Prior to this change / was not allowed in the build name. This was tested
with a CDash server and worked. In addition the safe build name was not
used everywhere. This caused mismatched build names to be in the xml
files going to CDash which caused different rows to be created for the
same build.
This allows ctest_update to get the current version without actually
changing the repository. This is useful when using Jenkins or an
external project to update the source to a specific version, but you
still want the current version to show up in CDash.
Split the monolithic cmCPackIFWGenerator source into three parts:
cmCPackIFWGenerator, cmCPackIFWInstaller, and cmCPackIFWPackage
to isolate the implementations somewhat.
Add a COMMON option to the cpack_ifw_configure_component command to make
the data component common to its parent group.
Change the default packaging method to ONE_PACKAGE_PER_COMPONENT.
This is not incompatible because the CPack IFW generator has not
yet been in a release.
This option instructs configure_package_config_file to consider paths
that are not absolute as relative to the INSTALL_PREFIX directory
instead of relative to the CMAKE_INSTALL_PREFIX directory.
If this argument is not passed, the default behaviour is to use
the CMAKE_INSTALL_PREFIX variable.
The new signature is therefore:
configure_package_config_file(<input> <output> INSTALL_DESTINATION <path>
[PATH_VARS <var1> <var2> ... <varN>]
[NO_SET_AND_CHECK_MACRO]
[NO_CHECK_REQUIRED_COMPONENTS_MACRO]
[INSTALL_PREFIX <path>])
The implementation of CMP0047 detects the QNX compiler id internally,
then changes it to GNU after the language is fully enabled. This is too
late to correctly set the old IS_GNU variables, so set them as part of
the policy OLD behavior.
Reported-by: Mike Lattanzio <mlattanzio@blackberry.com>
The implementation of CMP0047 detects the QNX compiler id internally,
then changes it to GNU after the language is fully enabled. This is too
late to correctly set the old IS_GNU variables, so set them as part of
the policy OLD behavior.
Reported-by: Mike Lattanzio <mlattanzio@blackberry.com>
When $<TARGET_PROPERTY> names a build property like COMPILE_DEFINITIONS
but not the usage requirement (INTERFACE_) version of it, the value
should be that used to build the target. It should not be influenced by
a dependent 'head' target like usage requirements are.
Extend the CompileDefinitions test with a case covering the corrected
behavior.
Reviewed-by: Stephen Kelly <steveire@gmail.com>
5bf1c5cc Help: Document add_custom_command PRE_BUILD/PRE_LINK for exe/lib only (#15059)
2a58c872 Help: Revise and format 'add_custom_command' docs
cb88742d Help: Remove stray content from 'add_custom_command' docs
The context->Makefile is never NULL, but our checks for it convince
Clang scan-build that it might be NULL. Then it warns about later
unchecked uses. Drop the unnecessary checks.
Let clients do their own research on the current strength of each
hash algorithm to choose what is best for their needs.
Suggested-by: Nico Schlömer <nico.schloemer@gmail.com>
State explicitly that the PRE_LINK mode is not for targets created by
the add_custom_target command. The existing wording for PRE_BUILD being
treated as PRE_LINK by non-VS generators will now imply this restriction
for PRE_BUILD too.
Format the reStructuredText markup manually. Organize the command
options into a definition list. Use inline markup to cross-reference
related documents.
