openssl/Configurations/README
Richard Levitte 473a954771 Document how to force linking with static libraries
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1889)
2016-11-10 15:51:43 +01:00

674 lines
30 KiB
Text

Configurations of OpenSSL target platforms
==========================================
Target configurations are a collection of facts that we know about
different platforms and their capabilities. We organise them in a
hash table, where each entry represent a specific target.
In each table entry, the following keys are significant:
inherit_from => Other targets to inherit values from.
Explained further below. [1]
template => Set to 1 if this isn't really a platform
target. Instead, this target is a template
upon which other targets can be built.
Explained further below. [1]
sys_id => System identity for systems where that
is difficult to determine automatically.
cc => The C compiler command, usually one of "cc",
"gcc" or "clang". This command is normally
also used to link object files and
libraries into the final program.
cxx => The C++ compiler command, usually one of
"c++", "g++" or "clang++". This command is
also used when linking a program where at
least one of the object file is made from
C++ source.
cflags => Flags that are used at all times when
compiling C object files.
cxxflags => Flags that are used at all times when
compiling C++ object files. If unset, it
gets the same value as cflags.
defines => As an alternative, macro definitions may be
present here instead of in `cflags'. If
given here, they MUST be as an array of the
string such as "MACRO=value", or just
"MACRO" for definitions without value.
shared_cflag => Extra compilation flags used when
compiling for shared libraries, typically
something like "-fPIC".
(linking is a complex thing, see [3] below)
ld => Linker command, usually not defined
(meaning the compiler command is used
instead).
(NOTE: this is here for future use, it's
not implemented yet)
lflags => Flags that are used when linking apps.
shared_ldflag => Flags that are used when linking shared
or dynamic libraries.
plib_lflags => Extra linking flags to appear just before
the libraries on the command line.
ex_libs => Extra libraries that are needed when
linking.
ar => The library archive command, the default is
"ar".
(NOTE: this is here for future use, it's
not implemented yet)
arflags => Flags to be used with the library archive
command.
ranlib => The library archive indexing command, the
default is 'ranlib' it it exists.
unistd => An alternative header to the typical
'<unistd.h>'. This is very rarely needed.
shared_extension => File name extension used for shared
libraries.
obj_extension => File name extension used for object files.
On unix, this defaults to ".o" (NOTE: this
is here for future use, it's not
implemented yet)
exe_extension => File name extension used for executable
files. On unix, this defaults to "" (NOTE:
this is here for future use, it's not
implemented yet)
thread_scheme => The type of threads is used on the
configured platform. Currently known
values are "(unknown)", "pthreads",
"uithreads" (a.k.a solaris threads) and
"winthreads". Except for "(unknown)", the
actual value is currently ignored but may
be used in the future. See further notes
below [2].
dso_scheme => The type of dynamic shared objects to build
for. This mostly comes into play with
engines, but can be used for other purposes
as well. Valid values are "DLFCN"
(dlopen() et al), "DLFCN_NO_H" (for systems
that use dlopen() et al but do not have
fcntl.h), "DL" (shl_load() et al), "WIN32"
and "VMS".
perlasm_scheme => The perlasm method used to created the
assembler files used when compiling with
assembler implementations.
shared_target => The shared library building method used.
This is a target found in Makefile.shared.
build_scheme => The scheme used to build up a Makefile.
In its simplest form, the value is a string
with the name of the build scheme.
The value may also take the form of a list
of strings, if the build_scheme is to have
some options. In this case, the first
string in the list is the name of the build
scheme.
Currently recognised build scheme is "unified".
For the "unified" build scheme, this item
*must* be an array with the first being the
word "unified" and the second being a word
to identify the platform family.
multilib => On systems that support having multiple
implementations of a library (typically a
32-bit and a 64-bit variant), this is used
to have the different variants in different
directories.
bn_ops => Building options (was just bignum options
in the earlier history of this option,
hence the name). This a string of words
that describe properties on the designated
target platform, such as the type of
integers used to build up the bitnum,
different ways to implement certain ciphers
and so on. To fully comprehend the
meaning, the best is to read the affected
source.
The valid words are:
BN_LLONG use 'unsigned long long' in
some bignum calculations.
This has no value when
SIXTY_FOUR_BIT or
SIXTY_FOUR_BIT_LONG is given.
