Remove the stdio handle inheritance suppression originally added by
commit f262298bb0 (... do not inherit pipes in child procs for ctest so
it can kill them, 2007-09-11, v2.6.0~1136). It's not clear what problem
it was trying to solve, was only done in `ctest` and not `cmake`, and
since commit 9c3ffe2474 (BUG: fix problem with stdout and stderr not
showing up in ms dos shells, 2007-09-25, v2.6.0~1066) has not been done
in `ctest` launched under interactive consoles.
Furthermore, the code has been spuriously breaking stdio when `ctest` is
started with both stdout and stderr connected to the same pipe, such as
when `ctest --launch` is used under `ninja`. This is because it used
`DuplicateHandle` with `DUPLICATE_CLOSE_SOURCE` on the stdout handle and
then the stderr handle. If the handles are the same, then the stderr
handle becomes invalid in between these operations, leading to
likely-undefined behavior. Since commit 96b3dd329e
(cmCTestLaunchReporter: Replace cmsysProcess with cmUVProcessChain,
2023-07-26, v3.28.0-rc1~138^2~2) this became more noticeable because
`uv_spawn` performs additional verification on stdio handles.
This could be fixed by instead suppressing inheritance via
SetHandleInformation(h, HANDLE_FLAG_INHERIT, 0);
However, the functionality no longer seems necessary, so remove it.
When CMake is built against the MSYS runtime library, `uname()` returns
a name that depends on the `MSYSTEM` environment variable. Previously
we truncated `MSYS_...` to just `MSYS`, but outside `MSYSTEM=MSYS`
environments, names like `MINGW64_NT-10.0-22000` were reported.
The latter causes CMake to report an unsupported-platform error, which
users report as an issue when the real problem is that they should be
using a `MSYSTEM=MSYS` environment or a CMake that is not built against
the MSYS runtime.
For our purposes, if CMake is built against the MSYS runtime, the host
platform is always `MSYS`. Similarly for `CYGWIN`.
Extend commit c85524a94a (Ensure stdin, stdout, and stderr pipes are always
open, 2019-05-02, v3.15.0-rc1~171^2) to cover the `stdin`, `stdout`, and
`stderr` FILE streams from `<stdio.h>`.
Issue: #25625
Wide use of CMake 3.28.{1,0[-rcN]} has uncovered some hangs and crashes
in libuv SIGCHLD handling on some platforms, particularly in virtualization
environments on macOS hosts. Although the bug does not seem to be in CMake,
we can restore stability in the CMake 3.28 release series for users of such
platforms by reverting our new uses of libuv for process execution.
Revert implementation changes merged by commit 4771544386 (Merge topic
'replace-cmsysprocess-with-cmuvprocesschain', 2023-09-06, v3.28.0-rc1~138),
but keep test suite updates.
Issue: #25414, #25500, #25562, #25589
`isspace` takes `int` but documents that the value must be representable
by `unsigned char`, or be EOF. Use a wrapper to cast to `unsigned char`
to avoid sign extension while converting to `int`. This generalizes the
fix from commit 5e8c176e2a (cmExecuteProcessCommand: Cast c to unsigned
char before cast to int, 2024-01-05) to other `isspace` call sites.
This was detected by assertions in the MSVC standard library while
processing UTF-8 text.
Issue: #25561
`include-what-you-use` diagnostics, in practice, are specific to
the environment's compiler and standard library. Update includes
to satisfy IWYU for our CI job under Debian 12.
The VersionCompare() function converted version components to
'unsigned long' prior to comparing them. Any version components
too large for 'unsigned long' were treated as equal to ULONG_MAX.
This impacted operators like VERSION_GREATER, VERSION_LESS, and
VERSION_EQUAL. The new code does not limit the length of the
version components for valid comparisons.
Run the `clang-format.bash` script to update all our C and C++ code to a
new style defined by `.clang-format`. Use `clang-format` version 15.
* If you reached this commit for a line in `git blame`, re-run the blame
operation starting at the parent of this commit to see older history
for the content.
