Move this function definition to cexcept.swg from the code and only
define it if SWIG_swig_check_DEFINED is not defined yet, which both
simplifies the code in C.cxx and makes exception handling code more
flexible, as it's now possible to predefine SWIG_swig_check_DEFINED
in the code injected into the "cxxheader" section to replace the default
implementation. Show an example of doing this in the documentation and
document handling exceptions with C API too.
The changes above required adding a new "cxxcode" section, corresponding
to the "implementation" part of C++ wrappers and defining a new
SWIG_CXX_WRAPPERS preprocessor symbol to allow only adding C++-specific
code when C++ wrappers are actually generated. Also improve the
documentation of the C-specific sections in the manual.
This is much more convenient and allows checking if the shared pointer
is empty easily, unlike before, when it couldn't be done and adding
support for it would have required adding extra functions.
Also add a way to check whether an object is null in C++ wrappers of the
classes handled via shared pointers and static null() method for
creating null objects of such classes.
This makes using returning strings much simpler to use from C++ code as
the returned pointers don't have to be deleted manually -- although, of
course, this does require an extra allocation and copy and so should be
avoided for the very long strings.
Add a new runtime test showing how simple and convenient it is to use
the functions working with string using the C++ wrappers now.
Add typecheck typemaps for primitive types and string and call
Swig_overload_check() to ensure that we don't generate two wrappers
functions taking the same "const char*" type if we have overloads taking
it and "std::string" (or a reference) in the original code.
Avoid unnecessary heap allocations, just use temporary variables.
Actually update the string parameters passed by pointer/non-const
reference. This requires the pointers passed to actually be non-const,
so update the C-specific unit test runme to use a char buffer instead of
a literal string.
Also simplify the code copying the string contents to just use strdup()
(if there are ever any platforms where this POSIX functions is not
available, we could just define it ourselves once instead of using
strlen() + malloc() + memcpy() manually twice).
Use enum types instead of int for the enum-valued parameters and
function return values, this is more type-safe and clear for the users
of the library.
Change cpp_enum unit test to use C++ to check that C++ enum wrappers
can at least be compiled, but still use C API in it.
Note that enum whose underlying type is bigger than int still don't
work, but this is no different from what it was before, so just document
this limitation but don't do anything else about it for now.
This commit is best viewed ignoring whitespace-only changes.
This allows different threads to use the same library concurrently
without corrupting each other's exceptions.
Note that this requires C++11 support for wrapper compilation too. If
this really turns out to be a problem, we could always just #define
thread_local as compiler-specific equivalent such as "__thread" for gcc
or "__declspec(thread)" for MSVC, which are available since forever.
Ensure that we have only a single SWIG_CException_Raise() function
across all modules instead of having per-module functions and, worse,
per-module PendingException variables, which resulted in compile-time
errors and couldn't work anyhow because function checking for the
current exception didn't necessarily use the same "global" variable
where it was stored. More formally, old version resulted in ODR
violations and undefined behaviour.
The way we avoid it now is rather ugly and consists in excluding
SWIG_CException from wrapping using the hack in the source code which
postpones wrapping this class until the very end and checks if we had
encountered any %import directives and simply doesn't wrap it if we did.
The same code is used to define the special SWIG_CException_DEFINED
preprocessor symbol which is then used in the generated code to prevent
the SWIG_CException class declaration from being compiled as part of the
wrapper too (because this still happens due to %inline being used for
its definition, and there doesn't seem to be any better way to avoid
this).
This is definitely not pretty, but at least adding "throw(char*)" to a
couple of functions in mod_[ab].i test suite files works now instead of
failing (even without linking and running) as before. This commit
doesn't modify the test suite to avoid possible problems with the other
languages, however.
Check for the pending exception after every call to a wrapper function
not marked "noexcept" and throw a C++ exception if necessary.
Add C++ version of the exception example to show how this works.
Also change SWIG_CException to use member functions for checking for and
resetting pending exceptions, this seems better than having separate
functions for it and will make it easier to customize exception handling
later by just replacing SWIG_CException class with something else.
Note that we still use a global (and not a member) function for raising
the exception, but this one is not exported at all, and needs to be a
function in order to be easily callable from other modules (see the
upcoming commit).
These typemaps never worked, but this went unnoticed until now because
the nonsensical "ctype" expansion generated by them still compiled.
With types possibly having a namespace prefix, they didn't any longer,
so define them correctly for the objects using "$resolved_type" and
define the typemap for "void*[]" separately, as "$resolved_type" can't
be used for it.
Don't include the entire file, including its header comment, into the
generated code, but just the part that we want to appear there.
This looks nicer and is also more explicit and hence clear.
Enable the tests and support of shared_ptr in them for C (which required
disabling a previously passing, because not doing anything, attributes
test which is currently broken for unrelated reasons).
This is similar to be491506a (Java std::vector improvements for types
that do not have a default constructor., 2019-03-01) for Java, except we
don't have to bother with any compatibility constraints for this, not
yet used by anyone. module.
This reverts commit d89a95a48c as,
finally, it's not worth trying to use UTL typemaps from C: this required
bad hacks for std::vector and would be even worse for std::map, so don't
bother with them and just correct the "poor" C typemaps to be slightly
less poor and use them instead.
Use simple fixed typemap instead of trying to use the much more complex
one from the UTL which doesn't work for C.
Add a simple test case for std::map<>.
The existing typemaps didn't work at all and making them work would
require defining all the typemaps needed by the Unified Typemaps
Library, which seems to be geared towards dynamic languages.
Just implement completely straightforward (and not very convenient to
use) typemaps instead.