RC4_CHAR makes the basic RC4 unit of
calculation an unsigned char.
SIXTY_FOUR_BIT processor registers
are 64 bits, long is
32 bits, long long is
64 bits.
SIXTY_FOUR_BIT_LONG processor registers
are 64 bits, long is
64 bits.
THIRTY_TWO_BIT processor registers
are 32 bits.
EXPORT_VAR_AS_FN for shared libraries,
export vars as
accessor functions.
apps_extra_src => Extra source to build apps/openssl, as
needed by the target.
cpuid_asm_src => assembler implementation of cpuid code as
well as OPENSSL_cleanse().
Default to mem_clr.c
bn_asm_src => Assembler implementation of core bignum
functions.
Defaults to bn_asm.c
ec_asm_src => Assembler implementation of core EC
functions.
des_asm_src => Assembler implementation of core DES
encryption functions.
Defaults to 'des_enc.c fcrypt_b.c'
aes_asm_src => Assembler implementation of core AES
functions.
Defaults to 'aes_core.c aes_cbc.c'
bf_asm_src => Assembler implementation of core BlowFish
functions.
Defaults to 'bf_enc.c'
md5_asm_src => Assembler implementation of core MD5
functions.
sha1_asm_src => Assembler implementation of core SHA1,
functions, and also possibly SHA256 and
SHA512 ones.
cast_asm_src => Assembler implementation of core CAST
functions.
Defaults to 'c_enc.c'
rc4_asm_src => Assembler implementation of core RC4
functions.
Defaults to 'rc4_enc.c rc4_skey.c'
rmd160_asm_src => Assembler implementation of core RMD160
functions.
rc5_asm_src => Assembler implementation of core RC5
functions.
Defaults to 'rc5_enc.c'
wp_asm_src => Assembler implementation of core WHIRLPOOL
functions.
cmll_asm_src => Assembler implementation of core CAMELLIA
functions.
Defaults to 'camellia.c cmll_misc.c cmll_cbc.c'
modes_asm_src => Assembler implementation of cipher modes,
currently the functions gcm_gmult_4bit and
gcm_ghash_4bit.
padlock_asm_src => Assembler implementation of core parts of
the padlock engine. This is mandatory on
any platform where the padlock engine might
actually be built.
[1] as part of the target configuration, one can have a key called
'inherit_from' that indicate what other configurations to inherit
data from. These are resolved recursively.
Inheritance works as a set of default values that can be overridden
by corresponding key values in the inheriting configuration.
Note 1: any configuration table can be used as a template.
Note 2: pure templates have the attribute 'template => 1' and
cannot be used as build targets.
If several configurations are given in the 'inherit_from' array,
the values of same attribute are concatenated with space
separation. With this, it's possible to have several smaller
templates for different configuration aspects that can be combined
into a complete configuration.
instead of a scalar value or an array, a value can be a code block
of the form 'sub { /* your code here */ }'. This code block will
be called with the list of inherited values for that key as
arguments. In fact, the concatenation of strings is really done
by using 'sub { join(" ",@_) }' on the list of inherited values.
An example:
"foo" => {
template => 1,
haha => "ha ha",
hoho => "ho",
ignored => "This should not appear in the end result",
},
"bar" => {
template => 1,
haha => "ah",
hoho => "haho",
hehe => "hehe"
},
"laughter" => {
inherit_from => [ "foo", "bar" ],
hehe => sub { join(" ",(@_,"!!!")) },
ignored => "",
}
The entry for "laughter" will become as follows after processing:
"laughter" => {
haha => "ha ha ah",
hoho => "ho haho",
hehe => "hehe !!!",
ignored => ""
}
[2] OpenSSL is built with threading capabilities unless the user
specifies 'no-threads'. The value of the key 'thread_scheme' may
be "(unknown)", in which case the user MUST give some compilation
flags to Configure.
[3] OpenSSL has three types of things to link from object files or
static libraries:
- shared libraries; that would be libcrypto and libssl.
- shared objects (sometimes called dynamic libraries); that would
be the engines.
- applications; those are apps/openssl and all the test apps.
