In the Ninja generator we run all build rules from the top of the build
tree rather than changing into each subdirectory. Therefore we convert
all paths relative to the HOME_OUTPUT directory. However, the Convert
method on cmLocalGenerator restricts relative path conversions to avoid
leaving the build tree with a "../" sequence. Therefore conversions
performed for "subdirectories" that are outside the top of the build
tree always use full paths while conversions performed for
subdirectories that are inside the top of the build tree may use
relative paths to refer to the same files.
Since Ninja always runs rules from the top of the build tree we should
convert them using only the top-level cmLocalGenerator in order to
remain consistent. Also extend the test suite with a case that fails
without this fix.
Refactor the local generator creation API to accept a
cmState::Snapshot. Adjust MakeLocalGenerator to use the 'current'
snapshot in cases where there is no parent. Create the snapshot
for subdirectories in cmMakefile::AddSubdirectory.
This means that snapshots are now created at the point of extending the tree,
as appropriate, and independently of the cmLocalGenerator and cmMakefile they
represent the state for.
Global properties are already global in scope, so remove the
overload for specifying it and port users of the API.
The call from cmMakefile::GetProperty can be simplified because
the scope is only used during chaining, and there is no further
chaining after processing global properties.
A common idiom in CMake-based build systems is to have custom commands
that generate files not listed explicitly as outputs so that these
files do not have to be newer than the inputs. The file modification
times of such "byproducts" are updated only when their content changes.
Then other build rules can depend on the byproducts explicitly so that
their dependents rebuild when the content of the original byproducts
really does change.
This "undeclared byproduct" approach is necessary for Makefile, VS, and
Xcode build tools because if a byproduct were listed as an output of a
rule then the rule would always rerun when the input is newer than the
byproduct but the byproduct may never be updated.
Ninja solves this problem by offering a 'restat' feature to check
whether an output was really modified after running a rule and tracking
the fact that it is up to date separately from its timestamp. However,
Ninja also stats all dependencies up front and will only restat files
that are listed as outputs of rules with the 'restat' option enabled.
Therefore an undeclared byproduct that does not exist at the start of
the build will be considered missing and the build will fail even if
other dependencies would cause the byproduct to be available before its
dependents build.
CMake works around this limitation by adding 'phony' build rules for
custom command dependencies in the build tree that do not have any
explicit specification of what produces them. This is not optimal
because it prevents Ninja from reporting an error when an input to a
rule really is missing. A better approach is to allow projects to
explicitly specify the byproducts of their custom commands so that no
phony rules are needed for them. In order to work with the non-Ninja
generators, the byproducts must be known separately from the outputs.
Add a new "BYPRODUCTS" option to the add_custom_command and
add_custom_target commands to specify byproducts explicitly. Teach the
Ninja generator to specify byproducts as outputs of the custom commands.
In the case of POST_BUILD, PRE_LINK, and PRE_BUILD events on targets
that link, the byproducts must be specified as outputs of the link rule
that runs the commands. Activate 'restat' for such rules so that Ninja
knows it needs to check the byproducts, but not for link rules that have
no byproducts.
Changes in commit b9aa5041 (cmLocalGenerator: Simplify GetIncludeFlags
output formatting, 2014-03-04) caused Windows Resource Compiler include
directories to be computed as relative paths in the Ninja generator.
This breaks the cmcldeps handling of include paths. The reason for the
regression is that several cmLocalGenerator::GetIncludeFlags callers
treated the fourth "bool forResponseFile" argument as if it controlled
whether include directories were a full path. It actually did control
that by accident until the above commit.
Add an explicit "bool forceFullPaths" argument to GetIncludeFlags
and thread the value through ConvertToIncludeReference as needed.
Update GetIncludeFlags call sites that really wanted to control the
forResponseFile setting to be aware of the new argument. Extend the
VSResource test to cover this case.
The pre-defined 'console' pool is a non-buffered pool that runs with a
depth of 1. CMake re-run cannot be run concurrently and it will
eventually output something. A non-buffered pool allows to get it as
soon as possible
Also, generate the minimal required version of Ninja in the build file.
Until now the cmCustomCommandGenerator was used only to compute the
command lines of a custom command. Generalize it to get the comment,
working directory, dependencies, and outputs of custom commands. Update
use in all generators to support this.
Use the clang RemoveCStrCalls tool to automatically migrate the
code. This was only run on linux, so does not have any positive or
negative effect on other platforms.
As an INTERFACE_LIBRARY has no direct link dependencies, we can
short-circuit in cmGeneratorExpressionEvaluator and
in cmGlobalGenerator::CheckLocalGenerators.
As they do not generate any output directly, any generate- or install-
related code acn also be short-circuited. Many of the local generators
already do this.
Because only INTERFACE related properties make sense on INTERFACE_LIBRARY
targets, avoid setting other properties, for example via defaults.
Just enough to reach the BuildMacContentDirectory method and the
NeedRelinkBeforeInstall methods.
In the future, those methods can be moved to cmGeneratorTarget.
Drop the "vsProjectFile" argument from cmTarget::TraceDependencies. It
appears to be the modern equivalent to a hunk added in commit ba68f771
(...added new custom command support, 2003-06-03):
+ name = libName;
+ name += ".dsp.cmake";
+ srcFilesToProcess.push(name);
but was broken by refactoring at some point. The current behavior tries
to trace dependencies on a source file named the same as a target, which
makes no sense. Furthermore, in code of the form
add_executable(foo foo.c)
add_custom_command(OUTPUT "${somewhere}/foo" ... DEPENDS foo)
the "vsProjectFile" value "foo" matches source "${somewhere}/foo.rule"
generated to hold the custom command and causes the command to be added
to the "foo" target incorrectly.
Simply drop the incorrect source file trace and supporting logic.
Stephen Kelly
Brad King
James Johnston
Peter Collingbourne
Nils Gladitz
Sylvain Joubert
Ben Boeckel
Peter Kümmel