Enable the now passing unit tests and add a runme for one of them.
Remove the code related to "_result_ref" which was confusing and plain
wrong, as it generated something that compiled but crashed during
run-time due to the use of a pointer to an already destroyed stack
object.
Instead, correct the "out" typemap to create a new copy of the object,
which mostly works fine on its own, except that it depends on using
SwigValueWrapper if necessary, so add the call to cplus_value_type()
does this. This also required removing the code resolving typedefs in
the "type" attribute because it confused the base class logic and still
needs an explicit cast to the actual return type due to the use of (and
probable bug in) get_wrapper_func_return_type().
These changes mean that "cppouttype" typemap is not used any longer, so
remove it too.
A couple more tests pass now.
Use "..." instead of "{...}" to avoid generating extra lines with
opening/closing curly braces, which are completely unnecessary.
Also remove an unnecessary "if" check in "in" typemap, which was
useless.
No real changes.
There is no need to check if the input is non-null, as the effect of
both branches of "if" is the same anyhow.
Also get rid of an extra and unnecessary block.
No real changes.
They seem unnecessary too and look suspicious, notably the use of "static" in
the "out" typemap is almost certainly wrong.
As all the tests still pass without them, don't bother fixing these typemaps
but just remove them completely.
It's not clear why they were defined in the first place, but they broke
li_std_vector_ptr unit test, which now uses "T**", after the latest merge with
master, so just remove them to let the test pass again.
At the very least, this gives us a working vector<bool> and allows
"li_std_vector" unit test to pass. It is also just nice to reuse the existing
typemaps instead of having another copy of them.
As C doesn't have UTL, some parts of std_vector.i had to be excluded from it
however.
The existing typemaps didn't make much sense, simplify (there doesn't seem to
be any point in using $1_basetype when dealing with "bool") them and treat
"const bool&" as "bool", not as pointer, as this is much more convenient in C.
This also allows another unit test to pass.
Using longjmp was incompatible with using C++ objects in the code using the
wrappers, and using this C API from C++ to avoid ABI incompatibilities between
different C++ compilers is one of the main reasons for using this module.
Also, this required using a separate SwigObj instead of just using the real
object pointer which inevitably resulted in memory leaks whenever a non owned
object was returned from anywhere, e.g. from a member accessor or any method
returning pointer or reference.
Abandon the attempts to recreate C++ exceptions in C and just use a very
simple approach allowing to pass an error message out of band after any
function call in a global variable. An alternative could be to add a special
"out" error parameter to each and every function, but this risked being too
verbose, especially for the functions which don't really throw, and the calls
to SWIG_PendingException_get() won't need to be made explicitly when using a
C++ wrapper around the generated C API in the future.
This simplifies both the module and the generated code, in particular we don't
need any runtime code at all any more and there is no need for an extra level
of indirection for every object.
It also makes a couple more tests pass.
Use the same ctype for wrapper declarations and definitions and just expand
$resolved_type differently in the two cases.
This simplifies the typemaps and ensures that the declarations and definitions
use the same types, at least for all non-object parameters.
We can always just use the "unconst" one because it works when the type is
used for a return value and is the same as the other one when it's used for a
parameter anyhow because top-level const is ignored in function signatures, so
don't bother defining two sets of macros when one will do.
The sole reason for including this header seems to be to get size_t
declaration, so include the smallest header sufficient for this.
It would be even better to use a fragment for all size_t typemaps to only
include it if necessary, but it doesn't seem to be worth doing this right now.
Member function pointers can't possibly be represented as function pointers,
they have a strictly bigger size and attempting to do it resulted in code
which, with a lot of bad casts, compiled, but crashed during run-time.
The proper solution is to represent C++ method pointers with an appropriate
opaque type, but for now this remains broken -- just make it explicitly broken
instead of pretending that it works when it actually has no chance to.
Let "namespace_spaces" unit test pass, as it's not really related to the
function pointers, by adding an explicit SWIGC test to it.
Pass void pointers directly to C, don't handle them objects which was done by
default previously because we didn't define any specific typemaps for them.
This allows to reenable another test.
Ensure that any commas inside SWIG_STR() used in "out" typemaps for class
types are not interpreted as macro argument separators by using an extra pair
of parenthesis around the type.
Don't populate the typenames array for each and every created object and don't
search it when deleting every object, this is O(N) in number of classes and is
completely impractical for any big library where N can easily be several
thousands. Use destructor function which will correctly destroy the object
instead.
Also don't store each created objects in the global registry, there doesn't
seem to be any point in it.
And, in fact, don't bother with the typenames at all, just use another
pseudo-virtual function for checking whether a class inherits from a class
with the given name.
There is unfortunately one problem with the new approach: it doesn't work when
the same C++ type is wrapped under different names as this results in multiple
specializations of SWIG_derives_from<> for this type. But this is a relatively
rare situation (but which does arise in template_default2 unit test, which had
to be disabled) and could be fixed in the future by completely resolving the
type, including the default template parameters values, and checking if
SWIG_derives_from had been already specialized for it. In the meanwhile, this
regression is not a big deal compared to the advantages of the new approach.
It doesn't make any sense to use protected visibility for the symbols in the
proxy code, the only use for it would be if it could be applied to the
original functions being wrapped, to prevent them from being shadowed by the
proxy versions, but this cannot be done, in general.
Apparently this was already understood as the code using SWIGPROTECT was
commented out, but ef85d0d43f didn't give any
reason for doing this, nor for leaving this code.
Assume there was none and just remove it completely now to avoid confusion.
