path: root/docs/manpages-4/pidl.1.xml

<?xml version="1.0" encoding="iso-8859-1"?>
<!DOCTYPE refentry PUBLIC "-//Samba-Team//DTD DocBook V4.2-Based Variant V1.0//EN" "http://www.samba.org/samba/DTD/samba-doc">
<refentry id="pidl.1">

<refmeta>
	<refentrytitle>pidl</refentrytitle>
	<manvolnum>1</manvolnum>
</refmeta>


<refnamediv>
	<refname>pidl</refname>
	<refpurpose>IDL Compiler written in Perl</refpurpose>
</refnamediv>

<refsynopsisdiv>
	<cmdsynopsis>
		<command>pidl</command>
		<arg choice="opt">--help</arg>
		<arg choice="opt">--output OUTNAME</arg>
		<arg choice="opt">--parse</arg>
		<arg choice="opt">--dump</arg>
		<arg choice="opt">--header</arg>
		<arg choice="opt">--parser</arg>
		<arg choice="opt">--server</arg>
		<arg choice="opt">--template</arg>
		<arg choice="opt">--eparser</arg>
		<arg choice="opt">--diff</arg>
		<arg choice="opt">--keep</arg>
		<arg choice="req">idlfile</arg>
	</cmdsynopsis>
</refsynopsisdiv>

<refsect1>
	<title>DESCRIPTION</title>

	<para>pidl is an IDL compiler written in Perl that aims to be somewhat 
		compatible with the midl compiler. IDL stands for 
		"Interface Definition Language".</para>

	<para>pidl can generate stubs for DCE/RPC server code, DCE/RPC 
		client code and ethereal dissectors for DCE/RPC traffic.</para>

	<para>IDL compilers like <emphasis>pidl</emphasis> take a description 
		of an interface as their input and use it to generate C 
		(though support for other languages may be added later) code that 
		can use these interfaces, pretty print data sent 
		using these interfaces, or even generate ethereal 
		dissectors that can parse data sent over the 
		wire by these interfaces. </para>

	<para>pidl takes IDL files in the same format that is used by midl, 
		converts it to a .pidl file (which contains pidl's internal representation of the interface) and can then generate whatever output you need.
		.pidl files should be used for debugging purposes only. Write your 
		interface definitions in (midl) .idl format.
	</para>

	<para>
		The goal of pidl is to implement a IDL compiler that can be used 
		while developing the RPC subsystem in Samba (for 
		both marshalling/un-marshalling and debugging purposes).</para>

</refsect1>

<refsect1>
	<title>OPTIONS</title>

	<variablelist>
		<varlistentry>
		<term>--help</term>
		<listitem><para>
		Show list of available options.</para></listitem>
		</varlistentry>
		
		<varlistentry>
		<term>--output OUTNAME</term>
		<listitem><para>Write output files to OUTNAME.*, e.g. 
				OUTNAME.pidl. If --output is not used, the name of 
				the input IDL file is used without the extension and the dot 
				before the extension.
		</para></listitem>
		</varlistentry>
		
		<varlistentry>
		<term>--parse</term>
		<listitem><para>
				Tell pidl the files specified are (midl-style) IDL files.</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--dump</term>
		<listitem><para>
				Convert .pidl files to (midl-style) IDL files. FIle will be named OUTNAME.idl.</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--header</term>
		<listitem><para>
				Generate a C header file for the specified interface. File will be named OUTNAME.h.</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--parser</term>
		<listitem><para>
				Generate a C file capable of parsing data sent using the interface. 
				File will be named OUTNAME.c.
		</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--server</term>
		<listitem><para>
		Generate boilerplate for the RPC server that implements 
		the interface. Generates OUTNAME_s.c</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--template</term>
		<listitem><para>
		Generate stubs for a RPC server that implements 
		the interface. Output will be written to stdout.
		</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--eparser</term>
		<listitem><para>
		Generate an Ethereal dissector (in C) for the interface. Output will 
		be written to packet-dcerpc-OUTNAME.c.
		</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--diff</term>
		<listitem><para>
		Convert an IDL file to a pidl file and then back to a 
		IDL file and see if there are any differences with the 
		original IDL file. Useful for debugging pidl.</para></listitem>
		</varlistentry>


		<varlistentry>
		<term>--keep</term>
		<listitem><para>
		Tell pidl to keep the pidl files (used as intermediate files 
		between the IDL files and the parser/server/etc code). Useful 
		for debugging pidl.</para></listitem>
		</varlistentry>
	</variablelist>
</refsect1>

