max_align_t does not depend on any pre-C99 or !C++ language feature.
This structure is in use in 3rd party essential C++ code as an extension
for older language revisions and in gnu99 code in the NetBSD distribution
(RTLD's build rules define -std=gnu11 just for exposing this struct).
Exposing max_align_t from the central NetBSD header avoid duplicate
definition in 3rd party code that could differ and produce ABI mismatched
binaries between -std= revisions.
This problem does not exist on OSs like Linux as they get this namespace
visibility defined inside LLVM or GNU toolchain headers. NetBSD ships with
its own stddef.h, rather than relying on a toolchain and its internal
extensions.
the main effect of this is to make GCC and other libiberty using
tools use /tmp instead of /var/tmp for compiler temp files,
which can be a bottleneck on larger systems.
a survey of other platforms shows only OSX also uses /var/tmp,
everyone else has switched to /tmp long ago.
cons: some smaller systems may have a smaller /tmp than /var/tmp,
and this may cause builds to fail with out of space earlier.
point the build to /var/tmp using TMPDIR in this case.
one can argue that setting TMPDIR would work around this, but we
want to have the effect for all users without having special setup.
Remove const from the 2nd argument.
const char ** and char ** are incompatible types and it was a cost to keep
the technically incompatible form for a more purist variation. NetBSD was
almost the last alive OS to still keep the const argument (known leftovers:
Minix and Illumos).
Keep the const form for the internal purposes inside citrus and rump.
Address the build breakage fallout in the same change.
There are no ABI changes.
Change accepted by core@.
The existing definition caused issues as GCC only provides _Static_assert
when building C11 code.
This follows the C standard: static_assert available since C11.
Fixes https://rt.perl.org/Public/Bug/Display.html?id=134023
- -Wstrict-prototypes is not available for C++, so don't try to
ignore it for C++.
- remove many _DIAGASSERT() checks against not NULL for functions
with arguments with nonnull attributes. in two cases, leave
code behind that should set defaults to "(null)".
- use -Wno-error=frame-address for i386 mcount, as it seems valid
to assume the caller will have a frame.fair
functions are used for destructors of thread_local objects.
If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.
Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.
This as discussed on current-users in the thread
entitled:
Proposal: new libc/libutil functions to map SIGXXXX <-> "XXXX"
that can be found (starting at):
http://mail-index.netbsd.org/current-users/2017/04/28/msg031600.html
These functions provide the mechanism to enable applications
to divorce themselves from internal details of the signal
implementation.
Libc minor bumped, prototypes in <signal.h>, sets lists updated (and sorted).
One and all: feel free to improve the sources & man page (etc), but
please do not change the function signatures without discussion.
joerg