#

# CDDL HEADER START

#

# The contents of this file are subject to the terms of the

# Common Development and Distribution License (the "License").

# You may not use this file except in compliance with the License.

#

# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE

# or http://www.opensolaris.org/os/licensing.

# See the License for the specific language governing permissions

# and limitations under the License.

#

# When distributing Covered Code, include this CDDL HEADER in each

# file and include the License file at usr/src/OPENSOLARIS.LICENSE.

# If applicable, add the following below this CDDL HEADER, with the

# fields enclosed by brackets "[]" replaced with your own identifying

# information: Portions Copyright [yyyy] [name of copyright owner]

#

# CDDL HEADER END

#

#

# Copyright 2006 Sun Microsystems, Inc. All rights reserved.

# Use is subject to license terms.

#

# ident "%Z%%M% %I% %E% SMI"

#

Writing Library Makefiles in ON

===============================

Introduction

------------

This document guides you through the gnarly process of writing library

Makefiles for the ON consolidation. It assumes that you're comfortable with

make(1) and are somewhat familiar with the ON Makefile standards outlined in

/shared/ON/general_docs/make_std.txt.

Makefile Overview

-----------------

Your library should consist of a hierarchical collection of Makefiles:

lib/<library>/Makefile:

This is your library's top-level Makefile. It should contain rules

for building any ISA-independent targets, such as installing header

files and building message catalogs, but should defer all other

targets to ISA-specific Makefiles.

lib/<library>/Makefile.com

This is your library's common Makefile. It should contain rules

and macros which are common to all ISAs. This Makefile should never

be built explicitly, but instead should be included (using the make

include mechanism) by all of your ISA-specific Makefiles.

lib/<library>/<isa>/Makefile

These are your library's ISA-specific Makefiles, one per ISA

(usually sparc and i386, and often sparcv9 and amd64). These

Makefiles should include your common Makefile and then provide any

needed ISA-specific rules and definitions, perhaps overriding those

provided in your common Makefile.

To simplify their maintenance and construction, $(SRC)/lib has a handful of

provided Makefiles that yours must include; the examples provided throughout

the document will show how to use them. Please be sure to consult these

Makefiles before introducing your own custom build macros or rules.

lib/Makefile.lib:

This contains the bulk of the macros for building shared objects.

lib/Makefile.lib.64

This contains macros for building 64-bit objects, and should be

included in Makefiles for 64-bit native ISAs.

lib/Makefile.rootfs

This contains macro overrides for libraries that install into /lib

(rather than /usr/lib).

lib/Makefile.targ

This contains rules for building shared objects.

The remainder of this document discusses how to write each of your Makefiles

in detail, and provides examples from the libinetutil library.

The Library Top-level Makefile

------------------------------

As described above, your top-level library Makefile should contain

rules for building ISA-independent targets, but should defer the

building of all other targets to ISA-specific Makefiles. The

ISA-independent targets usually consist of:

install_h

Install all library header files into the proto area.

Can be omitted if your library has no header files.

check

Check all library header files for hdrchk compliance.

Can be omitted if your library has no header files.

_msg

Build and install a message catalog.

Can be omitted if your library has no message catalog.

Of course, other targets (such as `cstyle') are fine as well, as long as

they are ISA-independent.

The ROOTHDRS and CHECKHDRS targets are provided in lib/Makefile.lib to make

it easy for you to install and check your library's header files. To use

these targets, your Makefile must set the HDRS to the list of your library's

header files to install and HDRDIR to the their location in the source tree.

In addition, if your header files need to be installed in a location other

than $(ROOT)/usr/include, your Makefile must also set ROOTHDRDIR to the

appropriate location in the proto area. Once HDRS, HDRDIR and (optionally)

ROOTHDRDIR have been set, your Makefile need only contain

install_h: $(ROOTHDRS)

check: $(CHECKHDRS)

to bind the provided targets to the standard `install_h' and `check' rules.

Similar rules are provided (in $(SRC)/Makefile.msg.targ) to make it easy for

you to build and install message catalogs from your library's source files.

To install a catalog into the catalog directory in the proto area, define the

POFILE macro to be the name of your catalog, and specify that the _msg target

depends on $(MSGDOMAINPOFILE). The examples below should clarify this.

To build a message catalog from arbitrarily many message source files, use

the BUILDPO.msgfiles macro.

include ../Makefile.lib

POFILE = libfoo.po

MSGFILES = $(OBJECTS:%.o=%.i)

# ...

$(POFILE): $(MSGFILES)

$(BUILDPO.msgfiles)

_msg: $(MSGDOMAINPOFILE)

include $(SRC)/Makefile.msg.targ

Note that this example doesn't use grep to find message files, since that can

mask unreferenced files, and potentially lead to the inclusion of unwanted

messages or omission of intended messages in the catalogs. As such, MSGFILES

should be derived from a known list of objects or sources.