Very roughly speaking, linking is done like this (words in braces
represent the configuration settings documented at the beginning
of this file):
shared libraries:
{ld} $(CFLAGS) {shared_ldflag} -shared -o libfoo.so \
-Wl,--whole-archive libfoo.a -Wl,--no-whole-archive \
{plib_lflags} -lcrypto {ex_libs}
shared objects:
{ld} $(CFLAGS) {shared_ldflag} -shared -o libeng.so \
blah1.o blah2.o {plib_lflags} -lcrypto {ex_libs}
applications:
{ld} $(CFLAGS) {lflags} -o app \
app1.o utils.o {plib_lflags} -lssl -lcrypto {ex_libs}
Historically, the target configurations came in form of a string with
values separated by colons. This use is deprecated. The string form
looked like this:
"target" => "{cc}:{cflags}:{unistd}:{thread_cflag}:{sys_id}:{lflags}:{bn_ops}:{cpuid_obj}:{bn_obj}:{ec_obj}:{des_obj}:{aes_obj}:{bf_obj}:{md5_obj}:{sha1_obj}:{cast_obj}:{rc4_obj}:{rmd160_obj}:{rc5_obj}:{wp_obj}:{cmll_obj}:{modes_obj}:{padlock_obj}:{perlasm_scheme}:{dso_scheme}:{shared_target}:{shared_cflag}:{shared_ldflag}:{shared_extension}:{ranlib}:{arflags}:{multilib}"
Build info files
================
The build.info files that are spread over the source tree contain the
minimum information needed to build and distribute OpenSSL. It uses a
simple and yet fairly powerful language to determine what needs to be
built, from what sources, and other relationships between files.
For every build.info file, all file references are relative to the
directory of the build.info file for source files, and the
corresponding build directory for built files if the build tree
differs from the source tree.
When processed, every line is processed with the perl module
Text::Template, using the delimiters "{-" and "-}". The hashes
%config and %target are passed to the perl fragments, along with
$sourcedir and $builddir, which are the locations of the source
directory for the current build.info file and the corresponding build
directory, all relative to the top of the build tree.
To begin with, things to be built are declared by setting specific
variables:
PROGRAMS=foo bar
LIBS=libsomething
ENGINES=libeng
SCRIPTS=myhack
EXTRA=file1 file2
Note that the files mentioned for PROGRAMS, LIBS and ENGINES *must* be
without extensions. The build file templates will figure them out.
For each thing to be built, it is then possible to say what sources
they are built from:
PROGRAMS=foo bar
SOURCE[foo]=foo.c common.c
SOURCE[bar]=bar.c extra.c common.c
It's also possible to tell some other dependencies:
DEPEND[foo]=libsomething
DEPEND[libbar]=libsomethingelse
(it could be argued that 'libsomething' and 'libsomethingelse' are
source as well. However, the files given through SOURCE are expected
to be located in the source tree while files given through DEPEND are
expected to be located in the build tree)
It's also possible to depend on static libraries explicitely:
DEPEND[foo]=libsomething.a
DEPEND[libbar]=libsomethingelse.a
This should be rarely used, and care should be taken to make sure it's
only used when supported. For example, native Windows build doesn't
support build static libraries and DLLs at the same time, so using
static libraries on Windows can only be done when configured
'no-shared'.
For some libraries, we maintain files with public symbols and their
slot in a transfer vector (important on some platforms). It can be
declared like this:
ORDINALS[libcrypto]=crypto
The value is not the name of the file in question, but rather the
argument to util/mkdef.pl that indicates which file to use.
One some platforms, shared libraries come with a name that's different
from their static counterpart. That's declared as follows:
SHARED_NAME[libfoo]=cygfoo-{- $config{shlibver} -}
The example is from Cygwin, which has a required naming convention.
Sometimes, it makes sense to rename an output file, for example a
library:
RENAME[libfoo]=libbar
That lines has "libfoo" get renamed to "libbar". While it makes no
sense at all to just have a rename like that (why not just use
"libbar" everywhere?), it does make sense when it can be used
conditionally. See a little further below for an example.
In some cases, it's desirable to include some source files in the
shared form of a library only:
SHARED_SOURCE[libfoo]=dllmain.c
For any file to be built, it's also possible to tell what extra
include paths the build of their source files should use:
INCLUDE[foo]=include
In some cases, one might want to generate some source files from
others, that's done as follows:
GENERATE[foo.s]=asm/something.pl $(CFLAGS)
GENERATE[bar.s]=asm/bar.S
The value of each GENERATE line is a command line or part of it.
