winebuild(1) Wine dll builder

SYNOPSIS

winebuild [options] [inputfile...]

DESCRIPTION

winebuild generates the assembly files that are necessary to build a Wine dll, which is basically a Win32 dll encapsulated inside a Unix library.

winebuild has different modes, depending on what kind of file it is asked to generate. The mode is specified by one of the mode options specified below. In addition to the mode option, various other command-line option can be specified, as described in the OPTIONS section.

MODE OPTIONS

You have to specify exactly one of the following options, depending on what you want winebuild to generate.
--dll
Build an assembly file from a .spec file (see SPEC FILE SYNTAX for details), or from a standard Windows .def file. The .spec/.def file is specified via the -E option. The resulting file must be assembled and linked to the other object files to build a working Wine dll. In this mode, the input files should be the list of all object files that will be linked into the final dll, to allow winebuild to get the list of all undefined symbols that need to be imported from other dlls.
--exe
Build an assembly file for an executable. This is basically the same as the --dll mode except that it doesn't require a .spec/.def file as input, since an executable need not export functions. Some executables however do export functions, and for those a .spec/.def file can be specified via the -E option. The executable is named from the .spec/.def file name if present, or explicitly through the -F option. The resulting file must be assembled and linked to the other object files to build a working Wine executable, and all the other object files must be listed as input files.
--def
Build a .def file from a spec file. The .spec file is specified via the -E option. This is used when building dlls with a PE (Win32) compiler.
--implib
Build a .a import library from a spec file. The .spec file is specified via the -E option.
--resources
Generate a .o file containing all the input resources. This is useful when building with a PE compiler, since the PE binutils cannot handle multiple resource files as input. For a standard Unix build, the resource files are automatically included when building the spec file, so there's no need for an intermediate .o file.

