***********************************************************************
* Perl for GUARDIAN *
***********************************************************************
FORWARD
-------
Welcome to Perl (Version 5.004) for Guardian. This is version 2 of the
Guardian Port.
This file contains information you will need to:
* Run Perl on your Guardian NSK system
* Compile Perl on your Guardian NSK system (should that be necessary)
* Install Perl on your Guardian NSK system (jump to end of document)
Take a look at the file whatsnew.txt for a description of the
changes since the last upload.
This port has only been running on a D35 version of Guardian and has been
tested on G06 (using the D44 C compiler & the G06 NMC compiler). If the
executable file doesn't run on your system, try recompiling (using the
supplied instructions). If it still doesn't run, give me a call and maybe
I can help you figure out what to do next.
This document contains the following sections:
* INTRODUCTION
Some remarks on why I tried to do this port (_why_ indeed?)
* FILE ORGANIZATION
How the Perl subdirectories got mapped into Guardian file names
* LIBRARIES
How libraries work under Guardian
* HOW TO USE PERL ON GUARDIAN
An explanation of the command line
* STUFF THAT APPEARS TO WORK
A discussion of what works
* STUFF THAT WON'T WORK
A discussion of what doesn't (or can't ever) work.
* PERFORMANCE
A brief remark on performance.
* EXTENDING PERL FOR GUARDIAN WITH C ROUTINES
A pointer to another text file to read
* TEST SUITE RESULTS
A discussion of what happened when I ran the validation suite
* COMPILATION INSTRUCTIONS
Some tools for recompiling and re-binding Perl for Guardian
* TACL SUPPORT
An explanation of the TACL macros
* LICENSING
This describes the LICENSE (it points you at the LICENSE file).
* INSTALLATION INSTRUCTIONS
How to install this beast.
* README DOCUMENT CHANGE HISTORY
A list of changes made to this document and the software over time.
If you make any changes to the software please update this document
to reflect them.
First, a caveat: I'm not a UNIX person and only a moderately experienced C
programmer (most of my work has been in TAL). So the choices I made about
what to get working and the approach I took for getting it working may seem
odd or idiosyncratic or even just plain wrong. So, I am very open to
suggestions about how to improve this port. Please feel free to call or
write with any ideas or thoughts you might have.
Enjoy!
Carl Adler
IDX, Systems Corp.
Seattle, Washington
(206) 689-1479
[email protected]
(206) 542-6291
[email protected]
INTRODUCTION
------------
This first bit of verbiage is a discussion of why we are trying
to do this port and some of the peculiarities associated with
implementing Perl on Guardian.
Perl was built to help people do interesting work on UNIX
platforms, primarily related to system managment. It has been
successfully ported to a number of other platforms which resemble
UNIX in several important aspects:
* The file system is tree structured
* Process creation is not a capital crime
* Most of the UNIX utilities or something like them have
already been ported.
Gurdian does not resemble UNIX in any of these aspects. So what
did we hope to gain by trying to do this port?
* TACL doesn't do regular expressions; Perl does.
* Many people know Perl; not many know TACL.
* There are books you can buy to help you learn Perl. There
are no books on TACL that I'm aware of.
* For people who know C, Perl is probably easier to learn
than TACL.
* With a little luck, any Perl scripts generated for our use
on Guardian, will be useful on other platforms. It's a
cinch that our TACL macros won't be useful.
* There are a ton of Perl scripts available in the world that might
be usable to people on a Tandem (obviously, those that are used
to help manage UNIX systems won't be).
FILE ORGANIZATION
-----------------
On tree structured file systems, Perl is usually in something
like:
/usr/perl/...
and there is a tree under this with "bin", "lib", etc.
Here's what we did:
PERL is currently installed on volume $SYS2 and resides in
a set of subvols, each of whose names is prefixed with "PERL":
PERL PERLO executeable, TACL library, PERLLOCL (config &
init) PERL bootstrap macro and this file.
PERLMNT Perl Maintanence macros
PERLSRC Source code (program and header files)
PERLOBJ Object code (one per source compile)
PERLL01 - PERLL09 Library directories (see the AAREADME file in
each subvol for it's original UNIX name).
PERLT01 - PERLT07 /t directories (see the README file in each
subvol for it's orignal UNIX name).
PERLLIB - Libraries added by implementer and site
PERLEXT - This is where we've stashed code to implement our
extensions (at the time of this writing there are two).
PERLEXTT - PERL scripts we used to test our extensions
LIBRARIES
---------
Starting with the 10Sep1998 version of this port, the way we
handle libraries has changed.
You no longer need to specifically have a define for each library
file. You only need defines for sub-volumes. In particular:
=PERL_LIBRARY, CLASS SEARCH, SUBVOL0(...)
=PERL_LIBRARY_FILE, CLASS SEARCH, SUBVOL1(...)
=PERL_LIBRARY_TEXT, CLASS SEARCH, SUBVOL2(...)
=PERL_LIBRARY_MATH, CLASS SEARCH, SUBVOL3(...)
=PERL_LIBRARY_GETOPT, CLASS SEARCH, SUBVOL4(...)
=PERL_LIBRARY_SYS, CLASS SEARCH, SUBVOL5(...)
=PERL_LIBRARY_SEARCH, CLASS SEARCH, SUBVOL6(...)
=PERL_LIBRARY_TIME, CLASS SEARCH, SUBVOL7(...)
=PERL_LIBRARY_TIE, CLASS SEARCH, SUBVOL8(...)
=PERL_LIBRARY_LOCAL, CLASS SEARCH, SUBVOL9(...)
=PERL_LIBRARY_CSTM, CLASS SEARCH, SUBVOL10(...)
=PERL_LIBRARY_EXT, CLASS SEARCH, SUBVOL11(...)
where "..." is a list of subvols that hold the library files.
** Library Sub-volumes.
In the distribution, there are 9 sub-volumes: PERLL01 thru PERLL09
which map to various library sub-directories in the original CPAN
distribution. The PERLL01 sub-volume maps to the top level "LIB"
dictory. The rest map to lower level directies.
We have added a sub-volume called PERLLIB for IDX or Guardian
specific libraries.
=PERL_LIBRARY_LOCAL always maps to PERLLIB.
** Search DEFINES.
Each one of these defines maps to one of these sub-volumes (via
the contents of the "...". You can put home-grown libraries that
you want to make available to everyone on the system in
PERLLIB (and map that subvol to "..._LOCAL". Individual user
libraries can be mapped to any subvolume desired via the
"..._CSTM" define (these work more or less like TACLLOCL and
TACLCSTM).
** How it Works.
In your PERL script you could say:
use strict;
and internally, PERL for Guardian will use the =PERL_LIBRARY
search define to find the file: $vol.PERLL01.STRICT.
For portability reasons, we have to be able to support, for example:
use File::Find;
In the UNIX world: "File::Find" maps to "./File/Find.pm"
which is a relative path from the current directory (note that
through the @INC mechanism, UNIX and DOS versions of perl
accomplish what we're doing here via SEARCH defines. When
PERL for Guardian sees this usage, it will strip the "file"
and append it to the main LIBRARY define, creating the define:
=PERL_LIBRARY_FILE
PERL for Guardian then uses the =PERL_LIBRARY_FILE search define
to locate $vol.PERLL02.FIND. ONLY two levels are supported
in PERL for Guardian. That is:
use foo::bar::zot;
is equivalent to:
use bar::zot;
Internally, PERL converts:
"strict" and "File::Find"
to:
"strict%" and "File/Find%"
which acts as a trigger for invoking these mechansims.
