Remove the 1.67 upper-bound on compiler feature computation so that with
newer versions we at least get it mostly right. Leave a comment with
notes about updating features for future versions.
Restore path suffixes incorrectly removed by commit a62d50ec56 (Modules:
Replace coded PATHS with PATH_SUFFIXES, 2017-11-20, v3.11.0-rc1~293^2).
Hints do not participate in the usual bin/lib subdirectory search that
`<PackangeName>_ROOT` or `CMAKE_PREFIX_PATH` exhibit.
Fixes: #19185
The Gentoo case added by commit 1673923c30 (FindBoost: Add support for
Boost 1.67 with Python version suffixes, 2018-03-18, v3.11.0~3^2) left
out the `.` version component separator and instead duplicated the RPM
case. Add the missing `.` now.
Fixes: #18743
VS 2019 Update 1 will fix its redist directories to be named `VC142`
instead of `VC141`. It will also use cl `19.21` instead of `19.20`
so we can use that to distinguish the versions.
Fixes: #19131
In some cases GCC reports *relative* implicit include directories. They
are computed adaptively with respect to the current working directory
such that the effective implicit include directory is an unchanging
absolute path. Teach our implicit include directory extraction to
recognize such paths and normalize them.
Fixes: #19133
Since VS 2017's v141 toolset there is no longer a simple equation to
calculate the redist name, dll version, and VS IDE version from just the
MSVC toolset version. Refactor the logic to use hard-coded values and
warn when a new version is not supported.
Fixes: #19125
Since commit 8f8d056051 (ARMCC: Fix identification of ARM compiler when
it defines GNU macros, 2019-03-20, v3.14.1~10^2) we consider ARMCC
before Clang or GNU compilers. Since armclang also defines
`__ARMCC_VERSION` it is now mistaken for ARMCC. Extend the check for
ARMCC to also verify that `__clang__` is not defined.
Issue: #19065
This was forgotten in commit 626c51f47b (VS: Update for Visual Studio
2019 Preview 2, 2019-01-24, v3.14.0-rc1~74^2) when the toolset was
first renumbered to `v142`.
Issue: #19125
The implicit include directory extraction added by commit 5990ecb741
(Compute implicit include directories from compiler output, 2018-12-07,
v3.14.0-rc1~108^2) leaves paths like `/usr/lib/../include` unchanged.
Fix the logic to canonicalize such paths (e.g. to `/usr/include`)
as we do for implicit link directories already. This is important
to ensure the set of implicit directories is represented in the same
form as the include directories that will be compared to them.
Issue: #19095
Both commit 88ecfd8ba1 (CPackIFW: Add some options, 2016-11-11,
v3.8.0-rc1~248^2) and commit e5089c562c (CPackIFW: Add some options,
2017-01-24, v3.8.0-rc1~53^2) added some options to this macro's
documentation and implementation, but the actual processing of the
parsed options was never added. Add it now.
Reviewed-by: Konstantin Podsvirov <konstantin@podsvirov.pro>
Our documented standard for find module variable names is to match the
case of the find module package name. This was overlooked when the
module was added by commit 84e7920b3a (FindFontconfig: Add module to
find Fontconfig, 2018-09-27, v3.14.0-rc1~523^2).
The module was released with the upper case names in CMake 3.14.0, so
fix it to have camel case names in 3.14.1. This is incompatible but
anyone using a given release series should be using the latest patch on
it and we've made breaking fixups on newly released features like this
before.
Reported-by: Christophe Giboudeaux <christophe@krop.fr>
Fixes: #19094
Any `include_directories()` calls in toolchain files are used during our
ABI detection step even though it does not include any system headers.
Since commit 5990ecb741 (Compute implicit include directories from
compiler output, 2018-12-07, v3.14.0-rc1~108^2), that check is also used
to detect implicit include directories. Any `include_directories()` in
a toolchain file are detected as implicit and later excluded from
explicit specification on compiler command lines, thus breaking the
purpose of the calls in the first place.
Fix the implicit include directory detection step to avoid using paths
from `include_directories()` calls in the toolchain file.
Fixes: #19079
Since commit 11da882a12 (Apple: Introduce separate system name for iOS,
tvOS, and watchOS, 2018-01-15, v3.14.0-rc1~14^2~1) we support setting
`CMAKE_SYSTEM_NAME` to `iOS`. Existing iOS toolchain files already
set `IOS` as a short-hand variable, so do the same here.
Update the component added by commit 513e77550d (FindPython: Introduce
NumPy component, 2018-12-12, v3.14.0-rc1~95^2). The `numpy/`
sub-directory should not be part of the include directory. It should be
part of the `#include` line.
Java 9 restructured the standard location of the AWT libraries due to
the removal of the JRE/JDK separation. We should check all possible
combinations of subdirectories to the Java root directories to ensure
that the libraries will be found after an upgrade.
Furthermore, a root directory would contain both, include and library
paths, so the search should be unified to ease maintenance on the
module.
In commit b6f6cac378 (ExternalProject: add LOG_DIR option that allows
overriding of log location, 2018-10-12, v3.14.0-rc1~515^2~1) the log
directory got its own option. The intention was to fall back to the
stamp directory by default. However, the implementation actually only
falls back to the same default as the stamp directory and does not
consider a custom stamp dir.
Update the default log dir computation to fall back to whatever is the
final selection for the stamp dir.
Fixes: #19000
Since commit e9a1ddc594 (FindThreads: Replace the pthread symbol
checking in libc., 2018-11-18, v3.14.0-rc1~292^2) we check libc for
`pthread_kill` instead of `pthread_create`. However, on FreeBSD
`pthread_kill` is in libc but not `pthread_create`. Discussion in the
original merge request for the above commit also considered
`pthread_key_create`, `pthread_self`, and `pthread_attr_init`. Every
symbol seems to have some reason it is not an appropriate choice.
Revert to the pre-3.14 behavior of using `pthread_create` pending
further investigation.
The `FindOctave` module added by commit 170bcb6fdc (FindOctave: Add
module to find GNU octave, 2018-11-17, v3.14.0-rc1~283^2) has a few
problems in its implementation that need to be worked out before the
module can be included in a CMake release. These were missed during
review. Remove the module for now. It can be restored later with a
fresh review.
Issue: #18991
When using a VS generator with `-T llvm`, MSBuild relies on the "LLVM
Compiler Toolchain" VS Extension. This does not put `clang-cl` in the
`PATH` inside the build, and LLVM no longer provides a `cl` replacement
either. Therefore we need another way to extract the path to the
`CMAKE_{C,CXX}_COMPILER`. Fortunately the LLVM VS integration provides
a `$(ClangClExecutable)` macro we can reference to get the path.
Fixes: #18983
In commit e9a1ddc594 (FindThreads: Replace the pthread symbol checking
in libc., 2018-11-18, v3.14.0-rc1~292^2) we switched to checking for
`pthread_kill` in libc but did not update the symbol check's header file
to match. Add `signal.h` to get `pthread_kill`. Keep `pthread.h`
anyway since the purpose of the check is to verify that the pthread API
works.
Fixes: #18984
Alex Turbov