openssl/NOTES-WINDOWS.md
Neil Horman bf74cf35cf Fixes for defaults code
Fix up some indenting, and ensure that the run_once routines don't get
defined if OSSL_WINCTX isn't defined to avoid compiler errors

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24450)
2024-07-09 04:01:44 -04:00

284 lines
10 KiB
Markdown

Notes for Windows platforms
===========================
- [Native builds using Visual C++](#native-builds-using-visual-c)
- [Native builds using Embarcadero C++Builder](
#native-builds-using-embarcadero-cbuilder)
- [Native builds using MinGW](#native-builds-using-mingw)
- [Linking native applications](#linking-native-applications)
- [Hosted builds using Cygwin](#hosted-builds-using-cygwin)
There are various options to build and run OpenSSL on the Windows platforms.
"Native" OpenSSL uses the Windows APIs directly at run time.
To build a native OpenSSL you can either use:
Microsoft Visual C++ (MSVC) C compiler on the command line
or
Embarcadero C++Builder
or
MinGW cross compiler
run on the GNU-like development environment MSYS2
or run on Linux or Cygwin
"Hosted" OpenSSL relies on an external POSIX compatibility layer
for building (using GNU/Unix shell, compiler, and tools) and at run time.
For this option, you can use Cygwin.
Native builds using Visual C++
==============================
The native builds using Visual C++ have a `VC-*` prefix.
Requirement details
-------------------
In addition to the requirements and instructions listed in `INSTALL.md`,
these are required as well:
### Perl
We recommend Strawberry Perl, available from <http://strawberryperl.com/>
Please read NOTES.PERL for more information, including the use of CPAN.
An alternative is ActiveState Perl, <https://www.activestate.com/ActivePerl>
for which you may need to explicitly build the Perl module Win32/Console.pm
via <https://platform.activestate.com/ActiveState> and then download it.
### Microsoft Visual C compiler.
Since these are proprietary and ever-changing we cannot test them all.
Older versions may not work. Use a recent version wherever possible.
### Netwide Assembler (NASM)
NASM is the only supported assembler. It is available from <https://www.nasm.us>.
Quick start
-----------
1. Install Perl
2. Install NASM
3. Make sure both Perl and NASM are on your %PATH%
4. Use Visual Studio Developer Command Prompt with administrative privileges,
choosing one of its variants depending on the intended architecture.
Or run `cmd` and execute `vcvarsall.bat` with one of the options `x86`,
`x86_amd64`, `x86_arm`, `x86_arm64`, `amd64`, `amd64_x86`, `amd64_arm`,
or `amd64_arm64`.
This sets up the environment variables needed for `nmake.exe`, `cl.exe`,
etc.
See also
<https://docs.microsoft.com/cpp/build/building-on-the-command-line>
5. From the root of the OpenSSL source directory enter
- `perl Configure VC-WIN32` if you want 32-bit OpenSSL or
- `perl Configure VC-WIN64A` if you want 64-bit OpenSSL or
- `perl Configure VC-WIN64-ARM` if you want Windows on Arm (win-arm64)
OpenSSL or
- `perl Configure VC-WIN64-CLANGASM-ARM` if you want Windows on Arm (win-arm64)
OpenSSL with assembly support using clang-cl as assembler or
- `perl Configure VC-CLANG-WIN64-CLANGASM-ARM` if you want Windows on Arm (win-arm64)
OpenSSL using clang-cl as both compiler and assembler or
- `perl Configure VC-WIN32-HYBRIDCRT` if you want 32-bit OpenSSL dependent
on the Universal CRT or
- `perl Configure VC-WIN64A-HYBRIDCRT` if you want 64-bit OpenSSL dependent
on the Universal CRT or
- `perl Configure` to let Configure figure out the platform
6. `nmake`
7. `nmake test`
8. `nmake install`
For the full installation instructions, or if anything goes wrong at any stage,
check the INSTALL.md file.
Installation directories
------------------------
On most Unix platforms installation directories are determined at build time via
constant defines. On Windows platforms however, installation directories are
determined via registry keys, as it is common practice to build OpenSSL and
install it to a variety of locations.
The following keys:
`\\HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\OpenSSL-<version>-<ctx>\OPENSSLDIR`
`\\HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\OpenSSL-<version>-<ctx>\ENGINESDIR`
`\\HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\OpenSSL-<version>-<ctx>\MODULESDIR`
Can be administratively set, and openssl will take the paths found there as the
values for OPENSSLDIR, ENGINESDIR and MODULESDIR respectively.
To enable the reading of registry keys from windows builds, add
`-DOPENSSL_WINCTX=<string>`to the Configure command line. This define is used
at build-time to construct library build specific registry key paths of the
format:
`\\HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432node\OpenSSL-<version>-<ctx>`
Where `<version>` is the major.minor version of the library being
built, and `<ctx>` is the value specified by `-DOPENSSL_WINCTX`. This allows
for multiple openssl builds to be created and installed on a single system, in
which each library can use its own set of registry keys.
Note the installer available at <https://github.com/openssl/installer> will set
these keys when the installer is run.
A summary table of behavior on Windows platforms
