glib-genmarshal

glib-genmarshal — C code marshaller generation utility for GLib closures

Synopsis

glib-genmarshal [OPTION...] [FILE...]

Description

glib-genmarshal is a small utility that generates C code marshallers for callback functions of the GClosure mechanism in the GObject sublibrary of GLib. The marshaller functions have a standard signature, they get passed in the invoking closure, an array of value structures holding the callback function parameters and a value structure for the return value of the callback. The marshaller is then responsible to call the respective C code function of the closure with all the parameters on the stack and to collect its return value.

glib-genmarshal takes a list of marshallers to generate as input. The marshaller list is either read from standard input or from files passed as additional arguments on the command line.

Marshaller list format

The marshaller lists are processed line by line, a line can contain a comment in the form of

1
# this is a comment

or a marshaller specification of the form

RTYPE:PTYPE
RTYPE:PTYPE,PTYPE
RTYPE:PTYPE,PTYPE,PTYPE

(up to 16 PTYPEs may be present).

The RTYPE part specifies the callback's return type and the PTYPEs right to the colon specify the callback's parameter list, except for the first and the last arguments which are always pointers.


Parameter types

Currently, the following types are supported:

VOID

indicates no return type, or no extra parameters. If VOID is used as the parameter list, no additional parameters may be present.

BOOLEAN

for boolean types (gboolean)

CHAR

for signed char types (gchar)

UCHAR

for unsigned char types (guchar)

INT

for signed integer types (gint)

UINT

for unsigned integer types (guint)

LONG

for signed long integer types (glong)

ULONG

for unsigned long integer types (gulong)

INT64

for signed 64bit integer types (gint64)

UINT64

for unsigned 64bit integer types (guint64)

ENUM

for enumeration types (gint)

FLAGS

for flag enumeration types (guint)

FLOAT

for single-precision float types (gfloat)

DOUBLE

for double-precision float types (gdouble)

STRING

for string types (gchar*)

BOXED

for boxed (anonymous but reference counted) types (GBoxed*)

PARAM

for GParamSpec or derived types (GParamSpec*)

POINTER

for anonymous pointer types (gpointer)

OBJECT

for GObject or derived types (GObject*)

VARIANT

for GVariant types (GVariant*)

NONE

deprecated alias for VOID

BOOL

deprecated alias for BOOLEAN

Options

--header

Generate header file contents of the marshallers.

--body

Generate C code file contents of the marshallers.

--prefix=PREFIX

Specify marshaller prefix. The default prefix is `g_cclosure_marshal'.

--skip-source

Skip source location remarks in generated comments.

--stdinc

Use the standard marshallers of the GObject library, and include gmarshal.h in generated header files.

--nostdinc

Do not use the standard marshallers of the GObject library, and skip gmarshal.h include directive in generated header files.

--internal

Mark generated functions as internal, using G_GNUC_INTERNAL.

--valist-marshallers

Generate valist marshallers, for use with g_signal_set_va_marshaller().

-v, --version

Print version information.

--g-fatal-warnings

Make warnings fatal, that is, exit immediately once a warning occurs.

-h, --help

Print brief help and exit.

-v, --version

Print version and exit.

Example

To generate marshallers for the following callback functions:

1
2
3
4
5
6
7
8
9
void   foo (gpointer data1,
            gpointer data2);
void   bar (gpointer data1,
            gint     param1,
            gpointer data2);
gfloat baz (gpointer data1,
            gboolean param1,
            guchar   param2,
            gpointer data2);

The marshaller.list file has to look like this:

VOID:VOID
VOID:INT
FLOAT:BOOLEAN,UCHAR

and you call glib-genmarshal like this:

glib-genmarshal --header marshaller.list > marshaller.h
glib-genmarshal --body marshaller.list > marshaller.c

The generated marshallers have the arguments encoded in their function name. For this particular list, they are

g_cclosure_user_marshal_VOID__VOID(),
g_cclosure_user_marshal_VOID__INT(),
g_cclosure_user_marshal_FLOAT__BOOLEAN_UCHAR().

They can be used directly for GClosures or be passed in as the GSignalCMarshaller c_marshaller; argument upon creation of signals:

1
2
3
4
5
6
7
8
GClosure *cc_foo, *cc_bar, *cc_baz;

cc_foo = g_cclosure_new (NULL, foo, NULL);
g_closure_set_marshal (cc_foo, g_cclosure_user_marshal_VOID__VOID);
cc_bar = g_cclosure_new (NULL, bar, NULL);
g_closure_set_marshal (cc_bar, g_cclosure_user_marshal_VOID__INT);
cc_baz = g_cclosure_new (NULL, baz, NULL);
g_closure_set_marshal (cc_baz, g_cclosure_user_marshal_FLOAT__BOOLEAN_UCHAR);

See also

glib-mkenums(1)