In the stock PERL for Guardian distribution the define
=PERL_LIBRARY lists every library subvol as search subvols. Thus
You never have to qualifify a library name unless it is duplicated
in which case you will want to qualify it. For example, suppose
you wanted to create an alternate version of "find" for use at
your site. You would put in in the PERLLIB sub-volume and in
your script you would say:
use local::find
and PERL for Guardian would use the library $vol.PERLLIB.FIND
instead of $vol.PERLL01.FIND.
NOTE: The Perl error message: "Can't locate <use file> in @INC"
has been changed to: "Unable to resolve <use file> to a file".
When you get this error it typically means the DEFINE that
was selected by PERL (based on <use file>) does not reference
as sub-volume that contains the file you are trying to USE.
ANOTHER NOTE: The Perl error message (or something like it):
Undefined subroutine &main::find called at foo line 6
occurs when you have a "use" statement that references a library
file that did exist but references package names that don't. Notice
in the previous example where we did (first letter of each is in caps):
use File::Find;
This is because the package name associated with File Finder Library
is:
package File::Find;
While Guardian file names are not case sensitive, package names are.
That is:
use file::find;
will compile and execute correctly, but the subroutine call
find (...);
will not.
** When do the DEFINES get created?
When you execute the PERL macro in PERLMAC, the first thing that
happens is that a macro called PERL.PERLLOCL is executed. This
macro contains all of the defines for the standard PERL library
directories (or at least many of them). PERLMAC maintains a
timestamp which tells it not run the PERLLOCL macro again unless
it has been modified.
The PERL macro will also execute a macro called PERLCSTM if it
finds it in your default sub-volume. This PERLCSTM is a macro
which should contain, among other things:
ADD DEFINE =PERL_LIBRARY_<suffix>, CLASS SEARCH SUBVOL10 <mylib>
where:
<mylib> is a subvolume name where you want to put your own
personal library files.
To access your own library files you would say:
use <suffix>::<file>
where:
<suffix> is some string you invent for the higher level
directory
and:
<file> is the name of a file in the subvolume <mylib>
HOW TO USE PERL ON GUARDIAN
---------------------------
1. Use normal PERL run line syntax. For example:
PERL -w PSCRIPT ARG1 ARG2
Assuming <vol>.PERL has been added to your PMSEARCHLIST, "PERL"
in this case is really a macro that will execute the PERL
interpreter. The first time you execute PERL in any particular
TACL session, the PERL macro will initialize your TACL session
by creating a bunch of variables and defines that make PERL
work correctly on GUARDIAN.
Other macros are available for debugging PERL. Look inside
<vol>.PERL.PERLMAC for details.
STUFF THAT APPEARS TO WORK
--------------------------
In general anything that operates on variables and does no I/O
or system calls appears to work. Some I/O and system calls work:
1. Basic "line oriented" file I/O AND ENSCRIBE I/O works.
Line oriented files are EDIT files and UNSTRUCTURED files.
You must include the "\n" character to terminate a line oriented
record. ENSCRIBE is different. See 1.3 below for details.
1.1 Perl can read and write UNSTRUCTURED files. This means
for example, that, a file of binary data that has been
FTP'd to the Tandem can be processed.
PAY ATTENTION TO THIS:
UNSTRUCTURED files must have ODDUNSTR flag set
in order for Perl to be able to write to them.
Otherwise you will get the following error:
"Not a C file"
You can use either <> or "read" to read UNSTRUCTURED files.
However, suppose $D1.FOO.BAR is a an UNSTRUCTURED file
containing binary (i.e. non-displayable) information. The
following script:
$foo = "\$D1.FOO.BAR";
open F,$foo;
$d = <F>;
will read all of the information from the file $D1.FOO.BAR
into the variable $d until either EOF or until a "\n"
character is found. That is, for "line oriented" files the
<FILE> will read one line and a line is alway terminated with
the "\n" character. NOTE: the "\n" character will be
included in the information in $d. If there is no "\n"
character in a file, <FILE> will read the entire file into
memory. So far we have not seen a limit on how large the file
can be in this case.
1.2 Perl scripts can know what kind of file you are using.
The following script demonstrates this:
stat($ARGV[0]); # get "status" of file
if (!defined -e _) { # does not exist if undefinded
print "$ARGV[0] doesn't exist\n";
exit 0;
}
print "Directory\n" if -d _; # if file was <template>
print "Block Special\n" if -b _; # if file was ENSCRIBE
print "File\n" if -f _; # if file was line oriented
print "Text\n" if -T _; # if content is displayable
print "Binary\n" if -B _; # if content isn't
That is:
-d returns true if the filename is actually a template
-b returns true if the file is an ENSCRIBE file
-f returns true if the file is a line oriented file
-T returns true if the first 512 bytes or the first
ENSCRIBE record contains only displayable data.
-B returns true if the first 512 bytes or the first
ENSCRIBE record contains non-displayable data.
and, if the result of any of these tests is !defined then
the file you are testing doesn't exist.
1.3 Perl can read and write ENSCRIBE files but only through the use
of the following commands:
- sysopen
- sysread
- syswrite
If you try to use "open" on an ENSCRIBE file you'll get an
error (in $!):
"Not a C file"
If you are reading ENSCRIBE files you must specify a length
that is >= the record length of the file or you will truncate
your data. For example:
$file = "\$CUST15.HHDATA.ACCT";
sysopen (ACCT, $file, 0);
$status = sysread (ACCT, $rec, 4000);
The "sysread" will return up to 4000 bytes from an ACCT record.
Also, $rec will contain RAW data. That is, if the record
had binary data in it, then $rec will have it as well.
You must write your scripts knowing the record layout.
To write to ENSCRIBE files use "syswrite". The length
of the string you writing must make sense in terms of the
record length of the file. For example, if the record length
of the ENSCRIBE record was 127, then the variable that you
write should be 127 bytes long.
When writing to ENSCRIBE files, DO NOT include "\n" character
at the end of the string. ENSCRIBE files are record oriented
(1 syswrite = 1 record) and the "\n" will be included in the
data which you probably don't want.
When writing to KEY SEQUENCED files you must make sure that
the primary key is unique. For example, suppose you were
writing to a KEY SEQUENCED file with KEYOFF = 0, KEYLEN = 10.
The first 10 bytes of every variable you write will be the
primary key.
When writing to RELATIVE RECORD files, if you write out a zero
length string, there will not be a record written. That is
if you have say: $foo = "" and then do:
syswrite (OUTFILE, $foo, length($foo));
No record will be writen if OUTFILE is RELATIVE RECORD.
EOF doesn't work on ENSCRIBE files (it always returnes TRUE).
For sysread you can always use "defined" to test the result
of the read (I've doctored up sysread so that even Perl
files return !defined at end of file). Read on for more.
And, if the result of sysread or syswrite is !defined then
an error has occured. $! in a numeric context returns the
error code. $! in a string context returns an error message.
The message is usually somethine like: "Guardian or user
define error <num>". In particular, you should quit your
read/write loops when the result of system is !defined. For
example:
while (defined sysread (FILE, $var, $rdlen)) {
...process $var...
}
if ($! != 1) {
print "Error on FILE: $!\n";
}
Writing is similar.
Errors are the usual Guardian suspects:
21 number of bytes is longer than record
45 file full
43 disk file
1 end of file
etc.
2. OPENDIR/CLOSEDIR/READDIR works but it has some funny Guardian
quirks. Here are some examples that work (that is, READDIR
will actually give you a file name)
a. opendir DIR, $CUST20.F5OBEY
Each readdir will return a file in this directory.
b. opendir DIR, $CUST20.F5OBEY.A*
The mask is expanded and readdir will return an
item from this list.
c. opendir DIR, $CUST20.FAOBEY.A
same as b.