It is usually preferable to run the source through the C preprocessor prior

to extracting messages. To do this, use the ".i" suffix, as shown in the

above example. If you need to skip the C preprocessor, just use the native

(.[ch]) suffix.

The only time you shouldn't use BUILDPO.msgfiles as the preferred means of

extracting messages is when you're extracting them from shell scripts; in

that case, you can use the BUILDPO.pofiles macro as explained below.

To build a message catalog from other message catalogs, or from source files

that include shell scripts, use the BUILDPO.pofiles macro:

include ../Makefile.lib

SUBDIRS = $(MACH)

POFILE = libfoo.po

POFILES = $(SUBDIRS:%=%/_%.po)

_msg := TARGET = _msg

# ...

$(POFILE): $(POFILES)

$(BUILDPO.pofiles)

_msg: $(MSGDOMAINPOFILE)

include $(SRC)/Makefile.msg.targ

The Makefile above would work in conjunction with the following in its

subdirectories' Makefiles:

POFILE = _thissubdir.po

MSGFILES = $(OBJECTS:%.o=%.i)

$(POFILE): $(MSGFILES)

$(BUILDPO.msgfiles)

_msg: $(POFILE)

include $(SRC)/Makefile.msg.targ

Since this POFILE will be combined with those in other subdirectories by the

parent Makefile and that merged file will be installed into the proto area

via MSGDOMAINPOFILE, there is no need to use MSGDOMAINPOFILE in this Makefile

(in fact, using it would lead to duplicate messages in the catalog).

When using any of these targets, keep in mind that other macros, like

XGETFLAGS and TEXT_DOMAIN may also be set in your Makefile to override or

augment the defaults provided in higher-level Makefiles.

As previously mentioned, you should defer all ISA-specific targets to your

ISA-specific Makefiles. You can do this by:

1. Setting SUBDIRS to the list of directories to descend into:

SUBDIRS = $(MACH)

Note that if your library is also built 64-bit, then you should

also specify

$(BUILD64)SUBDIRS += $(MACH64)

so that SUBDIRS contains $(MACH64) if and only if you're compiling

on a 64-bit ISA.

2. Providing a common "descend into SUBDIRS" rule:

$(SUBDIRS): FRC

@cd $@; pwd; $(MAKE) $(TARGET)

FRC:

3. Providing a collection of conditional assignments that set TARGET

appropriately:

all := TARGET= all

clean := TARGET= clean

clobber := TARGET= clobber

install := TARGET= install

lint := TARGET= lint

The order doesn't matter, but alphabetical is preferable.

4. Having the aforementioned targets depend on SUBDIRS:

all clean clobber install lint: $(SUBDIRS)

The `all' target must be listed first so that make uses it as the

default target; the others might as well be listed alphabetically.

As an example of how all of this goes together, here's libinetutil's

top-level library Makefile (license notice and copyright omitted):

include ../Makefile.lib

HDRS = libinetutil.h

HDRDIR = common

SUBDIRS = $(MACH)

$(BUILD64)SUBDIRS += $(MACH64)

all := TARGET = all

clean := TARGET = clean

clobber := TARGET = clobber

install := TARGET = install

lint := TARGET = lint

.KEEP_STATE:

all clean clobber install lint: $(SUBDIRS)

install_h: $(ROOTHDRS)

check: $(CHECKHDRS)

$(SUBDIRS): FRC

@cd $@; pwd; $(MAKE) $(TARGET)

FRC:

include ../Makefile.targ

The Common Makefile

-------------------

In concept, your common Makefile should contain all of the rules and

definitions that are the same on all ISAs. However, for reasons of

maintainability and cleanliness, you're encouraged to place even

ISA-dependent rules and definitions, as long you express them in an

ISA-independent way (e.g., by using $(MACH), $(TARGETMACH), and their kin).

(TARGETMACH is the same as MACH for 32-bit targets, and the same as MACH64

for 64-bit targets).

The common Makefile can be conceptually split up into four sections:

1. A copyright and comments section. Please see the prototype

files in usr/src/prototypes for examples of how to format the

copyright message properly. For brevity and clarity, this

section has been omitted from the examples shown here.

2. A list of macros that must be defined prior to the inclusion of

Makefile.lib. This section is conceptually terminated by the

inclusion of Makefile.lib, followed, if necessary, by the

inclusion of Makefile.rootfs (only if the library is to be

installed in /lib rather than the default /usr/lib).

3. A list of macros that need not be defined prior to the inclusion

of Makefile.lib (or which must be defined following the inclusion

of Makefile.lib, to override or augment its definitions). This

section is conceptually terminated by the .KEEP_STATE directive.

