| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
Before this patch `NDEBUG` was force-disabled, preventing a build with
debug asserts disabled.
After this patch `NDEBUG` works again when passed as a custom build
option, e.g.: `-DCMAKE_C_FLAGS=-DNDEBUG`
Previously submitted as #988, which was merged, but the commit vanished
from master and ended up missing from both 3.8.3 and 3.9.0 releases.
|
|\ |
|
| |
| |
| |
| |
| | |
With ASM support the builds either exit with an assert or hang
(with asserts disabled).
|
|\ \ |
|
| |/
| |
| |
| |
| |
| |
| |
| | |
Visual studio does not define __BYTE_ORDER__ so all architectures
were detected as LITTLE_ENDIAN since both __BYTE_ORDER__ and
__ORDER_LITTLE_ENDIAN__ would evaluate to 0 and compare equal. This
updates the checks to use CMakes detection of endianness, with a hard
error, if this also fails.
|
|/
|
|
|
| |
As it enables a lot of spammed warnings that are not part of W4. This
reduces the warnings a lot when compiling LibreSSL in CLion for me.
|
| |
|
|
|
|
|
|
| |
Syncing this up with autotools.
Also use the built-in `MINGW` variable.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Notice that just like in autotools, this detection also doesn't take
into account the targeted OS version. Meaning it detects `strtonum` even
if targeting e.g. macOS older than release v11 Big Sur (which introduced
this funcitions), if the SDK declares it. Wrong detection will either
cause a binary broken on older macOS and/or trigger compiler warnings.
Ref: https://github.com/libressl/portable/issues/928#issuecomment-1850178282
Ref: https://github.com/libressl/portable/issues/928#issuecomment-1850276298
Prerequisite:
https://github.com/libressl/portable/issues/928#issuecomment-1850356408
|
|\ |
|
| |
| |
| |
| |
| |
| | |
This makes it accept values consistently. Before this patch
mingw-w64, Apple and SunOS did not accept a CPU if it had
a suffix or prefix (e.g. a triplet), while other targets did.
|
| | |
|
|/
|
|
|
|
|
|
| |
- `-DCPPFLAGS`: probably a copy-paste typo from the initial CMake
commit.
- `-DNO_CRYPT`: `NO_CRYPT` is no longer used in the source and this
macro is no longer set by autotools.
|
|
|
|
|
| |
Reported here:
https://github.com/libressl/portable/pull/935#issuecomment-1798345787
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of including a full copy of libcrypto and libssl in
libtls-static, link existing libcrytpo-static and libssl-static to
the test targets.
This wasn't causing any issue, just unnecessarily duplicating
a lot of objects.
|
|/ |
|
| |
|
|
|
|
| |
2022's preprocessor
|
|\ |
|
| | |
|
| | |
|
| | |
|
|/ |
|
| |
|
|
|
|
|
| |
This allows users to install the CMake configs without forcing a
`LibreSSL` directory after `LIBRESSL_INSTALL_CMAKEDIR`.
|
|
|
|
|
|
|
|
| |
The configs can be consumed by setting LibreSSL_DIR to the build
directory, or after installation using CMAKE_PREFIX_PATH/LibreSSL_DIR.
For compatibility, the EXPORT_NAME of targets and the LIBRESSL_*
variables are set to match the names used in FindLibreSSL.
|
|\ |
|
| | |
|
| | |
|
|/ |
|
|
|
|
|
| |
Fixes #683. These are not compiler-agnostic, and can interfere with user
overrides as well. The defaults in Cmake are reasonable.
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This adds a getopt implementation for compatibility where it is not
available, removing a couple of regress patches.
Note, this is a slightly modified copy from OpenBSD libc that doesn't
expose getopt_long, which has dependency conflicts with Windows system
headers and isn't needed anyway.
|
|/
|
|
|
|
|
|
|
|
|
| |
based on discussion in https://github.com/libressl/portable/issues/266
and https://bugs.python.org/issue23524 adjust the compat layer for
Windows to use _get_osfhandle in combination with
_set_thread_local_invalid_parameter_handler if applicable to more
reliably determine if a handle is a socket, file, or closed socket.
This prevents assertions when calling tls_close on an already-closed
socket.
|
|
|
|
|
|
| |
Sync CMAKE_SYSTEM_PROCESSOR to CMAKE_OSX_ARCHITECTURES. This doesn't
support universal binaries, but does allow cross-compiling for a single
architecture by setting -DCMAKE_OSX_ARCHITECTURES=(arm64|x86_64)
|
| |
|
| |
|
| |
|
|
|
|
| |
handle x86 as option for 32-bit x86 on Windows
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|