Calling assert() on a condition that's always false is not an
appropriate way to exit after emitting "Fatal error [...]" because
if NDEBUG is defined the assert() becomes a no-op and the error
stops actually being fatal.
Use `#pragma GCC poison` (supported since GCC 3, maybe earlier) when
compiling with GCC to help prevent direct uses being introduced for
functions which DOH provides a wrapper for.
Exit() is a wrapper for exit() by default, but SetExitHandler() allows
specifying a function to call instead.
This means that failures within DOH (e.g. Malloc() failing due to lack
of memory) will now perform cleanup such as removing output files.
This commit also cleans up exit statuses so SWIG should now reliably
exit with status 0 if the run was successful and status 1 if there was
an error (or a warning and -Werror was in effect).
Previously in some situations SWIG would try to exit with the status set
to the number of errors encountered, but that's problematic - for
example if there were 256 errors this would result in exit status 0 on
most platforms. Also some error statuses have special meanings e.g.
those defined by <sysexits.h>.
Also SWIG/Javascript tried to exit with status -1 in a few places (which
typically results in exit status 255).
Previously code in the SWIG tool didn't handle allocation failures
well. Most places didn't check for NULL return from
malloc()/realloc()/calloc() at all, typically resulting in undefined
behaviour, and some places used assert() to check for a NULL return
(which is a misuse of assert() and such checks disappear if built with
NDEBUG defined leaving us back with undefined behaviour).
All C allocations are now done via wrapper functions (Malloc(),
Realloc() and Calloc()) which emit and error and exit with non-zero
status on failure, so a non-NULL return can be relied upon.
Fixes#1901.
Specifying a value on the typemap method now gives an error, e.g.:
%typemap(argout=123) char * ""
The old way of specifying a language name in the typemap attributes
is no longer supported (it has been deprecated for 16 years).
Closes#891
Both function annotations and variable annotations are turned on using the
"python:annotations" feature. Example:
%feature("python:annotations", "c");
struct V {
float val;
};
The generated code contains a variable annotation containing the C float type:
class V(object):
val: "float" = property(_example.V_val_get, _example.V_val_set)
...
Python 3.5 and earlier do not support variable annotations, so variable
annotations can be turned off with a "python:annotations:novar" feature flag.
Example turning on function annotations but not variable annotations globally:
%feature("python:annotations", "c");
%feature("python:annotations:novar");
or via the command line:
-features python:annotations=c,python:annotations:novar
Closes#1951
Python function annotations containing C/C++ types are no longer
generated when using the -py3 option. Function annotations support
has been moved to a feature to provide finer grained control.
It can be turned on globally by adding:
%feature("python:annotations", "c");
or by using the command line argument:
-features python:annotations=c
The implementation is designed to be expandable to support different
annotations implementations. Future implementations could implement
something like the following for generating pure Python types:
%feature("python:annotations", "python");
or typing module types to conform to PEP-484:
%feature("python:annotations", "typing");
Closes#1561
Issue #735
- Improved documentation for using declarations.
- Issue new warning WARN_LANG_USING_NAME_DIFFERENT when there
is a conflict in the target language name to be used when
introducing a method via a using declaration. Previously
the method was silently ignored. Issue #1840. Issue #655.
An invalid preprocessor expression is reported as a pair of
warnings with the second giving a more detailed message from the
expression evaluator. Previously SWIG prefixed the second message
with "Error:" - that was confusing as it's actually only a warning
by default so we've now dropped this prefix.
Before:
x.i:1: Warning 202: Could not evaluate expression '1.2'
x.i:1: Warning 202: Error: 'Floating point constant in preprocessor expression'
Now:
x.i:1: Warning 202: Could not evaluate expression '1.2'
x.i:1: Warning 202: Floating point constant in preprocessor expression
See #1465
A subexpression in parentheses lost its string/int type flag and
instead used whatever type was left in the stack entry from
previous use. In practice we mostly got away with this because
most preprocessor expressions are integer, but it could have
resulted in a preprocessor expression incorrectly evaluating as
zero. If -Wextra was in use you got a warning:
Warning 202: Error: 'Can't mix strings and integers in expression'
Fixes#1384
This commit handles multi-version support at runtime, it
fixes:
* 5.5.2 - with cutted long identifier name
* 6.0.0 - with full string identifier
* 6.1.0 - with 1 or 0 output argument
It also improves the codebase by:
* Using `Char(X)` instead of `DohCheck(X)` and `Data(X)`
* Using `Len(X)` instead of `strlen()`
* Correctly detecting old Scilab versions
Implementation is very similar to typedef implementation.
Issue #655 and closes#1488.
Testcase using_member.i.
Better implementation to that reverted in previous commit 3f36157b.
Symbol tables shown with -debug-csymbols and -debug-symbols now correct
and are similar to when using a typedef.
If it's not a recognised directive the scanner now emits MODULO and then
rescans what follows, and if the parser then gives a syntax error we
report it as an unknown directive. This means that `a%b` is now allowed
in an expression, and that things like `%std::vector<std::string>` now
give an error rather than being quietly ignored.
Fixes#300Fixes#368
This reverts commit d7e0aaa57d. I can't
get regextarget to work correctly, and the fortran branch no longer
depends on it since pcre support is optional.
The "command" encoder was mostly intended for use in `%rename` - most
uses can be achieved using the "regex" encoder, so we recommend using
that instead.
The "command" encoder suffers from a number of issues - as the
documentation for it admitted, "[it] is extremely slow compared to all
the other [encoders] as it involves spawning a separate process and
using it for many declarations is not recommended" and that it "should
generally be avoided because of performance considerations".
But it's also not portable. The design assumes that `/bin/sh` supports
`<<<` but that's a bash-specific feature so it doesn't work on platforms
where `/bin/sh` is not bash - it fails on Debian, Ubuntu and probably
some other Linux distros, plus most non-Linux platforms. Microsoft
Windows doesn't even have a /bin/sh as standard.
Finally, no escaping of the passed string is done, so it has potential
security issues (though at least with %rename the input is limited to
valid C/C++ symbol names).
Fixes#1806