4. A list of targets.

The first section is self-explanatory. The second typically consists of the

following macros:

LIBRARY

Set to the name of the static version of your library, such

as `libinetutil.a'. You should always specify the `.a' suffix,

since pattern-matching rules in higher-level Makefiles rely on it,

even though static libraries are not normally built in ON, and

are never installed in the proto area. Note that the LIBS macro

(described below) controls the types of libraries that are built

when building your library.

If you are building a loadable module (i.e., a shared object that

is only linked at runtime with dlopen(3dl)), specify the name of

the loadable module with a `.a' suffix, such as `devfsadm_mod.a'.

VERS

Set to the version of your shared library, such as `.1'. You

actually do not need to set this prior to the inclusion of

Makefile.lib, but it is good practice to do so since VERS and

LIBRARY are so closely related.

OBJECTS

Set to the list of object files contained in your library, such as

`a.o b.o'. Usually, this will be the same as your library's source

files (except with .o extensions), but if your library compiles

source files outside of the library directory itself, it will

differ. We'll see an example of this with libinetutil.

The third section typically consists of the following macros:

LIBS

Set to the list of the types of libraries to build when building

your library. For dynamic libraries, you should set this to

`$(DYNLIB) $(LINTLIB)' so that a dynamic library and lint library

are built. For loadable modules, you should just list DYNLIB,

since there's no point in building a lint library for libraries

that are never linked at compile-time.

If your library needs to be built as a static library (typically

to be used in other parts of the build), you should set LIBS to

`$(LIBRARY)'. However, you should do this only when absolutely

necessary, and you must *never* ship static libraries to customers.

ROOTLIBDIR (if your library installs to a nonstandard directory)

Set to the directory your 32-bit shared objects will install into

with the standard $(ROOTxxx) macros. Since this defaults to

$(ROOT)/usr/lib ($(ROOT)/lib if you included Makefile.rootfs),

you usually do not need to set this.

ROOTLIBDIR64 (if your library installs to a nonstandard directory)

Set to the directory your 64-bit shared objects will install into

with the standard $(ROOTxxx64) macros. Since this defaults to

$(ROOT)/usr/lib/$(MACH64) ($(ROOT)/lib/$(MACH64) if you included

Makefile.rootfs), you usually do not need to set this.

SRCDIR

Set to the directory containing your library's source files, such

as `../common'. Because this Makefile is actually included from

your ISA-specific Makefiles, make sure you specify the directory

relative to your library's <isa> directory.

SRCS (if necessary)

Set to the list of source files required to build your library.

This defaults to $(OBJECTS:%.o=$(SRCDIR)/%.c) in Makefile.lib, so

you only need to set this when source files from directories other

than SRCDIR are needed. Keep in mind that SRCS should be set to a

list of source file *pathnames*, not just a list of filenames.

LINTLIB-specific SRCS (required if building a lint library)

Set to a special "lint stubs" file to use when constructing your

library's lint library. The lint stubs file must be used to

guarantee that programs that link against your library will be able

to lint clean. To do this, you must conditionally set SRCS to use

your stubs file by specifying `LINTLIB := SRCS= $(SRCDIR)/$(LINTSRC)'

in your Makefile. Of course, you do not need to set this if your

library does not build a lint library.

LDLIBS

Appended with the list of libraries and library directories needed

to build your library; minimally "-lc". Note that this should

*never* be set, since that will inadvertently clear the library

search path, causing the linker to look in the wrong place for

the libraries.

Since lint targets also make use of LDLIBS, LDLIBS *must* only

contain -l and -L directives; all other link-related directives

should be put in DYNFLAGS (if they apply only to shared object

construction) or LDFLAGS (if they apply in general).

MAPFILES (if necessary)

Set to the list of mapfiles used to link each ISA-specific version

of your library. This defaults to `$(SRCDIR)/mapfile-vers' in

Makefile.lib, so you only need to change this if you have additional

mapfiles or your mapfile doesn't follow the standard naming

convention. If you have supplemental ISA-dependent mapfiles that

reside in the respective <isa> directories, you can augment

MAPFILES like this:

MAPFILES += mapfile-vers

CPPFLAGS (if necessary)

Appended with any flags that need to be passed to the C

preprocessor (typically -D and -I flags). Since lint macros use

CPPFLAGS, CPPFLAGS *must* only contain directives known to the C

preprocessor. When compiling MT-safe code, CPPFLAGS *must*

include -D_REENTRANT. When compiling large file aware code,

CPPFLAGS *must* include -D_FILE_OFFSET_BITS=64.

CFLAGS

Appended with any flags that need to be passed to the C compiler.

Minimally, append `$(CCVERBOSE)'. Keep in mind that you should

add any C preprocessor flags to CPPFLAGS, not CFLAGS.