The rule is this: OPENDIR will take whatever you pass it and
try to turn it into a template which is then passed to the
FILE_FINDFIRST_ Guardian procedure call. If you pass something
that doesn't have any wild cards (as in example "a", above),
OPENDIR will append a ".*" to the end of it. OPENDIR will
take "b" verbatum because it has wild cards in it. Example "c"
may look like a file name but OPENDIR needs to pass a template
so it converts "c" into "b". if "A" were really a file name
OPENDIR will still work, but READDIR will return all of the
files in FAOBEY that begin with the letter "A".
NOTE: the PERL run command in the TACL library will expand
templates before running the Perl executable unless the
template is surrounded by quotes. For exmample, suppose
the subvolume $D1.FOO had three files in it: BAR, DRAT, & ZOT.
And, further suppose you entered the following command (REMEMBER,
YOU MUST RUN PERL VIA THE PERLMAC MACRO):
PERL -W SCRIPT FOO.*
The command that actually gets executed is:
PERL -W SCRIPT BAR DRAT ZOT
In particular, note that SCRIPT will never see the reference
to "FOO.*".
You can get a template passed to a PERL script by surrounding it
with double quotes. For example, if you wanted FOO.* to be
seen by SCRIPT, you would enter the command like:
PERL -W SCRIPT "FOO.*"
SCRIPT can access this template as $ARGV[0].
3. SYSTEM works. What you do with it is execute TACL commands.
The C Runtime starts up a TACL the first time you use it and
this TACL is yours and yours only. This means you can do
things like:
system ('#push #out');
system ('#set #out $s.#junk');
system ('fileinfo $d1.junk.*');
The fileinfo will generate a list of the files in $d1.junk to
the spooler location mentioned. Note the use of single quotes
to prevent PERL from treating $s and $d1 as variables.
4. CHDIR seems to work (see PERLL02.FIND).
5. While(<>) seems to work as long as you don't try to pass it
any templates (this is why PERL command line expands templates
before running the PERL program).
6. Command switches that seem to work:
-w will display warnings
-h will display help
-V will display configuration (not that it tells you much)
-e but be careful because the command line parsing isn't
intutively obvious. Here's an example, this doesn't work
as you'd expect:
perl -e"$a=~1"; -e"print ""$a\n""";
but this version of the same thing does:
perl -e"$a = ~1"; -e"print ""$a\n""";
7. Requester/Server scripts. Perl scripts can act as either servers
or requesters.
A Perl script acting as a server accepts messages on $RECEIVE,
does something, and the replies. A perl script acting as a
requester, "prompts" a server for some work and accepts its reply
all in a single read request (Guardian WRITEREAD).
To "enable" reqeuster/server behavior, a script must open $RECEIVE
or a process using "sysopen" with "MODE=3".
The following is a simple server:
sysopen (R, "\$RECEIVE", 3);
while (defined sysread (R, $s, 200)){
print "Request: $s\n";
$s = "Ok";
syswrite (R, $s, length($s));
}
print "$!\n";
It opens $RECEIVE with mode = 3. Then until an error occurs
(which isn't likely) the script reads $RECEIVE, prints what it
received, and then responds with "Ok". This script will run
forever until you stop the PERL processor.
IMPORTANT NOTE: When you want Perl to execute a server script
Perl must itself be running as a named process. That is:
TACL > perl/name $ptzzz/smplsvr
This is so that requesters know where to send their messages.
The following is a simple requester:
sysopen (S, "\$PTZZZ", 3);
$s = ">";
sysread (S, $s, 200);
print $s;
This script opens a process named $PTZZZ (e.g. the Perl processor
running the server script above). It then sends a ">" as a prompt
to the process, gets back it's response, prints the response and
then terminates. Note that when "mode = 3" "sysread" works like
the COBOL verb: READ WITH PROMPT with the contents of buffer
containing the prompt and being replaced with the response.
Perl scripts can also do "simplex" (i.e. one-way) output to
processes and can do "simplex" (i.e. one-way) reads on $RECEIVE.
In this case, you use SYSOPEN with mode = 0 (for read) or 1
(for write).
[ed note: someday, when we're bored, we might try to implement
PATHSENDS, but don't hold your breath].
8. Perl can directly access the following Guardian system calls:
Perl Call System Call
--------- -----------
GUARDEXT::aborttmf ABORTTRANSACTION
GUARDEXT::begintmf BEGINTRANSACTION
GUARDEXT::endtmf ENDTRANSACTION
GUARDEXT::control CONTROL
GUARDEXT::setmode SETMODE
GUARDEXT::position POSITITION
GUARDEXT::keypos KEYPOSITIONX
In every case the parameters and return values are as
described in the Guardian Procedure Calls Manual. The following
script illustrates how to use the keypos routine to position
within a KEYSEQUENCED file before reading.
1 sysopen (HTLOG, "\$CUST15.HHDATA.HTLOG", 0)
or die "Couldn't open HTLOG: $!\n";
2 $fnum = fileno(HTLOG);
3 $keyval = "9802230800";
4 $keyspec = "\x00\x00";
5 $len = length $keyval;
6 $len = ($len << 8) | $len;
7 $pmode = 0;
8 $status = GUARDEXT::keypos($fnum, $keyval, $keyspec, $len,
$pmode);
9 print "$status\n";
10 sysread (HTLOG, $logrec, 300);
11 print "$logrec\n";
Line 1 opens an HTLOG file using sysopen.
line 2 gets the file number for that file
line 3 sets the date/time to start at
line 4 sets the key to "primary key"
line 5 & 6 sets up the length word (left half = righ half) as
described in the Procedure calls manual.
line 7 sets the positioning mode to approximate.
line 8 does the keypos
line 9 prints the status (1 = Okay, anything else is failure)
line 10 reads the first record at or after the one that has the
the date/time specified
line 11 prints out the record.
9. Libraries and packages seem to work. But there is an anomoly
you need to be careful about. Suppose you were to say:
use file::find;
thinking that you would be able to then say, somewhere in your
script:
find (\&callback, ...);
Because Guardian file names are case INSENSITIVE, PERL will be
able to locate the "find" library in PERLL02 via the
=PERL_LIBRARY_FILE define. So far so good.
However, since the package name is:
File:Find
(note that the first character is upshifted), the use
command never actually found anything to import into the main::
namespace.
So the moral of this story is when you want to "use" packages,
make sure you ask for them EXACTLY like they were declared
keeping case sensitivity in mind even though in terms of
finding the file, Guardian won't care about the case.
10. The MD5 package has been added. Here are the notes from the README
file. See PERLEXTT.MD5T2 for an example of how to use it.
Digest::MD5
---------------------------
This is a Perl extension interface to the RSA Data Security Inc. MD5
and MD2 Message Digest algorithms. This package also provide modules
which calculate HMAC digests.
<notes on installation, removed>
Copyright 1998 Gisle Aas.
Copyright 1995-1996 Neil Winton.
Copyright 1990-1992 RSA Data Security, Inc.
This library is free software; you can redistribute it and/or
modify it under the same terms as Perl itself.
The Digest::MD5 and Digest::MD2 modules are written by Gisle Aas
<
[email protected]>. They are based on MD5-1.7 made by Neil Winton. The
Digest::HMAC module is written by Graham Barr and Gisle Aas.
11. `cmd` doesn't work, but we've implemented a module called BACKTICK
which provides almost the same functionality:
SYNTAX: backtick ("<TACL command>")
Where:
<TACL command>
This argument is Required. It can be any valid command that can
be issued from the TACL prompt.
Description:
Backtick is a Perl function that allows a Perl program to issue
TACL commands. It sends a command, and returns the output from that
command. The output is returned as a reference to an array.
Usage:
The statement:
use backtick;
must be present in the Perl code using backtick, prior to any
backtick commands.
