Store the definitions in a cmLinkedTree in the cmMakefile. This can
be moved to cmState and then the tree will provide snapshotting
possibilities. It will also make the Closure copy created at
the start of each cmMakefile unnecesarry.
In cmMakefile::PushScope, a copy of the closure of keys initialized
in the parent scope is made. In PopScope, essentially the same copy
is inserted back into the parent. That means a lot of duplication
of strings and a lot of string comparisons. None of it is needed,
because the cmDefinitions keys already provide a canonical
representation of what is initialized.
The removal of the separate container also makes the variable handling
code more easy to reason about in general.
Before this patch, configuring llvm uses 200 KiB for the VarInitStack.
Overall peak memory consumption goes from 35.5 MiB to 35.1 MiB.
Presumably the intention here is to attempt to optimize memory by not
storing what is not needed. However, all keys need to be tracked
anyway to implement initialization tracking, and this special case
gets in the way of simplifying the implementation of that.
This doesn't change any observable effects because values set
to 0 are considered not to exist by the cmDefinitions API.
For some reason, using recursion here is faster to configure ParaView
than using a loop. Probably some compiler optimization is inhibited
by using a loop.
Co-Author: Brad King <brad.king@kitware.com>
Construct the final list directly in a named return value. Use
a single set to track bindings already found.
Co-Author: Brad King <brad.king@kitware.com>
d1b62185 Merge branch 'parent-scope-tests' into variable-pull-failure
5f414cef Revert "cmDefinitions: Don't store parent lookups"
e0c0b1ac test: add a test for PARENT_SCOPE with multiple scopes
064c415d test: add test for PARENT_SCOPE behavior
This reverts commit 5abfde6cb8.
The behaviors associated with implicit pulldown on variable lookup
seriously conflict with the optimizations made in these commits.
Basically, since values were copied upon variable lookup, not just on
PARENT_SCOPE, coupled with PARENT_SCOPE's behavior based on whether the
variable is in the current scope or not causes serious problems with not
storing a value for every variable at every scope.
The commit changed behavior of the following example, among other cases:
function(test_set)
set(blah "value2")
message("before PARENT_SCOPE blah=${blah}")
set(blah ${blah} PARENT_SCOPE)
message("after PARENT_SCOPE blah=${blah}")
endfunction()
set(blah value1)
test_set()
message("in parent scope, blah=${blah}")
Reported-by: Alex Merry <alex.merry@kde.org>
Reported-by: Ben Cooksley <bcooksley@kde.org>
Casts from std::string -> cmStdString were high on the list of things
taking up time. Avoid such implicit casts across function calls by just
using std::string everywhere.
The comment that the symbol name is too long is no longer relevant since
modern debuggers alias the templates anyways and the size is a
non-issue since the underlying methods are generated since it's
inherited.
This converts the CMake license to a pure 3-clause OSI-approved BSD
License. We drop the previous license clause requiring modified
versions to be plainly marked. We also update the CMake copyright to
cover the full development time range.
Previously each new variable scope (subdirectory or function call) in
the CMake language created a complete copy of the key->value definition
map. This avoids the copy using transitive lookups up the scope stack.
Results of queries answered by parents are stored locally to maintain
locality of reference.
The class cmDefinitions replaces cmMakefile::DefinitionsMap, and is
aware of its enclosing scope. Each scope stores only the definitions
set (or unset!) inside it relative to the enclosing scope.
Stephen Kelly
Brad King
Ben Boeckel
Ben Boeckel