CFLAGS64 (if necessary)

Appended with any flags that need to be passed to the C compiler

when compiling 64-bit code. Since all 64-bit code is compiled

$(CCVERBOSE), you usually do not need to modify CFLAGS64.

COPTFLAG (if necessary)

Set to control the optimization level used by the C compiler when

compiling 32-bit code. You should only set this if absolutely

necessary, and it should only contain optimization-related

settings (or -g).

COPTFLAG64 (if necessary)

Set to control the optimization level used by the C compiler when

compiling 64-bit code. You should only set this if absolutely

necessary, and it should only contain optimization-related

settings (or -g).

LINTFLAGS (if necessary)

Appended with any flags that need to be passed to lint when

linting 32-bit code. You should only modify LINTFLAGS in

rare instances where your code cannot (or should not) be fixed.

LINTFLAGS64 (if necessary)

Appended with any flags that need to be passed to lint when

linting 64-bit code. You should only modify LINTFLAGS64 in

rare instances where your code cannot (or should not) be fixed.

Of course, you may use other macros as necessary.

The fourth section typically consists of the following targets:

all

Build all of the types of the libraries named by LIBS. Must always

be the first real target in common Makefile. Since the

higher-level Makefiles already contain rules to build all of the

different types of libraries, you can usually just specify

all: $(LIBS)

though it should be listed as an empty target if LIBS is set by your

ISA-specific Makefiles (see above).

lint

Use the `lintcheck' rule provided by lib/Makefile.targ to lint the

actual library sources. Historically, this target has also been

used to build the lint library (using LINTLIB), but that usage is

now discouraged. Thus, this rule should be specified as

lint: lintcheck

Conspicuously absent from this section are the `clean' and `clobber' targets.

These targets are already provided by lib/Makefile.targ and thus should not

be provided by your common Makefile. Instead, your common Makefile should

list any additional files to remove during a `clean' and `clobber' by

appending to the CLEANFILES and CLOBBERFILES macros.

Once again, here's libinetutil's common Makefile, which shows how many of

these directives go together. Note that Makefile.rootfs is included to

cause libinetutil.so.1 to be installed in /lib rather than /usr/lib:

LIBRARY = libinetutil.a

VERS = .1

OBJECTS = octet.o inetutil4.o ifspec.o ifaddrlist.o eh.o tq.o

include ../../Makefile.lib

include ../../Makefile.rootfs

LIBS = $(DYNLIB) $(LINTLIB)

SRCDIR = ../common

COMDIR = $(SRC)/common/net/dhcp

SRCS = $(COMDIR)/octet.c $(SRCDIR)/inetutil4.c \

$(SRCDIR)/ifspec.c $(SRCDIR)/eh.c $(SRCDIR)/tq.c \

$(SRCDIR)/ifaddrlist.c

$(LINTLIB):= SRCS = $(SRCDIR)/$(LINTSRC)

LDLIBS += -lsocket -lc

CFLAGS += $(CCVERBOSE)

CPPFLAGS += -I$(SRCDIR)

.KEEP_STATE:

all: $(LIBS)

lint: lintcheck

pics/%.o: $(COMDIR)/%.c

$(COMPILE.c) -o $@ $<

$(POST_PROCESS_O)

include ../../Makefile.targ

The mapfile for libinetutil is named `mapfile-vers' and resides in $(SRCDIR),

so the MAPFILES definition is omitted, defaulting to $(SRCDIR)/mapfile-vers.

Note that for libinetutil, not all of the object files come from SRCDIR. To

support this, an alternate source file directory named COMDIR is defined, and

the source files listed in SRCS are specified using both COMDIR and SRCDIR.

Additionally, a special build rule is provided to build object files from the

sources in COMDIR; the rule uses COMPILE.c and POST_PROCESS_O so that any

changes to the compilation and object-post-processing phases will be

automatically picked up.

The ISA-Specific Makefiles

--------------------------

As the name implies, your ISA-specific Makefiles should contain macros and

rules that cannot be expressed in an ISA-independent way. Usually, the only

rule you will need to put here is `install', which has different dependencies

for 32-bit and 64-bit libraries. For instance, here are the ISA-specific

Makefiles for libinetutil:

sparc/Makefile:

include ../Makefile.com

install: all $(ROOTLIBS) $(ROOTLINKS) $(ROOTLINT)

sparcv9/Makefile:

include ../Makefile.com

include ../../Makefile.lib.64

install: all $(ROOTLIBS64) $(ROOTLINKS64)

i386/Makefile:

include ../Makefile.com

install: all $(ROOTLIBS) $(ROOTLINKS) $(ROOTLINT)

amd64/Makefile:

include ../Makefile.com

include ../../Makefile.lib.64

install: all $(ROOTLIBS64) $(ROOTLINKS64)

Observe that there is no .KEEP_STATE directive in these Makefiles, since all

of these Makefiles include libinetutil/Makefile.com, and it already has a

.KEEP_STATE directive. Also, note that the 64-bit Makefiles also include

Makefile.lib.64, which overrides some of the definitions contained in the

higher level Makefiles included by the common Makefile so that 64-bit

compiles work correctly.