Configure places no rules on the command line, except the the first
item muct be the generator file. It is, however, entirely up to the
build file template to define exactly how those command lines should
be handled, how the output is captured and so on.
Sometimes, the generator file itself depends on other files, for
example if it is a perl script that depends on other perl modules.
This can be expressed using DEPEND like this:
DEPEND[asm/something.pl]=../perlasm/Foo.pm
There may also be cases where the exact file isn't easily specified,
but an inclusion directory still needs to be specified. INCLUDE can
be used in that case:
INCLUDE[asm/something.pl]=../perlasm
NOTE: GENERATE lines are limited to one command only per GENERATE.
As a last resort, it's possible to have raw build file lines, between
BEGINRAW and ENDRAW lines as follows:
BEGINRAW[Makefile(unix)]
haha.h: {- $builddir -}/Makefile
echo "/* haha */" > haha.h
ENDRAW[Makefile(unix)]
The word within square brackets is the build_file configuration item
or the build_file configuration item followed by the second word in the
build_scheme configuration item for the configured target within
parenthesis as shown above. For example, with the following relevant
configuration items:
build_file => "build.ninja"
build_scheme => [ "unified", "unix" ]
... these lines will be considered:
BEGINRAW[build.ninja]
build haha.h: echo "/* haha */" > haha.h
ENDRAW[build.ninja]
BEGINRAW[build.ninja(unix)]
build hoho.h: echo "/* hoho */" > hoho.h
ENDRAW[build.ninja(unix)]
Should it be needed because the recipes within a RAW section might
clash with those generated by Configure, it's possible to tell it
not to generate them with the use of OVERRIDES, for example:
SOURCE[libfoo]=foo.c bar.c
OVERRIDES=bar.o
BEGINRAW[Makefile(unix)]
bar.o: bar.c
$(CC) $(CFLAGS) -DSPECIAL -c -o $@ $<
ENDRAW[Makefile(unix)]
See the documentation further up for more information on configuration
items.
Finally, you can have some simple conditional use of the build.info
information, looking like this:
IF[1]
something
ELSIF[2]
something other
ELSE
something else
ENDIF
The expression in square brackets is interpreted as a string in perl,
and will be seen as true if perl thinks it is, otherwise false. For
example, the above would have "something" used, since 1 is true.
Together with the use of Text::Template, this can be used as
conditions based on something in the passed variables, for example:
IF[{- $disabled{shared} -}]
LIBS=libcrypto
SOURCE[libcrypto]=...
ELSE
LIBS=libfoo
SOURCE[libfoo]=...
ENDIF
or:
# VMS has a cultural standard where all libraries are prefixed.
# For OpenSSL, the choice is 'ossl_'
IF[{- $config{target} =~ /^vms/ -}]
RENAME[libcrypto]=ossl_libcrypto
RENAME[libssl]=ossl_libssl
ENDIF
Build-file programming with the "unified" build system
======================================================
"Build files" are called "Makefile" on Unix-like operating systems,
"descrip.mms" for MMS on VMS, "makefile" for nmake on Windows, etc.
To use the "unified" build system, the target configuration needs to
set the three items 'build_scheme', 'build_file' and 'build_command'.
In the rest of this section, we will assume that 'build_scheme' is set
to "unified" (see the configurations documentation above for the
details).
For any name given by 'build_file', the "unified" system expects a
template file in Configurations/ named like the build file, with
".tmpl" appended, or in case of possible ambiguity, a combination of
the second 'build_scheme' list item and the 'build_file' name. For
example, if 'build_file' is set to "Makefile", the template could be
Configurations/Makefile.tmpl or Configurations/unix-Makefile.tmpl.
In case both Configurations/unix-Makefile.tmpl and
Configurations/Makefile.tmpl are present, the former takes
precedence.
The build-file template is processed with the perl module
Text::Template, using "{-" and "-}" as delimiters that enclose the
perl code fragments that generate configuration-dependent content.
Those perl fragments have access to all the hash variables from
configdata.pem.
The build-file template is expected to define at least the following
perl functions in a perl code fragment enclosed with "{-" and "-}".
They are all expected to return a string with the lines they produce.
generatesrc - function that produces build file lines to generate
a source file from some input.
