This restores the directories used before the MinGW support was added.
For some reason, in the CI the VS and Xcode builds do store the output
in directories not suffixed with Release or Debug, but this is not
usually the case.  As a consequence, on a typical build with those build
systems, this test failed.

Signed-off-by: Erik Boasson <eb@ilities.com>

This fixes incompatibilities with bison 3.8:

* returning user-defined values from yyparse is not supported, this
  works aruond that by storing a more meaningful return value in the
  parser state object;

* the token name table generated by %token-table is deprecated and has
  been for quite some time, and it seems that now the accompanying
  mapping to internal token numbers is gone as well, this uses a hack to
  work around that.

A follow-up fix is needed to remove the use of the token name table
entirely.  The hack here is attractive only in that it provides a quick
solution for the failing CI runs on macOS.

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

The analyzer doesn't do a good job if there are no SAL2 annotations, but
those annotations are only supported by Visual Studio, the documentation
is very incomplete and it is (nearly) impossible to get them right
without actually compiling the code with Visual Studio.  The latter is
what it makes it really impossible to use them.

That makes dealing with false positives from the Visual Studio analzyer
far more trouble than it is worth (especially considering we're also
using the clang and gcc analyzers and Coverity).

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Dennis Potman <dennis.potman@adlinktech.com>

Signed-off-by: Jeroen Koekkoek <jeroen@koekkoek.nl>

Signed-off-by: Jeroen Koekkoek <jeroen@koekkoek.nl>

The length returned by snprintf does not includes the ending
null. So the test must be changed to ensure that if the final length
if sadly 64, the library can be found all the time.

Also add little comments.

Signed-off-by: José Bollo <jose.bollo@iot.bzh>

A message is emitted to stderr when either the plugin can
not be loaded or the symbol 'generate' can't be located
in the loaded plugin.

Logging to sdterr is in wait of some better integration to
some other logging system.

Signed-off-by: José Bollo <jose.bollo@iot.bzh>

Signed-off-by: Erik Boasson <eb@ilities.com>

- The parser was not proceeding correctly in the case of the state
  being IDL_SCAN_ANNOTATION_APPL_SCOPED_NAME

Signed-off-by: Martijn Reicher <martijn.reicher@adlinktech.com>

- The parser was setting the node mask to IDL_ANY in stead of the
  correct mask for bool, char and string literals

Signed-off-by: Martijn Reicher <martijn.reicher@adlinktech.com>

Signed-off-by: Jeroen Koekkoek <jeroen@koekkoek.nl>

Signed-off-by: Dennis Potman <dennis.potman@adlinktech.com>

Case label destructors deleted the const-expr as opposed to
unreferencing it causing a use-after-free if the same enum
was used in more than one union.

Fixes eclipse-cyclonedds#946.

Signed-off-by: Jeroen Koekkoek <jeroen@koekkoek.nl>

Destructors are not executed for untyped midrule actions because Bison
does not know which destructor to run. annotation_appl nodes were not
cleaned up if the annotation_appl could not be successfully finalized.
Remove use of midrule actions in annotation_appl as some platforms still
ship with Bison 3.0.4.

Fixes eclipse-cyclonedds#950.

Signed-off-by: Jeroen Koekkoek <jeroen@koekkoek.nl>

…g 'C' as a language platform-specific macros are irrelevant for idlc, and this is a blocker for compiling mac universal binaries.

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

There is a maximum nesting depth for the configuration that is
determined by the internal configuration tables.  Simply aborting the
interpretation of the XML input on the first unrecognized element item
would guarantee never exceeding the maximum nesting depth.  As an error
handling strategy this is less user-friendly than ignoring some of the
input, attempting to interpret the remainder and reporting any further
issues.

This commit makes it abort the interpretation as soon as the maximum
nesting depth is reached, in all other cases it continues and reports as
many issues with the configuration as it can.

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

- This fixes eclipse-cyclonedds#968
- Add a byte swap to the participant GUID prefix

Signed-off-by: Martijn Reicher <martijn.reicher@adlinktech.com>

CMAKE_SOURCE_DIR always points to the top-level CMake project, so if
CycloneDDS was included as a gitmodule or pulled in using CMake
FetchContent, the CMake step would fail with errors like

```
CMake Error at cyclonedds/src/tools/idlc/CMakeLists.txt:52 (file):
file failed to open for reading (No such file or directory):
/home/robin/dds-project/src/ddsrt/include/getopt.h.in

CMake Error: File /home/robin/dds-project/src/ddsrt/src/getopt.c does not exist.
CMake Error at cyclonedds/src/tools/idlc/CMakeLists.txt:61 (configure_file):
configure_file Problem configuring file

CMake Error at cyclonedds/src/tools/idlc/CMakeLists.txt:95 (include):
include could not find load file:
/home/robin/dds-project/cmake/Modules/Generate.cmake
```

This commit replaces CMAKE_SOURCE_DIR with <PROJECT_NAME>_SOURCE_DIR, a
variable that always points to the source dir of the named project.

See:
https://cmake.org/cmake/help/v3.21/variable/CMAKE_SOURCE_DIR.html
https://cmake.org/cmake/help/v3.21/variable/PROJECT-NAME_SOURCE_DIR.html

Signed-off-by: Robin Lindén <_@robinlinden.eu>

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

gcc 11 warns that there is a conversion from 'long int'
to 'uint32_t' when assigning the value of PTHREAD_STACK_MIN
to tattr.stackSize.  It's pedantically correct; tattr.stackSize
is a uint32_t, and PTHREAD_STACK_MIN is either an integer (if
it is hard-coded), or the result of calling sysconf(), which
always returns a long.  In either case it is a signed value,
so we should explicitly cast it to uint32_t here.  This quiets
the warning.

Signed-off-by: Chris Lalancette <clalancette@openrobotics.org>

This makes idlc show an error message in case the maximum number
of instructions is reached. Because of 16-bits offset fields in the
instructions, the maximum number of instructions that can be used
is limited to 65k.

Signed-off-by: Dennis Potman <dennis.potman@adlinktech.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Erik Boasson <eb@ilities.com>

This suppresses deprecation warnings from OpenSSL 3.0. It is not a
proper fix, but the OpenSSL documentation on how to deal with the
deprecations is such that it will take some time to fix this.

Signed-off-by: Erik Boasson <eb@ilities.com>

Signed-off-by: Christophe Bedard <bedard.christophe@gmail.com>

Signed-off-by: Christophe Bedard <bedard.christophe@gmail.com>