CTF Data in Libraries

---------------------

By default, all position-independent objects are built with CTF data using

ctfconvert, which is then merged together using ctfmerge when the shared

object is built. All C-source objects processed via ctfmerge need to be

processed via ctfconvert or the build will fail. Objects built from non-C

sources (such as assembly or C++) are silently ignored for CTF processing.

Filter libraries that have no source files will need to explicitly disable

CTF by setting CTFMERGE_LIB to ":"; see libw/Makefile.com for an example.

More Information

----------------

Other issues and questions will undoubtedly arise while you work on your

library's Makefiles. To help in this regard, a number of libraries of

varying complexity have been updated to follow the guidelines and practices

outlined in this document:

lib/libdhcputil

Example of a simple 32-bit only library.

lib/libdhcpagent

Example of a simple 32-bit only library that obtains its sources

from multiple directories.

lib/ncad_addr

Example of a simple loadable module.

lib/libipmp

Example of a simple library that builds a message catalog.

lib/libdhcpsvc

Example of a Makefile hierarchy for a library and a collection

of related pluggable modules.

lib/lvm

Example of a Makefile hierarchy for a collection of related

libraries and pluggable modules.

Also an example of a Makefile hierarchy that supports the

_dc target for domain and category specific messages.

Of course, if you still have questions, please do not hesitate to send email

to the ON gatekeepers.

#

# CDDL HEADER START

#

# The contents of this file are subject to the terms of the

# Common Development and Distribution License (the "License").

# You may not use this file except in compliance with the License.

#

# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE

# or http://www.opensolaris.org/os/licensing.

# See the License for the specific language governing permissions

# and limitations under the License.

#

# When distributing Covered Code, include this CDDL HEADER in each

# file and include the License file at usr/src/OPENSOLARIS.LICENSE.

# If applicable, add the following below this CDDL HEADER, with the

# fields enclosed by brackets "[]" replaced with your own identifying

# information: Portions Copyright [yyyy] [name of copyright owner]

#

# CDDL HEADER END

#

#

# Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.

#

Mapfiles and versioning in illumos

==================================

1.0 Objective of this README

This README describes the engineering practices of creating and updating

visible library interfaces. It describes various kinds of actions that

typically occur as libraries are evolved, and shows how interface

specifications are affected or updated in accordance. It tells you what

you must do as a shared library developer if you:

1. Make interface additions to an existing library

- add a Public interface

- add a Private interface

2. Update an interface in an existing library

- remove an existing interface

- promote a Private interface to Public

- scope a Private interface to local

- move an interface from one library to another

- copy interfaces which are part of the standard to a new or

existing library

3. Introduce a new library

- source directory hierarchy

- creation of the "mapfile-vers" file

- Makefiles

4. Make an entire library obsolete before end-of-life

- introduce SUNWobsolete to the "mapfile-vers" file

-------------------------------------------------------------------------------

2.0 What's a mapfile?

Mapfiles are used to tell the link-editor ("ld") all sorts of things about

how to generate an executable file or a shared object from a collection of

relocatable objects, such as generated by a compiler. For all the gory

details, see the Solaris Linker and Libraries Guide.

There are two versions of the mapfile language accepted by the link-editor.

Version 1 derives from AT&T System V Release 4 Unix. Version 2 is a newer

syntax specific to Solaris and derivatives. All mapfiles in illumos are

required to use version 2 syntax. Note that every mapfile using version 2

syntax must start with the line:

$mapfile_version 2

Here, we are only concerned with specifying externally-visible interfaces

for shared libraries (shared objects) and with specifying their versions

for ABI (Application Binary Interface) purposes. For these purposes, we

only need to deal with a subset of the mapfile language.

There should be a "mapfile-vers" file associated with every shared library

and it should reside in the common source directory for that library, most

often in a "common" directory. This is the usual layout of a library's

top-level directory (usr/src/lib/libwombat):

Makefile amd64/ i386/ sparcv9/

Makefile.com common/ sparc/

The "common" directory contains the source files and other common files

for the library:

bat.c libwombat_impl.h mapfile-vers wom.c

libwombat.h llib-lwombat util.c wombat.c

The mapfile's name is, by convention, "mapfile-vers" because it is used

for only two purposes: to specify externally-visible interface names while

suppressing visibility of all other names, and to specify their respective

unique version names.

