Specify -pthread option which should be universally supported under
non-{MSW,Mac} platforms by now and which is required under Linux to link
the tests using boost::shared_ptr<>, which uses pthread_mutex_xxx().
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).
Also undo the changes to common.mk originally done in this branch and
rendered unnecessary by de5e0c865 (C++11 testing moved to a configure
option, 2013-10-08) on master.
This test doesn't really work, as directors support is not implemented
at all in C backend, but remove it from here to prevent reports from
"make check-failing" about "failing test passing".
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<>.
This is needed to find the SWIG-generated shared library when using C
too, as it's already the case for many (but, surprisingly, not all)
other target languages.
There doesn't seem to be any reason for using it rather than just
returning from main() as usual, and it provokes warnings about
implicitly declared function when compiling them.
They were not supported currently, but processing them resulted in just
a warning about missing ctype typemap and generated uncompilable code.
Give an error and don't generate any code at all now, which is more
clear and helpful.
Also exclude the part of kwargs_feature test using varargs from C test
suite.
This seems to be broken and sporadically results in "Text file busy"
errors when the tests are run actually in parallel, as is the case in
GitHub CI environment.
Don't add the builds known/allowed to fail and don't add all the
variations of Python builds, there are just too many of those, but do
add mostly all the rest, with the exceptions below:
Builds using JavaScript backend with node <= 10 had to be disabled as
well, as they just seem to be broken and there is no real motivation in
supporting these ancient versions.
Builds using D can't be run currently due to incompatibility between the
ancient dmd version being used and the earliest support Ubuntu version
(18.04) on GitHub Actions and not supporting newer versions in SWIG.
Builds using Scilab had to be disabled as running the test suite
reliably hangs. This would need to be debugged by somebody motivated in
doing it.
Note that some builds need to be run under Ubuntu 18.04 (Bionic) and not
the default 20.04 (Focal) as they use old dependencies that don't exist
in Focal any longer. In the future we ought to switch to using newer
versions of these dependencies.
It is not available out of the box in GitHub environment, unlike under
Travis CI, so install it ourselves.
Note that we do it manually because installing the rael-gc/rvm PPA
doesn't seem to work with weird failures in GitHub Actions environment.
We also can't use gpg2 --recv-keys in this environment as it doesn't
work, apparently due to a firewall.
This is probably not the greatest idea, and it would be better to change
the generated code to avoid these tests instead, but it was already done
like this for several tests, so just add two more of them that generate
warnings such as
../../member_funcptr_galore_wrap.cxx: In function ‘SwigV8ReturnValue _wrap_MemberFuncPtrs_aaa6(const SwigV8Arguments&)’:
../../member_funcptr_galore_wrap.cxx:3495:90: error: type qualifiers ignored on cast result type [-Werror=ignored-qualifiers]
3495 | result = (int)((MemberFuncPtrs const *)arg1)->aaa6((short (Funcs::*const )(bool) const)arg2);
| ^~~~
when using gcc 9.
This is similar to the previous commit, but in the other direction, with
the warning being generated for the implicitly-declared copy ctor due to
having an explicitly-declared assignment operator.
And the fix is different too because we need to have an assignment
operator in this test, so define the copy ctor explicitly too.
Defining a copy ctor results in -Werror=deprecated-copy when using
implicitly-generated assignment operator with recent gcc versions, so
simply remove this copy ctor, which was apparently never needed anyhow,
to avoid it and rely on the compiler generating both the copy ctor and
assignment operator implicitly.
This results in passing empty argument to configure which, in turn,
triggers "WARNING: you should use --build, --host, --target" from it as
this argument is apparently misinterpreted as the host name.
Executables can't be produced without this option using clang available
under Ubuntu 20.04, as linking fails with a lot of errors of the form
/usr/bin/ld: Swig/wrapfunc.o: relocation R_X86_64_32S against `.rodata'
can not be used when making a PIE object; recompile with -fPIE
They are supposed to be set to gcc-$GCC and g++-$GCC respectively, if
the suffix is defined, so do it.
Also use "compiler", rather than "CC", in the matrix for consistency
with .travis.yml and to avoid confusion between this property, which may
be empty, and CC environment variable, which is supposed to be always
defined.
Finally, show the path and the version of the compiler being used: this
was also done under Travis CI and there doesn't seem to be any reason
not to do it here.
There doesn't seem to be any good reason to use per-language caches, as
there are files that are common to all languages.
But we do want to use different keys for the different compilers and
even different compiler versions, as they yield different results when
compiling the same code.
Set the PATH to include the ccache directory for the subsequent steps,
including the "build" one, rather than only setting it locally for
configure, as this didn't have much effect at all.
Follow the convention of the GitHub built-in steps, such as the first
"Set up job" one and use capitalized verbs in imperative mood rather
than a mix of non-capitalized verbs and nouns.
This is a purely cosmetic change.
Bash is used by default under all platforms we're going to support here,
so there is no need to specify it explicitly. Moreover, by default the
appropriate shell options (such as "pipefail", as well as "e") are used,
and there is no need to override them.
There doesn't seem to be any reason to make them step-specific, so
simplify the workflow by doing it once and only once at the job level.
No real changes.
Don't rely on the auto-generated names which are not very useful: e.g.
currently something like "build (1, python, 3)" is shown, while this
commit changes it to just "python3" which is both shorter and more
readable.
This makes the matrix more compact, as we don't need to specify the OS
for the majority of the field (all of them now, but we'll add builds
using "macos-N" later, so it will still be necessary for some of them),
and the builds more stable than using which the current latest Ubuntu
version is.
