. The ELF kernel is capable of loading both ELF and a.out shared libraries. Actually, working with ELF binaries does make everything a bit slower, but the advantage gained by loadable modules is worth the loss in speed.

Once your system is up to ELF, run the Configure script for Perl. Make sure you answer yes to the question about dynamic loading. Then, for the type of dynamic loading, select the dl_open option. The rest of the questions are quite easy and should work with defaults. The Configure script will even run the make file for you to create Perl. Run make from within Configure when prompted at the end of the script.

After the Perl files are built, run the make test command. There should be no errors once the dust settles. If there are errors, your ELF distribution is not complete, and you should look at the errors when creating the make file and binaries in the /usr/src/linux tree. It will be a long night. A few files that will cause you grief are the GDBM* library files. The sources for these files are in the d package set for Linux. Copy these source files to /usr/src/linux. Compile and install one file at a time until the errors disappear.

I actually have found that the Perl installation test has been the most comprehensive test for checking my Linux ELF system to see whether everything is there. Until someone comes up with a better test, you may want to use this as your benchmark, too.

Partial Upgrades of Your Linux Kernel for Getting Perl Running

Linux source files are really not meant to be updated to ELF piece by piece. The note in the file UPGRADE.TXT points to that. Here's what the note says:

24-Aug-95 NOTE: Trying to upgrade to ELF Slackware from a.out Slackware will
undoubtedly cause you all kinds of problems. Don't do it.

Believe this message. The details of doing a partial upgrade are listed in the file ELF-HOWTO. However, the warning from the author of this note did hold. I followed the instructions in the HOWTO file to the letter…well, almost.

Even after getting the requisite binaries from the FTP sites listed in the HOWTO files, the programs that were supposed to work, didn't. Either the binaries were incompatible or I was missing files, or both. Modifying the LDPATH environment variable led to all sorts of problems.

The Real Danger During the Partial Upgrade

During this partial upgrade process, I accidentally deleted the /lib/ file. This really big error alone rendered my entire Linux system useless because just about every command in the Linux /bin tree uses this /lib/ file for execution. Not good.

The only way to recover from this was to perform the following steps:

  1. Place a Linux floppy in drive A. (If you do not have a boot disk, make one that works with your system and allows you to access the CD-ROM drive on your machine.)
  2. Do a hard reset.
  3. Mount the hard drive from the # prompt.
  4. Copy* to the /lib directory from the /live/lib directory on your Linux CD-ROM.
  5. Do a hard reset after the copy is over. Do not do a shutdown command.
  6. Remove the floppy disk during the reset process.

This is important enough to remember:
Do not remove the /lib/ file. Ever.


This chapter walked you through the installation steps for getting Perl on your system. Hopefully, things went smoothly and you did not have to resort to editing script files. However, in this chapter you should have gleaned enough information to be able to fix anything that is broken.