Output and Errors:
If there is an error in the TACL command, let's say the command
used was "foop info myfile", then backtick will return the error
message produced by TACL.
If there was no output produced from TACL, let's say the command
used was "volume $d1.myvol", then the returned output will be defined,
but blank.
If there was an error in backtick itself, let's say it can not open
TACL, then the returned output will be undefined.
Limitations:
Backtick will not work if Perl is being used as a server (see
the stuff previously about sysopen mode 3 I/O).
Efficiency concerns:
The first time backtick is used in a Perl program, it needs to
create and open a TACL session, which is significant overhead.
However, only the first backtick command in a Perl program has
this overhead, all other backtick commands do not need to start
TACL.
Example:
# Perl program that uses backtick
use backtick;
# print out the results of a FUP INFO command
print backtick ("fup info $data1.foo.*");
# capture the results of a TACL #output command
my $array_ref = backtick ("#output foobar");
# and print out the captured results
foreach (@$array_ref) {print $_, "\n";}
STUFF THAT WON'T WORK
---------------------
This list isn't fully enumerated yet. Please tell me if you try
something that you thought might work and doesn't.
In general, anything that appears to be a UNIX kind of command
like fork(), exec(), mkdir(), etc., won't work.
seek() Doesn't work.
PIPES don't work.
TCP/IP stuff doesn't work (although we could, in principle, make
them work).
File name globbing doesn't work (e.g. while(<FOO.*>)
File tests that ask questions about ownership, executability and
other OS dependent stuff don't work (see the discussion above
about file I/O for a list of the file tests that do work).
The @{<expression>} construct causes perl to crash. I don't know why
yet.
You can redirect STDOUT and STDERR to a disk file inside a perl
perl program but once you do, you can't switch back.
Printing arrays directly does not give you the results that you'd expect.
For example if @a contains 3 elements: "EL1", "EL2", and "EL3". Then
print @a generates "EL1EL2EL3" with maybe some garbage characters
interspersed. In general when you print arrays, you should do it from
within a foreach loop.
Command line switches that don't work:
-d A bunch of errors suggesting that perl is doing very
UNIXy kind of stuff
See below for details on the results from the suite of tests.
PERFORMANCE
-----------
Input seems acceptable. I was able to load a 13,300 line file into a
variable in about 4 + or - 1 seconds using:
undef $/;
$d = <INFILE>;
$/ = '\n';
I haven't done anything to optimize Output so it's a little slow.
Regular expression evaluation looks like it's about 25% to 50%
slower than on DOS.
If you notice anything else, let me know.
EXTENDING PERL FOR GUARDIAN WITH C ROUTINES
-------------------------------------------
Please read the file HOWTOEXT which is part of this release.
TEST SUITE RESULTS
------------------
The following summarizes the results of running the programs in
the /perl/t/... directories.
In general, they support my supposition that anything that perl does
internally works and only a few things that it does with the
outside world work. It turns out, that Perl is crashing either
during compilation or execution of 17 of the many dozens of test.
It also turns out that many of the tests fail because of hard coded
UNIX syntax file names. Sometimes (I only tried a few) by modifying
the names to GUARDIAN syntax I could get the scripts to run. If
I had to modify hard coded file names and, after doing so, the script
ran, I called it successful.
The other thing the test suite does, a lot, is add a hard coded
library reference to @INC. I had to eliminate those assignments
to get many of the scripts to run. This suggests that any scripts
that depend on being able to look in multiple places for a file
will fail.
The following is the "directory" listing for all of the test scripts.
You'll notice some of the names have been munged to 8 characters:
perl5.004/t/README
perl5.004/t/TEST
perl5.004/t/base/cond.t
perl5.004/t/base/if.t
perl5.004/t/base/lex.t
perl5.004/t/base/pat.t
perl5.004/t/base/term.t
perl5.004/t/cmd/elsif.t
perl5.004/t/cmd/for.t
perl5.004/t/cmd/mod.t
perl5.004/t/cmd/subval.t
perl5.004/t/cmd/switch.t
perl5.004/t/cmd/while.t
perl5.004/t/comp/cmdopt.t
perl5.004/t/comp/colon.t
perl5.004/t/comp/cpp.aux
perl5.004/t/comp/cpp.t
perl5.004/t/comp/decl.t
perl5.004/t/comp/multiline.t
perl5.004/t/comp/package.t
perl5.004/t/comp/proto.t
perl5.004/t/comp/redef.t
perl5.004/t/comp/script.t
perl5.004/t/comp/term.t
perl5.004/t/comp/use.t
perl5.004/t/harness
perl5.004/t/io/argv.t
perl5.004/t/io/dup.t
perl5.004/t/io/fs.t
perl5.004/t/io/inplace.t
perl5.004/t/io/pipe.t
perl5.004/t/io/print.t
perl5.004/t/io/read.t
perl5.004/t/io/tell.t
perl5.004/t/lib/abbrev.t
perl5.004/t/lib/anydbm.t
perl5.004/t/lib/autoloader.t
perl5.004/t/lib/basename.t
perl5.004/t/lib/bigint.t
perl5.004/t/lib/bigintpm.t
perl5.004/t/lib/checktree.t
perl5.004/t/lib/complex.t
perl5.004/t/lib/db-btree.t
perl5.004/t/lib/db-hash.t
perl5.004/t/lib/db-recno.t
perl5.004/t/lib/dirhand.t
perl5.004/t/lib/english.t
perl5.004/t/lib/env.t
perl5.004/t/lib/filecache.t
perl5.004/t/lib/filecopy.t
perl5.004/t/lib/filefind.t
perl5.004/t/lib/filehand.t
perl5.004/t/lib/filepath.t
perl5.004/t/lib/findbin.t
perl5.004/t/lib/gdbm.t
perl5.004/t/lib/getopt.t
perl5.004/t/lib/hostname.t
perl5.004/t/lib/io_dup.t
perl5.004/t/lib/io_pipe.t
perl5.004/t/lib/io_sel.t
perl5.004/t/lib/io_sock.t
perl5.004/t/lib/io_taint.t
perl5.004/t/lib/io_tell.t
perl5.004/t/lib/io_udp.t
perl5.004/t/lib/io_xs.t
perl5.004/t/lib/ndbm.t
perl5.004/t/lib/odbm.t
perl5.004/t/lib/opcode.t
perl5.004/t/lib/open2.t
perl5.004/t/lib/open3.t
perl5.004/t/lib/ops.t
perl5.004/t/lib/parsewords.t
perl5.004/t/lib/posix.t
perl5.004/t/lib/safe1.t
perl5.004/t/lib/safe2.t
perl5.004/t/lib/sdbm.t
perl5.004/t/lib/searchdict.t
perl5.004/t/lib/selectsaver.t
perl5.004/t/lib/socket.t
perl5.004/t/lib/soundex.t
perl5.004/t/lib/symbol.t
perl5.004/t/lib/texttabs.t
perl5.004/t/lib/textwrap.t
perl5.004/t/lib/timelocal.t
perl5.004/t/lib/trig.t
perl5.004/t/op/append.t
perl5.004/t/op/arith.t
perl5.004/t/op/array.t
perl5.004/t/op/assignwarn.t
perl5.004/t/op/auto.t
perl5.004/t/op/bop.t
perl5.004/t/op/chop.t
perl5.004/t/op/closure.t
perl5.004/t/op/cmp.t
perl5.004/t/op/cond.t
perl5.004/t/op/delete.t
perl5.004/t/op/do.t
perl5.004/t/op/each.t
perl5.004/t/op/eval.t
perl5.004/t/op/exec.t
perl5.004/t/op/exp.t
perl5.004/t/op/flip.t
perl5.004/t/op/fork.t
perl5.004/t/op/glob.t
perl5.004/t/op/goto.t
perl5.004/t/op/groups.t
perl5.004/t/op/gv.t
perl5.004/t/op/inc.t
perl5.004/t/op/index.t
perl5.004/t/op/int.t
perl5.004/t/op/join.t
perl5.004/t/op/list.t
perl5.004/t/op/local.t
perl5.004/t/op/magic.t
perl5.004/t/op/method.t
perl5.004/t/op/misc.t
perl5.004/t/op/mkdir.t
perl5.004/t/op/my.t
perl5.004/t/op/oct.t
perl5.004/t/op/ord.t
perl5.004/t/op/pack.t
perl5.004/t/op/pat.t
perl5.004/t/op/push.t
perl5.004/t/op/quotemeta.t
perl5.004/t/op/rand.t
perl5.004/t/op/range.t
perl5.004/t/op/re_tests
perl5.004/t/op/read.t
perl5.004/t/op/readdir.t
perl5.004/t/op/recurse.t
perl5.004/t/op/ref.t
perl5.004/t/op/regexp.t
perl5.004/t/op/repeat.t
perl5.004/t/op/runlevel.t
perl5.004/t/op/sleep.t
perl5.004/t/op/sort.t
perl5.004/t/op/split.t
perl5.004/t/op/sprintf.t
perl5.004/t/op/stat.t
perl5.004/t/op/study.t
perl5.004/t/op/subst.t
perl5.004/t/op/substr.t
perl5.004/t/op/sysio.t
perl5.004/t/op/taint.t
perl5.004/t/op/tie.t
perl5.004/t/op/time.t
perl5.004/t/op/undef.t
perl5.004/t/op/universal.t
perl5.004/t/op/unshift.t
perl5.004/t/op/vec.t
perl5.004/t/op/write.t
perl5.004/t/pragma/constant.t
perl5.004/t/pragma/locale.t
perl5.004/t/pragma/overload.t
perl5.004/t/pragma/strict-refs
perl5.004/t/pragma/strict-subs
perl5.004/t/pragma/strict-vars
perl5.004/t/pragma/strict.t
perl5.004/t/pragma/subs.t
perl5.004/t/pragma/warn-1global
perl5.004/t/pragma/warning.t
The following is a summary of the results of each of the above tests
An "*" means there is additional information available following the
summary.
