I thought I went over all these several years ago already, but
apparently not or maybe back then not all websites were HTTPS-ready. Now
they mostly are, except for maybe one or two.
Similarly as done in Aug 2024 in Corrade. When these were a part of the
function signature, they ended up being encoded into the exported
symbol. There are still cases of StridedArrayView slice() having
enable_if in the signature, which amounts to about 18 kB symbols in all
libMagnum*-d.so libraries, but apart from that this is the state before:
$ strings libMagnum*-d.so | grep enable_if | grep -v slice | wc -c
29591
And this is after. All of those are coming from STL, thus from
old or deprecated APIs that still use std::vector, std::tuple and such,
and from the few std::sort() uses.
$ strings libMagnum*-d.so | grep enable_if | grep -v slice | wc -c
4103
In a non-deprecated build it's just this, which is a 10x reduction.
Can't really do much about these maybe exceút for implementing my own
swap() specializations (sigh?), but I think it's fine.
$ strings libMagnum*-d.so | grep enable_if | grep -v slice | wc -c
2904
I also made it consistently use
typename std::enable_if<..., int>::type = 0
instead of
class = typename std::enable_if<...>::type
because the former works correctly also in presence of overloads and
having it used consistently everywhere makes it easier to grep & change
later. All SFINAE is now also excluded from Doxygen output, because it
doesn't make much sense there. It's better to just explain the
restriction in words than with this nasty hack.
This warning isn't enabled by default by either -Wall or -Wextra, and
for a good reason. When I do that, it fires also for all uses of
Platform::Application, basically any subclass causes that warning to be
printed as well, even if given subclass is used just once, in an
anonymous namespace, and only through the MAGNUM_APPLICATION_MAIN()
macro. The only solution on Clang is to make such subclasses `final`,
but on GCC not even that helps and one is forced to really make the base
destructor virtual. Which doesn't achieve anything, only adds yet
another entry to the vtable.
So, in the SceneGraph it fixes one particular use case where the warning
was triggered, and since the fix isn't so invasive I'm fine with that.
For Platform I have no idea how an acceptable fix would look like, so I
hope nobody needs any of that anytime soon.
See also https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102168, especially
the part saying "This warning should not be used.".
They're not parsed since 6b22a11170
(2020), so there's no point in keeping those workarounds. They're only
kept in utility application sources as they're parsed for pages, and in
tweakable implementations where it's easier to just copypaste the whole
ifdef expression from the header every time instead of modifying it to
not include DOXYGEN_GENERATING_OUTPUT.
Compared to Corrade, the improvement in compile time is about a minute
cumulative across all cores, or about 8 seconds on an 8-core system (~2
minutes before, ~1:52 after). Not bad at all. And this is with a
deprecated build, the non-deprecated build is 1:48 -> 1:41.
This was due to the internal counter in the object being just 2 bytes,
which was made in order to optimize the Object class size. In practice
however, not counting the vtables, the class has SEVEN pointers and then
a transformation part that's in most cases multiple of 8. Which means
that on 64 bit platforms there was always 5 bytes of padding next to
this 2-byte counter and 1-byte flags, thus no reason to have it 2-byte
anyway.
On 32-bit platforms this *might* cause the Object to get bigger, yes,
but as it's heap-allocated it's more likely that it's put on a
16-byte-aligned address and thus occupying a multiple of 16 bytes
anyway. In any case, this won't really hurt, because the SceneGraph
situation is bad enough already due to all the pointer chasings and
overhead from loose allocations.
Similar to the change done in Corrade, see the commit for details:
878624ac36
Wow, this is probably the most backwards-compatibility code I've ever
written. Can't wait until I can drop all that.
It limits the support for CMake 3.12+, but it's much less verbose and I
don't expect people to use ancient CMake versions with IDEs like Xcode
or VS anyway, so this should be fine.
Before the parent link was gone before destructing the children and that
just didn't make sense. The test added in previous commit now passes as
expected.
This is how it should be but isn't -- the Object inheritance makes it
so first the parent pointer is cleared and only then the children get
destructed. That doesn't really make sense (and I doubt any code relied
on that), so I'll flip it.
Should make new things more discoverable, avoid confusion when a
documented API isn't there and reduce the need for maintaining multiple
separate versions of the docs.
Vladimír Vondruš