-------------------------------------------------------------------------------

3.0 Contents of mapfile-vers

The structure of mapfile-vers is best explained by an example

(the license notification and copyright notice is omitted here

for brevity):

$mapfile_version 2

SYMBOL_VERSION ILLUMOS_0.2 { # Second interface change in illumos

global:

wb_notify;

} ILLUMOS_0.1;

SYMBOL_VERSION ILLUMOS_0.1 { # First interface change in illumos

global:

wb_poll;

} SUNW_1.2;

SYMBOL_VERSION SUNW_1.2 { # update to libwombat, Solaris 10

global:

wb_readv;

wb_stat;

wb_writev;

} SUNW_1.1;

SYMBOL_VERSION SUNW_1.1 { # first release of libwombat, Solaris 9

global:

wb_read;

wb_write;

};

SYMBOL_VERSION SUNWprivate { # private libwombat symbols

global:

wb_add;

wb_delete;

wb_search;

local:

*;

};

Each of these sections is a version declaration describing an ABI version of

the library containing the set of symbols exposed by the library to

external callers.

ABI versions must be constant, that is version ILLUMOS_0.2 in a given

library must always describe the same interface such that applications may

safely state their dependencies in terms of these versions and have a

constant and predictable ABI be exposed. This in effect means that once a

version is publicly visible, it may never be removed or have symbols added

to or removed from it.

ABI versions with the same major number should be upward compatible, that is

ILLUMOS_0.3 of a given library must contain all the interfaces in

ILLUMOS_0.2, and they must be compatible.

The private version, however, is special, and describes any private yet

exposed symbols within a library, and may change at any time (and without

notice). It exists purely to allow certain symbols to be of global scope

but not Public. Similarly, any symbols scoped local are private and may

change safely, even if the local statement happens to be within a public

version.

Interface changes made in illumos should be done with ILLUMOS_0.* versions,

introducing one version per interface change. In illumos, unlike Solaris,

symbol versions are not release-specific because of the requirement that

each be constant. No change should be made to add or remove symbols from

any pre-existing Public version.

The SUNW_*.* names were the Public version names of the library in Solaris.

There should be at most one version name for each release of Solaris, with

the minor number incremented by one over the previous version. No changes

should ever be made to SUNW_1.* versions.

So, for example, to add a new interface to libwombat in illumos one would add:

SYMBOL_VERSION ILLUMOS_0.3 { # Third update to libwombat in illumos

global:

wb_lseek;

} ILLUMOS_0.2;

Each version must inherit all symbols from its preceding version, specified at

the ending "}" for each version. The initial public version does not inherit

any symbols. The private version named either "SUNWprivate" for libraries

with private symbols pre-existing illumos, or "ILLUMOSprivate" otherwise

stands alone, inheriting nothing and being inherited by nothing.

The two lines in SUNWprivate:

local:

*;

ensure that no symbols other than those listed in the mapfile are visible to

clients of the library. If there is no private version, these two lines should

appear in the first public version.

For maintainability, the list of names in each version block should

be sorted in dictionary order (sort -d). Please comply.

The version 2 mapfile language supports a simple mechanism for conditional

input, in which lines in the mapfile apply only to a specific platform or

ELFCLASS (32/64-bit). This mechanism works very much like the #if/#endif

feature of the C preprocessor. For instance, the following mapfile declares

a version SUNW_1.1 that always exports a symbol foo, and also exports

the symbol bar on 32-bit sparc platforms:

$mapfile_version 2

SYMBOL_VERSION SUNW_1.1 {

foo;

$if _sparc && _ELF32

bar;

$endif

};

Conditional input can be used if there are ISA-specific library interfaces

not common to all instances of the library. It is the preferred method for

expressing platform specific items, as long as the differences are simple

(which is almost always the case). For example, see libproc, or, if you

are masochistic, libc or libnsl. In general, use of this feature should be

minimal.

In addition to conditional input, there is a second heavier weight mechanism

for expressing ISA-specific differences. In addition to the common mapfile:

common/mapfile-vers

some libraries may have ISA-specific supplemental mapfiles, one in each

of the ISA directories:

amd64/mapfile-vers

i386/mapfile-vers

sparc/mapfile-vers

sparcv9/mapfile-vers

The ISA-specific mapfiles look like the common mapfile, except that only

the ISA-specific names appear. The version names are the same as those

in the common mapfile, but only non-empty version instances are present

and no inheritance specification is present. The link-editor reads the

information from the common and ISA-specific mapfiles and merges them

in memory into a single description used to create the resulting object.

ISA-specific mapfiles were used with the version 1 mapfile language, which

lacked conditional input. Their use is rare now, as conditional input is

generally preferred. However, it is important to use conditional input

carefully, or the resulting mapfile can be extremly difficult to read.