<refsect1>
	<title>SYNTAX</title>

	<para>IDL files are always preprocessed using the C preprocessor.</para>

	<para>Each IDL file describes exactly one interface. Interfaces 
		can contain several C-like function definitions.</para>

	<para>Pretty much everything in an interface (the interface itself,
		functions, parameters) can have attributes (or properties 
		whatever name you give them). Attributes 
		always prepend the element they apply to and are surrounded 
		by square brackets ([]). Multiple attributes 
		are separated by comma's; arguments to attributes are 
		specified between parentheses. </para>

	<para>See the section COMPATIBILITY for the list of attributes that 
		pidl supports.</para>

	<para>C-style comments can be used.</para>
	
</refsect1>

<refsect1>
	<title>MIDL TYPES</title>

<para>
pidl uses slightly different types to midl by default. The following
defines in your MS IDL may make things easier to use the same IDL on
both platforms.
</para>

<programlisting>
#define unistr [string] wchar_t *
#define uint8 char
#define uint16 short
#define uint32 long
#define HYPER_T hyper
</programlisting>

<para>
	Let's look at the multiple ways you can encode an array.
</para>

<refsect2>
	<title>CONFORMANT ARRAYS</title>

	<para>
A conformant array is one with that ends in [*] or []. The strange
things about conformant arrays are:
</para>

<simplelist>
	<member>they can only appear as the last element of a structure</member>
	<member>the array size appears before the structure itself on the wire. </member>
</simplelist>

<para>
	So, in this example:
</para>

<programlisting>
	typedef struct {
		long abc;
		long count;     
		long foo;
		[size_is(count)] long s[*];
	} Struct1;
</programlisting>

<para>
it appears like this:
</para>

<programlisting>
[size_is] [abc] [count] [foo] [s...]
</programlisting>

<para>
the first [size_is] field is the allocation size of the array, and
occurs before the array elements and even before the structure
alignment.
</para>

<para>
Note that size_is() can refer to a constant, but that doesn't change
the wire representation. It does not make the array a fixed array.
</para>

<para>
midl.exe would write the above array as the following C header:
</para>

<programlisting>
   typedef struct {
		long abc;
		long count;     
		long foo;
		long s[1];
	} Struct1;
</programlisting>

<para>
pidl takes a different approach, and writes it like this:
</para>

<programlisting>
    typedef struct {
		long abc;
		long count;     
		long foo;
		long *s;
	} Struct1;
</programlisting>

</refsect2>

<refsect2>
	<title>VARYING ARRAYS</title>


<para>
A varying array looks like this:
</para>

<programlisting>
	typedef struct {
		long abc;
		long count;     
		long foo;
		[size_is(count)] long *s;
	} Struct1;
</programlisting>

<para>
This will look like this on the wire:
</para>

<programlisting>
[abc] [count] [foo] [PTR_s]    [count] [s...]
</programlisting>

</refsect2>

<refsect2>
	<title>FIXED ARRAYS</title>

<para>
A fixed array looks like this:
</para>

<programlisting>
    typedef struct {
	    long s[10];
    } Struct1;
</programlisting>

<para>
The NDR representation looks just like 10 separate long
declarations. The array size is not encoded on the wire.
</para>

<para>
pidl also supports "inline" arrays, which are not part of the IDL/NDR
standard. These are declared like this:
</para>

<programlisting>
    typedef struct {
	    uint32 foo;
	    uint32 count;
	    uint32 bar;
	    long s[count];
    } Struct1;
</programlisting>

<para>
This appears like this:
</para>

<programlisting>
[foo] [count] [bar] [s...]
</programlisting>

<para>
Fixed arrays are an extension added to support some of the strange
embedded structures in security descriptors and spoolss. 
</para>

</refsect2>
</refsect1>

<refsect1>
	<title>COMPATIBILITY WITH MIDL</title>

	<refsect2>
		<title>Asynchronous communication</title>

		<!--FIXME-->
	</refsect2>

	<refsect2>
		<title>Typelibs (.tlb files)</title>

		<!-- FIXME -->
	</refsect2>

	<refsect2>
		<title>Pointers</title>

		<para>Pidl does not support "full" pointers in the DCE meaning of the word. However, its "unique" pointer is compatible with MIDL's full ("ptr") pointer support. </para>

		<para>Pidl does not assume all top level pointers for functions are 
			"ref".</para>
	</refsect2>

	<refsect2>
		<title>Datagram support</title>

		<para>ncadg is not supported yet.</para>
	</refsect2>

<refsect2>
	<title>Supported properties (attributes is the MIDL term)</title>

	<para>
in, out, ref, length_is, switch_is, size_is, uuid, case, default, string, unique, ptr, pointer_default, v1_enum, object, helpstring, range, local, call_as, endpoint, switch_type, progid, coclass, iid_is.
	</para>