/perl/t/base - PERLT01
cond.t passed
if.t passed
*lex.t passed (Fixed 15SEP1997 - see note #4)
pat.t passed
term.t Failed (too much dependence on UNIX)
.
/perl/t/cmd - PERLT02
elsif.t passed
for.t passed
mod.t passed
subval.t passed
while.t passed
/perl/t/comp - PERLT03
cmdopt.t passed
colon.t passed
cppaux.t Failed (Can't emulate -P on #! line at perlt03.cppaux line 1)
decl.t passed
multili.t Failed (backtick)
package.t passed
proto.t passed
redef.t passed
script.t Failed (backtick)
term.t passed
use.t Failed (eval "use lib 1.01" & similar statements)
/perl/t/io - PERLT04
argv.t Failed (backtick)
dup.t Failed (open '>&xxx')
fs.t Failed (backtick, link, chmod, stat, rmdir, mkdir)
inplace.t Failed (backtick, inplace edits)
pipe.t Failed (fork not implemented)
read.t Failed (open '+>xxx' and seek)
tell.t Failed (seek, maybe tell)
/perl/t/lib - PERLT05
abbrev.t passed
anydbm.t Failed (DBM package not installed)
autoloa.t Failed (mkdir required, might work otherwise)
basenam.t passed
bigintp.t Failed (40 of 287 tests failed)
checktr.t Failed (depends on UNIX or UNIX like file system)
*complex.t passed (fixed 19SEP1997, see note #6)
dbbtree.t Failed (ext/ not installed)
dbhash.t Failed (ext/ not installed)
dbrecno.t Failed (ext/ not installed)
dirhand.t Failed (object creation)
filepat.t Failed (mkdir not supported)
findbin.t Failed (Cannot chdir to $D1.ADLERPT/ -- file name spec)
gdbm.t Failed (ext/ not installed)
*getopt.t passed (fixed 24SEP1997, see note #7)
hostnam.t Failed (hostname not implemented)
iodup.t Failed (/ext not installed)
iopipe.t Failed (/ext not installed)
iosel.t Failed (/ext not installed)
iosock.t Failed (/ext not installed)
iotaint.t Failed (/ext not installed)
iotell.t Failed (/ext not installed)
ioudp.t Failed (/ext not installed)
ioxs.t Failed (/ext not installed)
ndbm.t Failed (/ext not installed)
odbm.t Failed (/ext not installed)
opcode.t Failed (/ext not installed)
open2.t Failed (fork not implemented, /ext not installed)
open3.t Failed (fork not implemented, /ext not installed)
ops.t Failed (/ext not installed)
parsewo.t Failed (thinks subroutine name is a script name)
posix.t Failed (/ext not installed)
safe1.t Failed (/ext not installed)
safe2.t Failed (/ext not installed)
sdbm.t Failed (/ext not installed)
searchd.t Failed (tries to create file w/illegal Guardian name)
selects.t Failed (tries to create file w/illegal Guardian name)
socket.t Failed (TCP/IP not implemented)
soundex.t passed
symbol.t passed
texttab.t passed
textwra.t passed
timeloc.t passed (sort of: there's a wierd library deal going on)
*trig.t passed (Fixed 19SEP1997 see note #6)
/perl/t/op/*.* - PERLT06
append.t passed
arith.t passed
array.t passed (Fixed 19SEP1997 - see note #6)
assignw.t passed
auto.t passed
bop.t passed (Fixed 19SEP1997 - see note #6)
chop.t passed
closure.t Failed (closeure.t looks like it goes into a loop)
cmp.t passed
cond.t passed
delete.t passed
do.t passed
each.t passed
eval.t passed
exec.t Failed (exec not supported)
exp.t passed
flip.t passed
fork.t Failed (fork not supported)
glob.t Failed (file globbing not supported)
goto.t Failed (backtick not supported)
groups.t Failed (Security stuff not implemented)
gv.t passed
inc.t passed (Fixed 19SEP1997 - see note #6)
index.t passed
int.t passed
join.t passed
list.t passed
local.t passed (Fixed 19SEP1997 - see note #6)
magic.t Failed (Use of uninitialized value at perlt06.magict line 28)
method.t passed (this is odd: OO stuff had a problem above)
misc.t Failed (probably, but I'm not sure what real result is)
mkdir.t Failed (mkdir not supported)
my.t passed (Fixed 19SEP1997 - see note #6)
oct.t passed (fixed 18SEP1997 - was crashing during compilation)
ord.t passed
pack.t Failed (some tests failed)
pat.t passed
push.t passed (Fixed 19SEP1997 - see note #6)
quotme.t passed
rand.t passed (it seems, anyway)
range.t passed
readdir.t Failed (19SEP1997 - 1 passed, 2 failed, 3 no output)
read.t Failed (1 test of 4 failed)
recurse.t Failed (some functions never terminate in my lifetime)
ref.t passed
regexp.t Failed (Reg exp work, so I don't know what's going on here)
repeat.t passed
runleve.t Failed (tests failed)
sleep.t Failed (sleep returned incorrect value)
sort.t passed (Fixed 19SEP1997 - see note #6)
split.t passed
sprintf.t passed
stat.t Failed (stat not fully supported)
study.t passed
substr.t passed
subst.t Failed (3 out of 60 tests failed)
sysio.t Failed (50% of the tests failed)
taint.t Failed (Too late for "-T" option at perlt06.taintt line 1.)
tie.t Failed (tests failed)
time.t Failed (times not implemented at perlt06.timet line 8.)
undef.t passed
univers.t passed
unshift.t passed
vec.t passed
write.t Failed (backtick not implemented)
/perl/t/pragma/*.* - PERLT07
constan.t passed (Fixed 19SEP1997 - see note #6)
locale.t Failed (Too late for "-T" option at perlt07.localet line 1.)
overloa.t Failed (4 out of 115 tests failed)
strict.t Failed (no output)
subs.t Failed (Tests failed)
warning.t Failed (no output)
NOTES on tests:
1. /perl/t/base/lex.t (01SEP1997)
This failed when PERL tried parse a 10 digit number. The crash
resulted from an arithmetic overflow. Evidently, other C runtimes
will allow this to occur without any disasters.