-------------------------------------------------------------------------------

4.0 Making interface additions to an existing library

4.1 Adding a Public interface

Public interfaces should be added to a new ILLUMOS_ symbol version, with the

minor number incremented by one from the current highest version name. If

this is the first Public interface in the shared object, a new ILLUMOS_0.1

version name must be introduced.

The major revision number is incremented whenever an incompatible change is

made to an interface. This could be the case if an API changes so

dramatically as to invalidate dependencies. This should almost never occur

in practice. It also requires changing the suffix of the shared object

from, say, .so.1 to .so.2 and introducing code to continue to ship the .so.1

version of the library.

The minor revision number is incremented whenever one or more new interfaces

is added to a library. Once a version comes to exist in illumos, it is from

that point onward considered to be immutable.

4.2 Adding a Private interface

Private interfaces are the non-ABI interfaces of the library. Unlike

introducing a Public interface, a new entry is simply added to the

private version. No minor number increment is necessary.

If this interface happens to be the first Private interface introduced into

the library, the private version must be created (with no major.minor

version numbers). It inherits nothing, nothing inherits from it and it

should be named ILLUMOSprivate.

If the library already has Private interfaces in a SUNWprivate version, you

should use that. They may have numbered version names like SUNWprivate_m.n

(due to errors of the past). If so, just use the highest numbered private

version name to version the new interface. There is no need to introduce a

new private version name. Be careful not to use a lower numbered private

version name; doing so can cause runtime errors (as opposed to load time

errors) when running an application with older versions of the library.

There are also libraries in illumos that contain only private interfaces. In

such libraries, the private versions maybe legitimately be versioned and

they may be incremented to ensure that the programs that depend on them are

built and delivered as a integrated unit. A notable example of this is

libld.so (usr/src/cmd/sgs/libld), which contains the implementation of the

link-editor, the public interface to which is provided by the ld

command. When making a modification to the interface of such a library, you

should follow the convention already in place.

4.3 Historical handling of Solaris update releases.

To aid the understanding of our existing mapfiles, it is useful to note how

interface versions were handled as they interacted with update releases of

Solaris. Solaris update releases required careful coordination with the full

release currently under development to keep symbol versions constant between

releases.

Multiple update releases were generally shipped during the development of the

next full release of Solaris. It was impossible to know in advance the full

set of new interfaces in the next full release until it was complete. Some,

though not all, new interfaces were included in the intervening update

releases between full releases.

Consequently, the new version number for an update cannot be a minor

increment, but must be a micro increment to ensure that was a distinct

version between the two releases. For example, if Release N had version

number SUNW_1.3 and Release N+1 had SUNW_1.4, then interfaces added to

an update of Release N must have micro numbers such as SUNW_1.3.1,

SUNW_1.3.2, etc. (note that the micro number is not directly tied to the

update number: SUNW_1.3.1 may have appeared in Update 2). The micro versions form

an inheritance chain that is inserted between two successive minor versions.

For example, the mapfile-vers file for minor release "N+1" to reflect its

inclusion of micro releases will look like the following:

$mapfile_version 2

SYMBOL_VERSION SUNW_1.4 { # release N+1

global:

...

} SUNW_1.3.2;

SYMBOL_VERSION SUNW_1.3.2 { # micro release 2 (e.g., release NU3)

global:

...

} SUNW_1.3.1;

SYMBOL_VERSION SUNW_1.3.1 { # micro release 1 (e.g., release NU2)

global:

...

} SUNW_1.3;

SYMBOL_VERSION SUNW_1.3 { # release N

global:

...

} SUNW_1.2;

SYMBOL_VERSION SUNW_1.2 { # release N-1

global:

...

} SUNW_1.1;

SYMBOL_VERSION SUNW_1.1 { # first release

global:

...

};

SYMBOL_VERSION SUNW_private { # same in all releases

global:

...

local:

*;

};

The corresponding update/patch mapfile-vers file will be identical

except for the exclusion of SUNW_1.4.

Those interfaces which are only present in Release N+1 are always put

into the next minor version set, SUNW_1.4.

Thus when adding a new Public interface to an update release, both the mapfiles

of the update release and next full release should have been modified to be

consistent.

There have been several cases of accidental deviation from this scheme, and

existing mapfiles sometimes reflect this unfortunate state of affairs.

-------------------------------------------------------------------------------

5.0 How to update an interface in an existing library

5.1 Removing an existing interface

5.1.1 Moving a Public interface

No Public interfaces should ever be removed from any mapfile, as this will

break all existing consumers of that interface.

To move an interface from one library to (say) libc, the code has to be

deleted from the library and added to libc, then the mapfile for the

library has to have the interface's entry changed from:

getfoobar;

to:

getfoobar { TYPE = FUNCTION; FILTER = libc.so.1 };

This is an exception to the immutability of public symbol versions. See,

for example, libnsl's common/mapfile-vers file.

