The storage of a pointer means that the ownership and lifetime are
externally determined, which is harder to reason about. It also imposes
API constraints, requiring APIs to return references to backtraces.
This pointer storage was introduced in commit v3.1.0-rc1~425^2~3 (genex:
remove the need for backtraces, 2014-05-23). As backtraces are now cheap
to copy around, just do that instead.
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.
While tracing dependencies of a target, cmTargetTraceDependencies
follows sources by full path to determine if the source is to be
produced by a custom command. Commit 4959f341 (cmSourceFileLocation:
Collapse full path for directory comparisons., 2014-03-27) changed
the storage of target sources to be in the form of a normalized
path instead of an unnormalized path.
The path is followed by looking it up in a mapping via
cmMakefile::GetSourceFileWithOutput to acquire an appropriate
cmSourceFile. The mapping is populated with the OUTPUT components
of add_custom_command invocations, however it is populated with
unnormalized paths. This means that the tracing logic does not
find appropriate cmSourceFiles, and does not generate appropriate
build rules for the generated sources.
Normalize the paths in the OUTPUT components of add_custom_command
to resolve this.
The paths in the DEPENDS component of add_custom_command are also
not normalized, leading to the same problem again. Normalize the
depends paths after generator evaluation and expansion.
Rely on evaluation in cmCustomCommandGenerator for the generators.
When tracing target dependencies, depend on the union of dependencies
for all configurations.
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.
Most callers already have a std::string, on which they called c_str() to pass it
into these methods, which internally converted it back to std::string. Pass a
std::string directly to these methods now, avoiding all these conversions.
Those methods that only pass in a const char* will get the conversion to
std::string now only once.
Removing the Process() API and removing the parameters from the
constructor will allow cmGeneratorExpressions to be cached and evaluated
with multiple configs for example, such as when evaluating target
properties. This requires the creation of a new compiled representation
of cmGeneratorExpression. The cmListFileBacktrace remains in the
constructor so that we can record where a particular generator
expression appeared in the CMakeLists file.
Evaluate in the COMMAND arguments of custom commands the generator
expression syntax introduced in commit d2e1f2b4 (Introduce "generator
expressions" to add_test, 2009-08-11). These expressions have a syntax
like $<TARGET_FILE:mytarget> and are evaluated during build system
generation. This syntax allows per-configuration target output files to
be referenced in custom command lines.
The cmCustomCommandGenerator::GetCommand method completely replaces the
purpose of this method. Re-implement GetRealLocation inline at the only
remaining call site and remove it.
The Makefile, VS, and Xcode generators previously duplicated some custom
command line generation code. Factor this out into a separate class
cmCustomCommandGenerator shared by all generators.
Stephen Kelly
Brad King
Ben Boeckel
Rolf Eike Beer