So, if you get an arithmetic overflow with any of your scripts, it's
probably because a value has exceeded the largest value possible for
a 32 bit integer. For example, this will fail:
$foo = int ($bar); # where $bar is a float > 2147483647.0
2. /perl/t/lib/complex.t (03SEP1997)
/perl/t/lib/trig.t
Perl crashes during compilation when it tries to compile one or another
of the following:
use constant pi => 4 * atan2(1, 1);
use constant log10inv => 1 / log(10);
The failure is an arithmetic overflow in sv_upgrade. The double
4294967293.
is being cast into a 32 bit integer which C can't do because the
maximum integer value is 2147483647. We don't know where this
number comes from.
This means that you can't use the trig or complex number functions
from the library.
3. /perl/t/lib/getopt.t (03SEP1997)
This is similar to #2 above. The crash is an arithmetic overflow
during a cast from double to int. The location in the code is the
exact same place as #2 and the number that is being dealt with is
almost the same:
4294965757.
instead of:
4294967293.
Both of these numbers, when cast as (long long) end up looking like
negative 32 bit integers. In the first case it's FFFFF9FD or -1539
and in the second it looks like FFFFFFFD or -3.
This suggests that some arithmetic is going wrong somewhere or maybe
an ABS function. Too soon to speculate.
4. More on #2 and #3 above: (15SEP1997)
Here's a reliable case that causes the exact same crash
# t23 - complement test
#
#
$b = 17;
$b &= ~1;
print "$a, $b\n";
Evidently, Perl is treating ~1 as ULONG_MAX - 1 instead of -2. And
because ULONG_MAX is > LONG_MAX Perl converts this thing to a double.
Well that's okay until it has to convert it back to a long. That's
when we get into trouble. We can't convert directly from double to
long. We have to convert from double to unsigned long to long.
This has been fixed by changing the macros: I_32 and I_V as follows:
#ifdef CASTI32
# ifndef _GUARDIAN_TARGET
# define I_32(what) ((I32)(what))
# define I_V(what) ((IV)(what))
# else
# define I_32(what) ((I32)(UV)(what))
# define I_V(what) ((IV)(UV)(what))
# endif
#else
# etc. etc.
That is, instead of going directly from double to long we cast
to unsigned long first.
5. /perl/t/lib/complex.t (18SEP1997)
/perl/t/lib/trig.t
/perl/t/lib/getopt.t
See note #4 before reading this one.
After fixing the problem with "~" (complement), Perl is now crashing
during execution with the following error in both tests.
\PHAMIS.$:5:242:411332716 - Invalid function parameter
\PHAMIS.$:5:242:411332716 - From Perl_pp_padav + %246, UC.01
6. /perl/t/op/array.t (18SEP1997)
See note #3 before reading this one.
After fixing the arithmetic overflow problem, we are now able to
start executing the test. But now, Perl crashes during execution.
*** Run-time Error 040 ***
Invalid function parameter
From Perl_pp_rv2av + %1052, UC.02
Perl_runops + %143, UC.02
perl_run + %546, UC.00
main + %57, UC.00
_MAIN + %33, UC.00
There are two things to pay attention to here: first, this error is
occured in the following bit of script:
@foo = ();
$r = join(',', $#foo, @foo);
And went away when I added:
@foo = (1, 2, 3);
as the first statement. Then the test passed completely. Then,
secondly, when I removed the first line from the script, it worked
perfectly.
Go figure.
In the debugger the problem was occuring in the following
C statement:
Copy(AvARRAY(av), SP+1, maxarg, SV*);
which invokes the macro:
(void)memcpy((char*)(d),(char*)(s), (n) * sizeof(t))
AvARRAY(sv) is equal to 0, which is why memcpy is complaining.
When I tried this same code in the WIN32 version, there wasn't any
problem.
So, here's what I suspect:
Perl doesn't know that the array it's moving is empty and it doesn't
know that the destination doesn't really exist yet. So it happily
goes off to move 0 bytes into outerspace. Evidently in the
Guardian C compiler if outerspace is NULL then we get a run-time
error.
I think the fix is to change the Copy MACRO so that if the length
is zero, it doesn't try to execute memcpy.
Postscript to #6: I implmented the change to the copy MACRO and
that eliminated every single one of the crashes. A new symptom
showed up that is related to how I tortered the GETS routine
into performing.
7. /perl/t/lib/getopt.t (19SEP1997)
The test ran to completion, but when the script tried to close a
file:
*** Run-time Error 034 ***
Released space not allocated
From GUARDIAN_FCLOSE + %21, UC.03
PerlIO_close + %5, UC.01
Perl_io_close + %175, UC.00
Perl_do_close + %176, UC.00
Perl_pp_close + %63, UC.02
Perl_runops + %143, UC.02
perl_run + %546, UC.00
main + %57, UC.00
_MAIN + %33, UC.00
Perl is trying to close STDERR twice. It's the second time through
that is causing the error. Oddly, the the first time STDERR is
closed is during execution of pp_open.
The way we redirect STDERR in GUARDIAN C is to use an ASSIGN. I
think what's going on might have something to do with the special
status of STDERR in UNIX.
(21SEP1997)
The runtime library routine, fileno, returns 0, 1, 2 for stdin, stdout
and stderr respectively. these aren't good numbers. So, when perl
starts messing around with STDERR, it tries to do some stuff with
file number 2 and gets into trouble.
MAXSYSFD is a variable that controls (I think) which numbers are
defined as system files (i.e. stdin, stdout and sterr).
So, if we are to support the following construct:
open STDERR, ">foo" or die "stderr: $!\n";
print STDERR "this is an error?\n";
close STDERR;
and have something actually show up in foo, we're going to have to
allow open, close, etc. to work on these wierd files.
(24SEP1997)
Turns out the WIN32 version of perl doesn't quite work here either.
We'll make the GUARDIAN version work like WIN32.
Once you open STDERR (or STDOUT) with a different file, you can't
switch back to where STDERR was originally pointing.
So the above script will work but if you then try to print to
STDERR after the close, you'll get a "write on unopened file"
warning.
COMPILATION INSTRUCTIONS
------------------------
Unless you start messing the source code or if the executable provided
in this distribution doesn't work on your system, you shouldn't have
to re-compile the source. But that's the programmer's kiss of death so
here is what you do if you have to recompile:
A) Complete Recompilation
If you want to do a complete recompilation, purge all of the files in
PERLOBJ (this is where all of the non-executables end up).
There is a cheesy little MAKE macro that will recompile and rebind Perl.
To use it:
1. VOLUME PERLMNT
2. RUN MAKE MKPERL
3. GO TO COFFEE (I live in Seattle so that's obligatory)
It will do what you'd expect a cheesy little MAKE macro to do and either
generates an executable or quits when something goes wrong.
If you are compiling on a system that has NMC it will use it and NLD to
to build the PERLO executable (ignore all of the warnings that are
printed -- you can't turn them off in NMC).