Follow the rules for adding a new interface for the necessary changes

to libc's mapfile to accommodate the moved interface, including creating a

new version in libc for the symbol.

When merging an entire library into libc, the mapfile is changed to specify

the type of each public symbol similarly to the above:

getfoobar;

to:

getfoobar { TYPE = FUNCTION };

But rather than specifying the library on which we filter for each symbol,

the link-editor is invoked with '-F libc.so.1' to specify that our entire

symbol table is a filter on libc. For examples, see libaio and librt.

5.1.2 Removing a Private interface

Deletion of Private interfaces is allowed, but caution should be taken;

it should first be established that the interface is not being used.

To remove a Private interface, simply delete the corresponding entry

for that symbol from the mapfile's private version section.

Do not forget to delete these Public or Private interfaces from the library's

header files as well as from the code that implements the interfaces.

5.2 Promoting a Private interface to Public

This is similar to what's done when adding a Public interface. Promoting an

existing Private interface to a Public one only requires a change to the

existing interface definition. Private interfaces have the symbol version

name "ILLUMOSprivate" or "SUNWprivate" associated with them. To make the

interface a Public one, the interface must be added as if it were a new

public symbol, following those same rules and removed from the private

version.

As an example, if we were modifying libwombat.so.1 and its existing latest

version were ILLUMOS_0.3, any new ABI would be put into a version called

ILLUMOS_0.4. Therefore, whether you wish to promote an existing Private

interface to Public, or to introduce a new Public interface, this (next

successive minor numbered version level) would be the version that it would

be associated with.

5.3 Scoping a Private interface local

Any interfaces not present in the mapfile-vers file will be scoped local

due to the presence of the

local:

*;

lines discussed earlier. This ensures that such interfaces will not be visible

outside the library. To move an interface from Private to local scope, simply

remove the Private interface from the mapfile-vers file and the header file

to prevent it from being exported. This may require moving the Private

interface into a library-private header file. Scope reduction of Public

interfaces is forbidden.

For the interface to be used in more than one file within the library, it

should be in a header file that can be included by each file in the library

that uses the interface. For example:

#include "libprivate.h"

5.4 How to copy interfaces which are part of a standard to a new or existing

library

SYSVABI and SISCD are reserved version names for interfaces listed in the

System V Interface Definition and the Sparc Compliance Definition. Avoid using

these version names when copying the implementation of standard interfaces to

another library. Instead, use ILLUMOS_0.1 for a new library, and ILLUMOS_m.n for

an existing library (where m.n is the next version; i.e., if the

last version was ILLUMOS_0.8, then you should version the interfaces with

ILLUMOS_0.9).

-------------------------------------------------------------------------------

6.0 Introducing a new library

6.1 Directories

The normal discipline for introducing a new library in illumos is to create a

new subdirectory of usr/src/lib. The interface definition discipline is to

create a common/mapfile-vers file for the new library. If we were introducing

a new foo library, libfoo, we'd create usr/src/lib/libfoo containing:

Makefile amd64/ i386/ sparcv9/

Makefile.com common/ sparc/

The common subdirectory would contain the normal source files plus the

mapfile-vers file. See usr/src/lib/README.Makefiles for directions on

how to organize the Makefiles.

6.2 The mapfile

The new common/mapfile-vers file would contain:

$mapfile_version 2

SYMBOL_VERSION ILLUMOS_0.1 { # first release of libfoo

global:

...

};

SYMBOL_VERSION ILLUMOSprivate {

global:

...

local:

*;

};

If there are no Public interfaces, the ILLUMOS_0.1 section would be omitted.

If there are no Private interfaces, the ILLUMOSprivate section would be

omitted and the two lines:

local:

*;

would be moved into ILLUMOS_0.1.

To decide which interfaces are Public (part of the ABI) and which are

Private (unstable interfaces not intended to be used by third parties), the

heuristic which works to a first approximation is that if it has a man page

then it's Public.

For maintainability, the list of names in each version block should

be sorted in dictionary order (sort -d). Please comply.

-------------------------------------------------------------------------------

7.0 Make an entire library obsolete

7.1 Introduce SUNWobsolete version

Use this version name not for specific interfaces but for marking an entire

library as obsolete. The existing public/private version names are left

unchanged, but a new SUNWobsolete version is created with no symbols in it.

This becomes a tag by which the obsolescence of the library can be recognized.

There is no numbering of this version name.

$mapfile_version 2

SYMBOL_VERSION SUNWobsolete {

global:

SUNWobsolete; # This is the only way to do it.

} SUNW_1.2;

SYMBOL_VERSION ILLUMOS_0.2 {

...

You should continue to use the name SUNWobsolete even in illumos.

-------------------------------------------------------------------------------

8.0 Documentation

For further information, please refer to the following documents:

"Solaris Linker and Libraries Guide"