OPTIONS

--as-cmd=as-command
Specify the command to use to compile assembly files; the default is as.
-b, --target=cpu-manufacturer[-kernel]-os
Specify the target CPU and platform on which the generated code will be built. The target specification is in the standard autoconf format as returned by config.sub.
--cc-cmd=cc-command
Specify the C compiler to use to compile assembly files; the default is to instead use the assembler specified with --as-cmd.
-d, --delay-lib=name
Set the delayed import mode for the specified library, which must be one of the libraries imported with the -l option. Delayed mode means that the library won't be loaded until a function imported from it is actually called.
-D symbol
Ignored for compatibility with the C compiler.
-e, --entry=function
Specify the module entry point function; if not specified, the default is DllMain for dlls, and main for executables (if the standard C main is not defined, WinMain is used instead). This is only valid for Win32 modules.
-E, --export=filename
Specify a .spec file (see SPEC FILE SYNTAX for details), or a standard Windows .def file that defines the exports of the DLL or executable that is being built.
--external-symbols
Allow linking to external symbols directly from the spec file. Normally symbols exported by a dll have to be defined in the dll itself; this option makes it possible to use symbols defined in another Unix library (for symbols defined in another dll, a forward specification must be used instead).
-f option
Specify a code generation option. Currently -fPIC and -fasynchronous-unwind-tables are supported. Other options are ignored for compatibility with the C compiler.
--fake-module
Create a fake PE module for a dll or exe, instead of the normal assembly or object file. The PE module contains the resources for the module, but no executable code.
-F, --filename=filename
Set the file name of the module. The default is to use the base name of the spec file (without any extension).
-h, --help
Display a usage message and exit.
-H, --heap=size
Specify the size of the module local heap in bytes (only valid for Win16 modules); default is no local heap.
-I directory
Ignored for compatibility with the C compiler.
-k, --kill-at
Remove the stdcall decorations from the symbol names in the generated .def file. Only meaningful in --def mode.
-K flags
Ignored for compatibility with the C compiler.
--large-address-aware
Set a flag in the executable to notify the loader that this application supports address spaces larger than 2 gigabytes.
--ld-cmd=ld-command
Specify the command to use to link the object files; the default is ld.
-L, --library-path=directory
Append the specified directory to the list of directories that are searched for import libraries.
-l, --library=name
Import the specified library, looking for a corresponding libname.def file in the directories specified with the -L option.
-m16, -m32, -m64
Generate respectively 16-bit, 32-bit or 64-bit code.
-marm, -mthumb, -march=option, -mcpu=option
Set code generation options for the assembler.
-M, --main-module=module
When building a 16-bit dll, set the name of its 32-bit counterpart to module. This is used to enforce that the load order for the 16-bit dll matches that of the 32-bit one.
-N, --dll-name=dllname
Set the internal name of the module. It is only used in Win16 modules. The default is to use the base name of the spec file (without any extension). This is used for KERNEL, since it lives in KRNL386.EXE. It shouldn't be needed otherwise.
--nm-cmd=nm-command
Specify the command to use to get the list of undefined symbols; the default is nm.
--nxcompat=yes|no
Specify whether the module is compatible with no-exec support. The default is yes.
-o, --output=file
Set the name of the output file (default is standard output). If the output file name ends in .o, the text output is sent to a temporary file that is then assembled to produce the specified .o file.
-r, --res=rsrc.res
Load resources from the specified binary resource file. The rsrc.res file can be produced from a source resource file with wrc(1) (or with a Windows resource compiler).
This option is only necessary for Win16 resource files, the Win32 ones can simply listed as input files and will automatically be handled correctly (though the -r option will also work for Win32 files).
--save-temps
Do not delete the various temporary files that winebuild generates.
--subsystem=subsystem[:major[.minor]]
Set the subsystem of the executable, which can be one of the following:
console for a command line executable,
windows for a graphical executable,
native for a native-mode dll,
wince for a ce dll.
The entry point of a command line executable is a normal C main function. A wmain function can be used instead if you need the argument array to use Unicode strings. A graphical executable has a WinMain entry point.
Optionally a major and minor subsystem version can also be specified; the default subsystem version is 4.0.
-u, --undefined=symbol
Add symbol to the list of undefined symbols when invoking the linker. This makes it possible to force a specific module of a static library to be included when resolving imports.
-v, --verbose
Display the various subcommands being invoked by winebuild.
--version
Display the program version and exit.
-w, --warnings
Turn on warnings.

SPEC FILE SYNTAX

General syntax

A spec file should contain a list of ordinal declarations. The general syntax is the following:

ordinal functype [flagsexportname ( [args...] ) [handler]
ordinal variable [flagsexportname ( [data...] )
ordinal extern [flagsexportname [symbolname]
ordinal stub [flagsexportname [ (args...) ]
ordinal equate [flagsexportname data
comments

Declarations must fit on a single line, except if the end of line is escaped using a backslash character. The # character anywhere in a line causes the rest of the line to be ignored as a comment.

ordinal specifies the ordinal number corresponding to the entry point, or '@' for automatic ordinal allocation (Win32 only).

flags is a series of optional flags, preceded by a '-' character. The supported flags are:

-norelay
The entry point is not displayed in relay debugging traces (Win32 only).
-noname
The entry point will be exported by ordinal instead of by name. The name is still available for importing.
-ret16
The function returns a 16-bit value (Win16 only).
-ret64
The function returns a 64-bit value (Win32 only).
-register
The function uses CPU register to pass arguments.
-private
The function cannot be imported from other dlls, it can only be accessed through GetProcAddress.
-ordinal
The entry point will be imported by ordinal instead of by name. The name is still exported.
-arch=cpu[,cpu]
The entry point is only available on the specified CPU architecture(s). The names win32 and win64 match all 32-bit or 64-bit CPU architectures respectively. In 16-bit dlls, specifying -arch=win32 causes the entry point to be exported from the 32-bit wrapper module.