If your system does not have NMC the executable that gets generated is
called PERLUX (meaning PERL unaccellerated). You should accellerate
the code using:
> AXCEL PERLUX PERLO
> BIND STRIP PERLO
On a system with NMC, you can force the use of Guardian C by including:
"COMPILER CISC"
at the beginning of the MKPERL file.
B) Partial Recompilation
In some cases, after you know how the pieces fit together, you can recompile
individual files and just rebind (even though MAKE would recompile more
or even the whole thing).
There is a macro called PERLCOMP that compiles one source file.
> VOLUME PERLMNT
> RUN PERLCOMP <file>
where <file> is the name of the source code file without the C suffix.
PERLCOMP always uses NMC if it's present (you'll have to edit the file
to force it to use Guardian C).
Then when your ready to rebind everything you can do:
> RUN PERLBIND
This does the same bind that the MAKE does. PERLBIND always uses NLD if
it is present. You'll have to edit the file if you want to force the
use of Guardian C.
Again, partial recompilations are not recommended unless you kind of
know what you're doing. Particularly if you're making header file
changes.
TACL SUPPORT
------------
As alluded to above, in many places, we've added a bunch of TACL to
make the use of Perl on Guardian a little more natural than it would
otherwise be.
Aside from the source code maintanence macros mentioned in the previous
section the following TACL macros and libraries are distributed:
PERLLOCL
This is a file which creates all of the defines necessary for Perl
to find its libraries. It uses the TACL builtins ("#" commands)
instead of the ADD DEFINE commands and, as a result, it is very
fast (you can't tell that it exeutes).
PERLLOCL is executed under two different circumstances:
1) when the timestamp changes (i.e. when its been modified)
2) when a "signal" define is missing (i.e. when the defines have
never been loaded).
PERLMAC
This is a TACL library containing, among other things, the PERL
command that runs the PERLO executable. It is responsible for
keeping tabs on the current state of PERLLOCL and, if present
PERLCSTM (which is your own personal PERL wrapper).
When PERLMAC has been loaded, and you enter, for example:
> perl foo bar > outfile
the routines in PERLMAC build a new set of commands:
> SINK [#PURGE OUTFILE]
> PERL/OUT OUTFILE/FOO BAR
As we mentioned previously, these routines will also expand file name
template parameters into a space separated list.
PERL
This is a "bootstrap" macro. It's only job in life is to load PERLMAC
and then re-run the command you used. So, if your #PMSEARCHLIST
inclues the volume and subvolume that contains PERLMAC (and PERLO) the
PERL macro will first load PERLMAC and then re-execute exactly the
same thing you typed. Once the PERLMAC library has been loaded, the
only way you can execute this macro again is by fully qualifying it
because "PERL" is now a variable and variables take precendence over
file names.
BTPROMPT
As you may recall from reading about the stuff that doesn't work,
`cmd` doesn't work on Guardian. I thought about trying to make it
work, but in the end, the method I came up could be done using
Perl itself (along with the "sysio" functions I implemented). So
instead of making `cmd` work, we implemented a module called BACKTICK
(in PERLLIB). One of the components of this TACL macro which changes
the TACL prompt to something very recognizable. See "Stuff that works"
item 11.
As we began to use Guardian PERL for more serious work, we found the
need to add some more TACL level support (It would be a good idea to
review the material in the previous section labeled "LIBRARY" before
trying to read this stuff):
PARAM PERL-DEV
One of the problems that arises because we didn't port @INC
functionality very well is the need for a "private" development
libraries. These are places where developers could put alpha
or beta copies of ".pm" type scripts until they were ready for
prime time.
While the preferred method of handling this problem is using
PERLCSTM, described previously, sometimes you want something
even more "temporary". To that end we added support for the
PERL-DEV parameter. It allows you to specify a temporary
library subvolume where PERL can find test versions of libraries.
For example when you say in your script:
use FOOBAR;
PERL normally expects to find FOOBAR in one of the supported
library sub-volumes. However, when you are just starting out
to develop this module, you don't want anyone else to use it.
So you say:
TACL> PARAM PERL-DEV "<subvol>"
prior to running your PERL script. <subvol> is where FOOBAR
is located. This will add <subvol> to the list of sub-volumes
where PERL looks for library routines.
NOTES:
1. When PERL-DEV is in effect the search list order is:
- distributed modules (for example, strict)
- the PERL-DEV sub-volume
- PERLLIB
So if you are writing a new version of a module that was
in the original distribution, be sure to name it different.
2. the use of PERL-DEV will cause the re-execution of PERLLOCL
upon each execution of PERL. So, you should move your code
to PERLLIB after testing, and then do away with the PERL-DEV
parameter (CLEAR PARAM PERL-DEV).
RUNNING A PERL SCRIPT FROM A MACRO
When you want to execute a PERL script from within a TACL macro
you have to be careful to preserve global TACL variables that
the PERL support macros create and use.
The naive approach is:
?TACL MACRO
#FRAME
do some set-up work
:
perl <your script>
:
do some finish work
#UNFRAME
There is a minor performance hit if you do it this way. To avoid
this there is a routine you can run prior to the #FRAME. The
new macro would look like this:
?TACL MACRO
PERLINIT
#FRAME
do some set-up work
:
perl <your script>
:
do some finish work
#UNFRAME
Notice the invocation of PERLINIT prior to the #FRAME. What's
going on here is PERLINIT is invoking all of the TACL code
that would normally get invoked the first time you run PERL
from a particular TACL session. The TACL variables the PERL
support macros require are created outside the frame your
application macro creates.
Another macro is available that you can run after the #UNFRAME
or in your TACL exception handler code:
PERLKILL
This routine will delete all the global variables and
the =perl* definitions. This causes your TACL session to forget
that PERL ever existed.
As a rule, you don't need to use PERLINIT and PERLKILL if the
macro you are writing is run only once per TACL session (let's
say, as part of a NETBATCH job). If, on the other hand, you are
creating a tool that many people will be using throughout the day
then the use of PERLINIT, in particular, could save lots of
I/O and cycles. As always you've got to think about the
application and how it's likely to be used.
NNC NOTES
---------
The native mode compiler/linker assumes the OSS system calls
and the NMC runtime library both of which are different then
the Guardian C runtime. In particular, the "open" used by NMC
returns a different kind of number than the "open" in the C
runtime.
The original Guardian extension software (GUARD.C) made the
assumption that both "open" and "FILE_OPEN_" returned Guardian
file numbers. This was true enough for the code to work.
Under NMC, "open" returns file numbers starting at 0 as does
"FILE_OPEN_" so it was no longer possible to mix "open" files
with "FILE_OPEN_" files.
The modification we made to support both NMC and Guardian C
was to conceptually divide the file number space into OSS
numbers and GUARDIAN numbers. Whenever a file was opened using
"FILE_OPEN_" we returned the file number with MIN_GUARDIAN_FD
added to it. Whenever we processed any request for data using
a file number we tested the file number against this value.
If the number was less than MIN_GUARDIAN_FD then we used the
OSS system calls (close, read, write). If the number is greater
than or equal to MIN_GUARDIAN_FD then we used Guardian I/O
(READX, WRITEX, REPLYX, etc).
This scheme works regardless of the compiler used to build Perl.
Also, FASTGETS doesn't work with NMC so we disable it if the NMC
compiler is used. This means we always use the runtime GETC which
is a little slower but hopefully, NMC has optimized this a little.
The "CC" function in the PERLMNT.MAKE sub-volume is also modified.
If NMC is present then use it instead of Guardian C. An important
difference between NMC and Guardian C is the error checking level.
In Guardian C, we get away with all kinds of loose constructs
without any complaints from the compiler. However in NMC every
little cotten picking anomolie is picked up and complained about
with WARNING messages. Consequently, we have to ignore
completion codes = 1 for NMC and hope that the warning wasn't
anything we introduced.
