Refactoring in commit v3.8.0-rc1~445^2~2 (cmTarget: Move sanity checks
and computed property access to callers, 2016-10-13) exposed a typo in
commit v3.8.0-rc1~445^2~3 (cmGeneratorTarget: Implement
cmTargetPropertyComputer interface, 2016-10-13). Together they broke
the `$<TARGET_PROPERTY:mytgt,SOURCES>` generator expression in the case
that the `SOURCES` target property is populated in part by the
`target_sources` command. Add the missing `;`-separator.
Fixes: #17243
In case a list file contains a null terminated string that is continued
until a later space, the lexer token information got inconsistent:
e.g. an argument "TEST\0FOOBAR" is passed by the lexer as a token
char* = "TEST\0FOOBAR" and length 11
^^ note: ascii 0x00
Using strdup in cmListFileLexer leads lexer->token.text to be allocated
with size 5 and lexer->token.length to be set to 11
A subsequent call to this function with an argument of 5 < length <= 11
wrongly assumed a sufficiently sized buffer and therefore corrupted the
heap buffer. The program might crash due to this corruption.
The case "NullTerminatedArgument" is intentionally using a quite large
'rest' to increase the chance to actually hit the issue. It will reliably
crash with address sanitizer enabled though.
This fix addresses all rules where arbitrary characters are matched to
ignore \0 in order to fall through to the rule that matches an arbitrary
character as BadCharacter.
Signed-off-by: Matthias Maennich <matthias@maennich.net>
Create a `<LANG>_CPPCHECK` target property (initialized by a
`CMAKE_<LANG>_CPPCHECK` variable) to specify a `cppcheck` command line
to be run along with the compiler.
In commit 8ed03baa76 (gitattributes: prefer `eol=crlf` to `-crlf`,
2017-08-23) we left a few CRLF blobs in the repository. Some Git
versions get confused by text files with CRLF blobs. Convert them
to LF blobs. Use the `eol=crlf` attribute to tell Git to use CRLF
on checkout.
Some CPack tests require running
commands after the inclusion of
CPack.cmake and this patch enables
such tests to declare run_after_include_cpack
function which is run after the
inclusion.
Disable the feature added by commit v3.9.0-rc1~71^2~2 (find_*: Add a new
PackageRoot search path group, 2017-05-03) and remove documentation
added by commit v3.9.0-rc1~71^2 (find_*: Add docs for PackageRoot search
path group, 2017-05-03). Unfortunately the name `<pkg>_ROOT` may
already be set by projects for their own incompatible purposes. Disable
the behavior change for now to fix the regression for CMake 3.9. We can
restore it later with a policy.
In order to keep the implementation and tests working, add an
undocumented variable we can use in the tests to enable the behavior
before the policy is introduced.
Fixes: #17144
CMake will soon require both C++11 and libuv to build. Neither of
these works on HP-UX, so unfortunately we need to drop support for
the platform until someone can get them working.
Issue: #17137
Add a new gtest_discover_tests function to GoogleTest.cmake,
implementing dynamic test discovery (i.e. tests are discovered by
actually running the test executable and asking for the list of
available tests, which is used to dynamically declare the tests) rather
than the source-parsing approach used by gtest_add_tests. Compared to
the source-parsing approach, this has the advantage of being robust
against users declaring tests in unusual ways, and much better support
for advanced features such as parameterized tests.
A unit test, modeled after the TEST_INCLUDE_DIR[S] test, is also
included. Note that the unit test does not actually require that Google
Test is available. The new functionality does not actually depend on
Google Test as such; it only requires that the test executable lists
tests in the expected format when invoked with --gtest_list_tests, which
the unit test can fake readily.
When `string(REGEX REPLACE)` or `string(REGEX MATCHALL)` loop
internally, they store their matches, but they do not clear the previous
match from an earlier iteration. This can leave the contents of
`CMAKE_MATCH_<N>` with bogus values for later matches in the string if
they have groups which earlier matched a non-empty string, but now match
an empty string.
Fixes#17079.
c4de0a25 Add sha1sum, sha224sum, sha256sum, sha384sum and sha512sum to command mode
c4647d84 Change ComputeFileMD5 to ComputeFileHash
501a4fee Add some unit tests for md5sum
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !1021
- Added clearer structure by grouping the options into logical sections.
- Expanded the details for many of the options.
- Added Examples section to show how to use the various commands.
- Specifically highlighted that the contents of SOURCE_DIR may be lost
if a download method is also provided.
- Updated argument-matching regex to be more robust and account for the
varying leading spaces before keywords in the docs.
- Updated tests to account for slightly changed error messages.
The code
add_library(A OBJECT a.c)
target_sources(A PRIVATE $<TARGET_OBJECTS:A>)
used to crash CMake via infinite recursion while evaluating the
generator expression. Then the change in commit v3.9.0-rc1~266^2~1
(cmGeneratorTarget: Replace source classifier implementation,
2017-04-07) avoided the infinite recursion because GetKindedSources now
creates a map entry and initializes it once. If it is called again on
the same target during that initialization, the partially computed
results are returned. This is still wrong but does not crash.
Detect and diagnose this case instead.
Co-Author: Ben Boeckel <ben.boeckel@kitware.com>
Fixes: #16578
The logic added in commit v3.6.0-rc1~30^2 (Add a variable to specify
language-wide system include directories, 2016-05-24) to use
`CMAKE_<LANG>_STANDARD_INCLUDE_DIRECTORIES` incorrectly filters them by
`CMAKE_<LANG>_IMPLICIT_INCLUDE_DIRECTORIES`. Rather than recognizing
this, commit v3.8.0-rc1~60^2 (Android: Pass sysroot include directory
explicitly, 2017-01-20) worked around the problem by incorrectly
removing `/usr/include` from `CMAKE_<LANG>_IMPLICIT_INCLUDE_DIRECTORIES`
so it worked in `CMAKE_<LANG>_STANDARD_INCLUDE_DIRECTORIES`.
By not filtering out `/usr/include` from user-specified include
directories, we allow the code
include_directories(${CMAKE_SYSROOT}/usr/include)
to place the include directory too early on the command line.
Fix support for standard include directories to not be filtered by
implicit include directories, and do not remove `/usr/include` from the
list of implicit include directories for Android builds. Add a test
case to verify that an explicit `/usr/include` is ignored in favor
of the standard directory at the end.
Fixes: #17059
Revert the main logic change of commit v3.9.0-rc1~192^2 (bindexplib: fix
constants symbols export, 2017-04-26) and its test case. Unfortunately
some constants may be provided by multiple object files with different
`@...` suffixes, leading to ambiguous references. Revert support
pending further investigation.
Fixes: #17045
Since commit v3.9.0-rc1~58^2 (find_package: shorten output for missing
package in config mode, 2017-05-09) we print only the one line
- Could NOT find Foo (missing: Foo_DIR)
when package Foo cannot be found in CONFIG mode and it is not REQUIRED.
However, in the case that package configuration files were found but not
used, this one line message leaves out important information. This can
happen when a package configuration file sets `Foo_FOUND` to `FALSE` or
when its package version file does not match the requested version.
Restore the longer message in these cases. Otherwise a seemingly valid
explicit `Foo_DIR` setting appears to be silently ignored even if it was
considered.
Fixes: #17029