|`OSSL_WINCTX`|Registry key|OpenSSL Behavior |
|-------------|------------|------------------------------------------|
|Defined | Defined |OpenSSL Reads Paths from Registry |
|Defined | Undefined |OpenSSL returns errors on module/conf load|
|Undefined | N/A |OpenSSL uses build time defaults |
Special notes for Universal Windows Platform builds, aka `VC-*-UWP`
-------------------------------------------------------------------
- UWP targets only support building the static and dynamic libraries.
- You should define the platform type to `uwp` and the target arch via
`vcvarsall.bat` before you compile. For example, if you want to build
`arm64` builds, you should run `vcvarsall.bat x86_arm64 uwp`.
Native builds using Embarcadero C++Builder
=========================================
This toolchain (a descendant of Turbo/Borland C++) is an alternative to MSVC.
OpenSSL currently includes an experimental 32-bit configuration targeting the
Clang-based compiler (`bcc32c.exe`) in v10.3.3 Community Edition.
<https://www.embarcadero.com/products/cbuilder/starter>
1. Install Perl.
2. Open the RAD Studio Command Prompt.
3. Go to the root of the OpenSSL source directory and run:
`perl Configure BC-32 --prefix=%CD%`
4. `make -N`
5. `make -N test`
6. Build your program against this OpenSSL:
* Set your include search path to the "include" subdirectory of OpenSSL.
* Set your library search path to the OpenSSL source directory.
Note that this is very experimental. Support for 64-bit and other Configure
options is still pending.
Native builds using MinGW
=========================
MinGW offers an alternative way to build native OpenSSL, by cross compilation.
* Usually the build is done on Windows in a GNU-like environment called MSYS2.
MSYS2 provides GNU tools, a Unix-like command prompt,
and a UNIX compatibility layer for applications.
However, in this context it is only used for building OpenSSL.
The resulting OpenSSL does not rely on MSYS2 to run and is fully native.
Requirement details
- MSYS2 shell, from <https://www.msys2.org/>
- Perl, at least version 5.10.0, which usually comes pre-installed with MSYS2
- make, installed using `pacman -S make` into the MSYS2 environment
- MinGW[64] compiler: `mingw-w64-i686-gcc` and/or `mingw-w64-x86_64-gcc`.
These compilers must be on your MSYS2 $PATH.
A common error is to not have these on your $PATH.
The MSYS2 version of gcc will not work correctly here.
In the MSYS2 shell do the configuration depending on the target architecture:
./Configure mingw ...
or
./Configure mingw64 ...
or
./Configure ...
for the default architecture.
Apart from that, follow the Unix / Linux instructions in `INSTALL.md`.
* It is also possible to build mingw[64] on Linux or Cygwin.
In this case configure with the corresponding `--cross-compile-prefix=`
option. For example
./Configure mingw --cross-compile-prefix=i686-w64-mingw32- ...
or
./Configure mingw64 --cross-compile-prefix=x86_64-w64-mingw32- ...
This requires that you've installed the necessary add-on packages for
mingw[64] cross compilation.
Linking native applications
===========================
This section applies to all native builds.
If you link with static OpenSSL libraries, then you're expected to
additionally link your application with `WS2_32.LIB`, `GDI32.LIB`,
`ADVAPI32.LIB`, `CRYPT32.LIB` and `USER32.LIB`. Those developing
non-interactive service applications might feel concerned about
linking with `GDI32.LIB` and `USER32.LIB`, as they are justly associated
with interactive desktop, which is not available to service
processes. The toolkit is designed to detect in which context it's
currently executed, GUI, console app or service, and act accordingly,
namely whether to actually make GUI calls. Additionally, those
who wish to `/DELAYLOAD:GDI32.DLL` and `/DELAYLOAD:USER32.DLL` and
actually keep them off service process should consider implementing
and exporting from .exe image in question own `_OPENSSL_isservice` not
relying on `USER32.DLL`. E.g., on Windows Vista and later you could:
__declspec(dllexport) __cdecl BOOL _OPENSSL_isservice(void)
{
DWORD sess;
if (ProcessIdToSessionId(GetCurrentProcessId(), &sess))
return sess == 0;
return FALSE;
}
If you link with OpenSSL .DLLs, then you're expected to include into
your application code a small "shim" snippet, which provides
the glue between the OpenSSL BIO layer and your compiler run-time.
See also the OPENSSL_Applink manual page.
Hosted builds using Cygwin
==========================
Cygwin implements a POSIX/Unix runtime system (`cygwin1.dll`) on top of the
Windows subsystem and provides a Bash shell and GNU tools environment.
Consequently, a build of OpenSSL with Cygwin is virtually identical to the
Unix procedure.
To build OpenSSL using Cygwin, you need to:
* Install Cygwin, see <https://cygwin.com/>
* Install Cygwin Perl, at least version 5.10.0
and ensure it is in the $PATH
* Run the Cygwin Bash shell
Apart from that, follow the Unix / Linux instructions in INSTALL.md.
NOTE: `make test` and normal file operations may fail in directories
mounted as text (i.e. `mount -t c:\somewhere /home`) due to Cygwin
stripping of carriage returns. To avoid this, ensure that a binary
mount is used, e.g. `mount -b c:\somewhere /home`.