</refsect2>

<refsect2>
	<title>PIDL Specific properties</title>

<variablelist>
	<varlistentry><term>public</term>
		<listitem><para>
The [public] property on a structure or union is a pidl extension that
forces the generated pull/push functions to be non-static. This allows
you to declare types that can be used between modules. If you don't
specify [public] then pull/push functions for other than top-level
functions are declared static.
		</para></listitem>
	</varlistentry>
				
	<varlistentry><term>noprint</term>
		<listitem><para>
The [noprint] property is a pidl extension that allows you to specify
that pidl should not generate a ndr_print_*() function for that
structure or union. This is used when you wish to define your own
print function that prints a structure in a nicer manner. A good
example is the use of [noprint] on dom_sid, which allows the
pretty-printing of SIDs.
		</para></listitem>
	</varlistentry>

	<varlistentry><term>value</term>
		<listitem><para>
The [value(expression)] property is a pidl extension that allows you
to specify the value of a field when it is put on the wire. This
allows fields that always have a well-known value to be automatically
filled in, thus making the API more programmer friendly. The
expression can be any C expression, although if you refer to variables
in the current structure you will need to dereference them with
r->. See samr_Name as a good example.
		</para></listitem>
	</varlistentry>
		
	<varlistentry><term>relative</term>
		<listitem><para>
The [relative] property can be supplied on a pointer. When it is used
it declares the pointer as a spoolss style "relative" pointer, which
means it appears on the wire as an offset within the current
encapsulating structure. This is not part of normal IDL/NDR, but it is
a very useful extension as it avoids the manual encoding of many
complex structures.
		</para></listitem>
	</varlistentry>

	<varlistentry><term>subcontext(length)</term>
		<listitem><para>
				Specifies that a size of <replaceable>length</replaceable>
				bytes should be read, followed by a blob of that size, 
				which will be parsed as NDR.
		</para></listitem>
	</varlistentry>

	<varlistentry><term>flag</term>
		<listitem><para>
				Specify boolean options, mostly used for 
				low-level NDR options. Several options 
				can be specified using the | character.
				Note that flags are inherited by substructures!
		</para></listitem>
	</varlistentry>

	<varlistentry><term>nodiscriminant</term>
		<listitem><para>
The [nodiscriminant] property on a union means that the usual uint16
discriminent field at the start of the union on the wire is
omitted. This is not normally allowed in IDL/NDR, but is used for some
spoolss structures.
		</para></listitem>
	</varlistentry>

	<varlistentry><term>align</term>
		<listitem><para>
				Force the alignment of the field this attribute is placed 
				on to the number of bytes specified.
		</para></listitem>
	</varlistentry>
</variablelist>
</refsect2>

<refsect2>
	<title>Unsupported MIDL properties</title>

<para>aggregatable, appobject, async_uuid, bindable, control, cpp_quote, defaultbind, defaultcollelem, defaultvalue, defaultvtable, dispinterface, displaybind, dual, entry, first_is, helpcontext, helpfile, helpstringcontext, helpstringdll, hidden, idl_module, idl_quote, id, immediatebind, importlib, import, include, includelib, last_is, lcid, licensed, max_is, module, ms_union, no_injected_text, nonbrowsable, noncreatable, nonextensible, odl, oleautomation, optional, pragma, propget, propputref, propput, readonly, requestedit, restricted, retval, source, transmit_as, uidefault, usesgetlasterror, vararg, vi_progid, wire_marshal. </para>

</refsect2>

</refsect1>

<refsect1>
	<title>BUGS</title>

	<itemizedlist>
		<listitem><para>Input should be validated better.</para></listitem>
	</itemizedlist>

</refsect1>

<refsect1>
	<title>VERSION</title>

	<para>This man page is correct for version 4.0 of the Samba suite.</para>
</refsect1>

<refsect1>
	<title>SEE ALSO</title>

	<para><ulink url="http://msdn.microsoft.com/library/en-us/rpc/rpc/field_attributes.asp">Field Attributes [Remote Procedure Call]</ulink>, ethereal</para>

</refsect1>

<refsect1>
	<title>AUTHOR</title>

	&man.credits.samba;

	<para>pidl was written by Andrew Tridgell, Stefan Metzmacher, Tim 
		Potter and Jelmer Vernooij. </para>

	<para>This manpage was written by Andrew Tridgell and Jelmer Vernooij. </para>
	
</refsect1>

</refentry>