It's called like this:
generatesrc(src => "PATH/TO/tobegenerated",
generator => [ "generatingfile", ... ]
generator_incs => [ "INCL/PATH", ... ]
generator_deps => [ "dep1", ... ]
generator => [ "generatingfile", ... ]
incs => [ "INCL/PATH", ... ],
deps => [ "dep1", ... ],
intent => one of "libs", "dso", "bin" );
'src' has the name of the file to be generated.
'generator' is the command or part of command to
generate the file, of which the first item is
expected to be the file to generate from.
generatesrc() is expected to analyse and figure out
exactly how to apply that file and how to capture
the result. 'generator_incs' and 'generator_deps'
are include directories and files that the generator
file itself depends on. 'incs' and 'deps' are
include directories and files that are used if $(CC)
is used as an intermediary step when generating the
end product (the file indicated by 'src'). 'intent'
indicates what the generated file is going to be
used for.
src2obj - function that produces build file lines to build an
object file from source files and associated data.
It's called like this:
src2obj(obj => "PATH/TO/objectfile",
srcs => [ "PATH/TO/sourcefile", ... ],
deps => [ "dep1", ... ],
incs => [ "INCL/PATH", ... ]
intent => one of "lib", "dso", "bin" );
'obj' has the intended object file *without*
extension, src2obj() is expected to add that.
'srcs' has the list of source files to build the
object file, with the first item being the source
file that directly corresponds to the object file.
'deps' is a list of explicit dependencies. 'incs'
is a list of include file directories. Finally,
'intent' indicates what this object file is going
to be used for.
obj2lib - function that produces build file lines to build a
static library file ("libfoo.a" in Unix terms) from
object files.
called like this:
obj2lib(lib => "PATH/TO/libfile",
objs => [ "PATH/TO/objectfile", ... ]);
'lib' has the intended library file name *without*
extension, obj2lib is expected to add that. 'objs'
has the list of object files (also *without*
extension) to build this library.
libobj2shlib - function that produces build file lines to build a
shareable object library file ("libfoo.so" in Unix
terms) from the corresponding static library file
or object files.
called like this:
libobj2shlib(shlib => "PATH/TO/shlibfile",
lib => "PATH/TO/libfile",
objs => [ "PATH/TO/objectfile", ... ],
deps => [ "PATH/TO/otherlibfile", ... ],
ordinals => [ "word", "/PATH/TO/ordfile" ]);
'lib' has the intended library file name *without*
extension, libobj2shlib is expected to add that.
'shlib' has the corresponding shared library name
*without* extension. 'deps' has the list of other
libraries (also *without* extension) this library
needs to be linked with. 'objs' has the list of
object files (also *without* extension) to build
this library. 'ordinals' MAY be present, and when
it is, its value is an array where the word is
"crypto" or "ssl" and the file is one of the ordinal
files util/libeay.num or util/ssleay.num in the
source directory.
This function has a choice; it can use the
corresponding static library as input to make the
shared library, or the list of object files.
obj2dso - function that produces build file lines to build a
dynamic shared object file from object files.
called like this:
obj2dso(lib => "PATH/TO/libfile",
objs => [ "PATH/TO/objectfile", ... ],
deps => [ "PATH/TO/otherlibfile",
... ]);
This is almost the same as libobj2shlib, but the
intent is to build a shareable library that can be
loaded in runtime (a "plugin"...). The differences
are subtle, one of the most visible ones is that the
resulting shareable library is produced from object
files only.
obj2bin - function that produces build file lines to build an
executable file from object files.
called like this:
obj2bin(bin => "PATH/TO/binfile",
objs => [ "PATH/TO/objectfile", ... ],
deps => [ "PATH/TO/libfile", ... ]);
'bin' has the intended executable file name
*without* extension, obj2bin is expected to add
that. 'objs' has the list of object files (also
*without* extension) to build this library. 'deps'
has the list of library files (also *without*
extension) that the programs needs to be linked
with.
in2script - function that produces build file lines to build a
script file from some input.
called like this:
in2script(script => "PATH/TO/scriptfile",
sources => [ "PATH/TO/infile", ... ]);
'script' has the intended script file name.
'sources' has the list of source files to build the
resulting script from.
In all cases, file file paths are relative to the build tree top, and
the build file actions run with the build tree top as current working
directory.
Make sure to end the section with these functions with a string that
you thing is appropriate for the resulting build file. If nothing
else, end it like this:
""; # Make sure no lingering values end up in the Makefile
-}