Over time, as we work on this code, we should endeavor to fix
the things that NMC complains about. There are a lot of them
so this is sort of a life's work.
LICENSING
---------
Please read the LICENSE file in the PERL subvol. I have attempted to
be complient with all provisions. If I did not it was through my
inability to deal with the legalese and not through any malice.
1. All of the source code that makes up this port is contained in
the PERLSRC sub-volume.
2. If you want a copy of this source code, just ask via the phone numbers
or email addresses listed above.
3. NO WARRENTY: QUOTED FROM THE LICENSE.
11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
REPAIR OR CORRECTION.
12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
INSTALL INSTRUCTIONS
--------------------
****************************************************************************
NOTE: If you already have an earlier version of this port of Perl
for Guardian installed, you must remove it from you system before
trying to install this version. If you placed your own Perl scripts
in any of the pre-defined Perl sub-volumes, please set them aside while
doing this this install.
****************************************************************************
The following install instructions assume you are installing from the
ZIP file and it is on your PC. Also note that I don't talk about security
here at all. Once the Perl distribution is on the Tandem you'll need
to secure the files so that only authorized people can access them.
The archiver/de-archiver referenced in the following discussion is a set
of programs called PAK/UNPAK available from the ITUG library. These
programs use the Guardian programs BACKUP and RESTORE (respectively).
Step 1: Expand the ZIP file (using PKUNZIP or WINZIP or whatever).
PKUNZIP tandem-perl-v2.zip
What you'll end up with are the following Tandem files:
File Code What
--------- ---- -------------------------------------
GPERL 100 A self-extracting archive file
And the following ASCII files:
README.TXT This file
FILELIST.TXT The output from the PAK
WHATSNEW.TXT A short description of new stuff in this release
Step 2: Upload to Tandem.
Upload GPERL a single subvolume. Make sure your FTP is in Binary Mode.
The volume probably should NOT be $vol.subvol.PERL
Step 3: FUP ALTER file codes for GPERL.
FUP ALTER GPERL,CODE 100
Step 4: De-archive the perl files.
RUN GPERL/out <spooler>/ ,*.*.*,VOL <disk volume>,MYID, LISTALL
where: <spooler> is where you want the output to go
<disk volume> is the disk drive where you want PERL to end up
MYID is the RESTORE command to cause files to be created using the
currently logged on user.
LISTALL is the RESTORE command to generate a listing of all of the files
extracted from the archive.
*NOTE*: The comma in front of "*.*.*" is required because the command
tail is passed verbatum to RESTORE following the "<tape>" device
parameter.
Step 5: Run the PLINSTAL macro.
Once you get all of the files that belong to PERL onto the disk, you
need to run the PLINSTAL macro.
The volume and prefix are hard coded in a couple of places. The
PLINSTAL macro figures out what the volume and prefix is and fixes up
those places where it's been hard coded.
Don't freak-out. The hard coding is not any of the actual PERL software.
We hard code the volume and prefix in the PERLMAC TACL library which has
to know where everything is and can't figure it out on the fly (because
it's a library) and we also store the volume and prefix in a file called
PERLPFIX
This file is read by PERLLOCL to figure out where all of the PERL library
scripts are stored (see the section above on libraries).
To run the PLINSTAL macro, make sure you're logged on as someone who can
modify the files in the main PERL subvolume (where the executable is).
Then from whereever you happen to be enter at the TACL prompt:
RUN <vol>.<prefix>.PLINSTAL
Answer the question (Continue? Y/N) and that's it.
ADDITIONAL INSTALLATION NOTES:
------------------------------
PERL must physically reside on a single VOLUME (you can always fuss with
the files if you don't want this to be true, but we recommend you don't
mess around with anything until you've had some experience with this port).
The subvolume names PERL uses are all prefixed with a string 5 chracters
or less. For purposes of this explanation we will assume you are installing
PERL on $SYSTEM and you are using the suggested prefix of "PERL".
This PERL distribution consists of the following subvolumes:
$SYSTEM.PERL
$SYSTEM.PERLEXT
$SYSTEM.PERLEXTT
$SYSTEM.PERLL01
$SYSTEM.PERLL02
$SYSTEM.PERLL03
$SYSTEM.PERLL04
$SYSTEM.PERLL05
$SYSTEM.PERLL06
$SYSTEM.PERLL07
$SYSTEM.PERLL08
$SYSTEM.PERLL09
$SYSTEM.PERLLIB
$SYSTEM.PERLMD5
$SYSTEM.PERLMNT
$SYSTEM.PERLOBJ
$SYSTEM.PERLSRC
$SYSTEM.PERLT01
$SYSTEM.PERLT02
$SYSTEM.PERLT03
$SYSTEM.PERLT04
$SYSTEM.PERLT05
$SYSTEM.PERLT06
$SYSTEM.PERLT07
Again, the volume you choose may be different as might be the prefix.
An aditional sub-volume is distributed: $SYSTEM.PERLPL which contains
application PERL scripts and is not technically part of PERL itself.
These scripts can be anywhere you want. These scripts were developed
locally for a variety of uses.
Since the original sub-volume names are prefixed with "PERL", it is likely
that you won't want to go through the trouble of changing this. However,
if you already have a sub-volume on the target disk with same name as
one in the above list, you'll need to either move it, or move perl to
a temporary disk and then use the RENAME command to change sub-volume names
before moving all of the files to their final home.
The volume and prefix are hard coded in a couple of places. The
PLINSTAL macro figures out what the volume and prefix is and fixes up
those places where it's been hard coded.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
DOCUMENT/SOFTWARE CHANGE HISTORY
--------------------------------
Date Who What
--------- --- ---------------------------------------------------
27Jan1998 CDA Added info about ARRAYS not printing directly via
print @array;
30Jan1998 CDA It is possible to read Type U files files and to
ask Perl whether a file is an edit file or not
<see below for a change>
08Feb1998 CDA It is no longer possible for a script to tell
whether a file is an EDIT file or an UNSTRUCTURED
file. It doesn't need to because both kinds of files
are treated the same way: as "line oriented".
It is now possible to read and write ENSCRIBE files.
A script can ask weather the file is an Enscribe
file or not. Perl can read any kind of
unstructured file but can write to only ODDUNSTR
files.
24Feb1998 CDA Perl scripts can be written as servers or requesters.
Perl scripts have access to the following Guardian
routines: ABORTTRANSACTION, BEGINTRANSACTION,
ENDTRANSACTION, KEYPOSITION, POSITION, CONTROL AND
SETMODE.
24Mar1998 FLC No longer need to load perl.perlmac prior to
running perl. Perlmac will be loaded via the PERL
bootstrap macro if it had not already been loaded
via a previous PERL command in the session. The
PERL.PERL 100 file has been renamed to
PERL.PERL0.
10Apr1998 CDA Fixed it up for the upload to CPAN.
10Sep1998 CDA No longer need defines for individual library
files. We now use FILENAME_RESOLVE_ and "SEARCH"
defines.
11Nov1998 CDA Added MD5 extensions and a documentation file
called HOWTOEXT which explains how to extend
PERL for guardian. Added PERLEXT library
subvolume. Fixed BYTEORDER in CONFIGH (turns
out it was ending up 0x1234 and it should be
0x4321 and evidently only MD5 cared).
20Jan1999 CDA Made Perl compilable by both Guardian C (CISC) and
NMC (RISC). See "NMC Notes" for details. Also,
fixed PERLMNT.MAKE and PERL.PERL TACL macros so they
will work at non-IDX sites (references were made to
routines/macros that are locally defined). Combined
README and README2 files into a single file.