Function ordinals

Syntax:
ordinal functype [flagsexportname ( [args...] ) [handler]

This declaration defines a function entry point. The prototype defined by exportname ( [args...] ) specifies the name available for dynamic linking and the format of the arguments. '@' can be used instead of exportname for ordinal-only exports.

functype should be one of:

stdcall
for a normal Win32 function
pascal
for a normal Win16 function
cdecl
for a Win16 or Win32 function using the C calling convention
varargs
for a Win16 or Win32 function using the C calling convention with a variable number of arguments
thiscall
for a Win32 function using the thiscall calling convention (first parameter in %ecx register on i386)

args should be one or several of:

word
(16-bit unsigned value)
s_word
(16-bit signed word)
long
(pointer-sized integer value)
int64
(64-bit integer value)
int128
(128-bit integer value)
float
(32-bit floating point value)
double
(64-bit floating point value)
ptr
(linear pointer)
str
(linear pointer to a null-terminated ASCII string)
wstr
(linear pointer to a null-terminated Unicode string)
segptr
(segmented pointer)
segstr
(segmented pointer to a null-terminated ASCII string).
Note: The 16-bit and segmented pointer types are only valid for Win16 functions.

handler is the name of the actual C function that will implement that entry point in 32-bit mode. The handler can also be specified as dllname.function to define a forwarded function (one whose implementation is in another dll). If handler is not specified, it is assumed to be identical to exportname.

This first example defines an entry point for the 32-bit GetFocus() call:

@ stdcall GetFocus() GetFocus

This second example defines an entry point for the 16-bit CreateWindow() call (the ordinal 100 is just an example); it also shows how long lines can be split using a backslash:

100 pascal CreateWindow(ptr ptr long s_word s_word s_word \
    s_word word word word ptr) WIN_CreateWindow

To declare a function using a variable number of arguments, specify the function as varargs and declare it in the C file with a '...' parameter for a Win32 function, or with an extra VA_LIST16 argument for a Win16 function. See the wsprintf* functions in user.exe.spec and user32.spec for an example.

Variable ordinals

Syntax:
ordinal variable [flagsexportname ( [data...] )

This declaration defines data storage as 32-bit words at the ordinal specified. exportname will be the name available for dynamic linking. data can be a decimal number or a hex number preceded by "0x". The following example defines the variable VariableA at ordinal 2 and containing 4 ints:

2 variable VariableA(-1 0xff 0 0)

This declaration only works in Win16 spec files. In Win32 you should use extern instead (see below).

Extern ordinals

Syntax:
ordinal extern [flagsexportname [symbolname]

This declaration defines an entry that simply maps to a C symbol (variable or function). It only works in Win32 spec files. exportname will point to the symbol symbolname that must be defined in the C code. Alternatively, it can be of the form dllname.symbolname to define a forwarded symbol (one whose implementation is in another dll). If symbolname is not specified, it is assumed to be identical to exportname.

Stub ordinals

Syntax:
ordinal stub [flagsexportname [ (args...) ]

This declaration defines a stub function. It makes the name and ordinal available for dynamic linking, but will terminate execution with an error message if the function is ever called.

Equate ordinals

Syntax:
ordinal equate [flagsexportname data

This declaration defines an ordinal as an absolute value. exportname will be the name available for dynamic linking. data can be a decimal number or a hex number preceded by "0x".

AUTHORS

winebuild has been worked on by many people over the years. The main authors are Robert J. Amstadt, Alexandre Julliard, Martin von Loewis, Ulrich Weigand and Eric Youngdale. Many other people have contributed new features and bug fixes. For a complete list, see the git commit logs.

BUGS

It is not yet possible to use a PE-format dll in an import specification; only Wine dlls can be imported.

Bugs can be reported on the Wine bug tracker

AVAILABILITY

winebuild is part of the Wine distribution, which is available through WineHQ, the Wine development headquarters