* See the parent commit for instructions to rebase a change across this
style transition commit.
Fixes: #24315
On Windows, a file may be inaccessible for a short time after it is
created. This occurs for various reasons, including indexing, antivirus
tools, and NTFS's asynchronous semantics. Add an `INPUT_MAY_BE_RECENT`
option to tell CMake that the input file may have been recently created
so that we can retry a few times to read it.
On Windows, `environ` is encoded by `CP_ACP`, which may be different
from `KWSYS_ENCODING_DEFAULT_CODEPAGE`. When environment variables
include a unicode character, they may be corrupted. Use `_wenviron`
instead.
Add a cross-platform wrapper over mkdtemp. This will allow us to create
guaranteed-unique directories. On POSIX platforms, this is simply a
wrapper over mkdtemp. On Windows, we take a brute-force approach using
C++11's random facilities and relying on attempts to create an existing
directory resulting in an error. (This approach is very possibly how
mkdtemp is implemented internally, and should be suitable for any
platform if needed, although at present it only uses a partial set of
substitution characters since Windows likely implies a case-insensitive
file system.)
When `cmake -E env` is given the `--modify` flag, try to parse the
following argument as an `ENVIRONMENT_MODIFICATION` operation and apply
it to the environment.
This generalizes `--unset=`:
1. When implementing `ENVIRONMENT_MODIFICATION` features for other CMake
commands, the `MYVAR=OP:VALUE` strings do not need to be translated
to OP-specific flags.
2. This provides a natural and consistent extension point to introduce
new operations without introducing very many flags.
3. Users need to learn only one syntax to access the same functionality.
There is one difference between the behavior here as compared to CTest's
interpretation of the `ENVIRONMENT_MODIFICATION` test property.
The `MYVAR=reset:` command when run in `cmake -E env` will reset `MYVAR`
to whatever its value was when `cmake -E env` launched, rather than try
to checkpoint after plain `MYVAR=VALUE` options. This makes `MYVAR=VALUE`
and `--modify MYVAR=set:VALUE` semantically equivalent.
Going through the same internal API for both `ENVIRONMENT` and
`ENVIRONMENT_MODIFICATION` properties will make it easier to implement
checkpointing for `MYVAR=reset:` more efficiently if the need ever
presents itself. It also makes the two-stage nature of the environment
mutation clearer in the code itself.
Revert commit 5101d586c4 (Windows: Prefer junctions for directory
symlinks, 2022-02-22, v3.24.0-rc1~575^2). Junctions do not support
`../` and other non-canonical paths. Revert their use pending further
investigation.
Fixes: #23781
Issue: #23257
The fallback path boils down to putenv(). Calling that with a "=" sets the
variable to an empty string.
Use cmSystemTools::UnPutEnv() instead, which correctly handles unsetting
variables on a variety of systems.
Rename the booleans 's_ErrorOccured' and 's_FatalErrorOccured' to
's_ErrorOccurred' and 's_FatalErrorOccurred', respectively.
Rename the getters and setters to 'Get[Fatal]ErrorOccurred' and
'Set[Fatal]ErrorOccurred', and fix all uses across the codebase.
Similar to GNU tar add a --touch option to the tar extract command to
skip extracting the timestamps from the files in the archive
effectively touching them as if they were just created.
Issue: #22746
Update the approach added by commit afb7f6e4ff (cmake: Add '-E
create_symlink' support on Windows, 2018-06-11, v3.13.0-rc1~75^2) to use
junctions, as suggested [here](https://superuser.com/a/1291446/140450).
This allows them to work under security limitations on Windows.
Fixes: #23257
Since commit 2e1149874d (cmSystemTools: Support multiple binary formats,
2021-06-14, v3.22.0-rc1~575^2) the `file(RPATH_...)` operations fail on
files that are not ELF or XCOFF format. Previously the RPATH operations
tolerated files of unknown format if the goal was to produce a file with
an empty RPATH. Restore this tolerance in order to support setting an
empty RPATH on GNU ld scripts.
Fixes: #22963
