I didn't notice them in my manual builds because I mounted tmpfs on
$LFS/tmp, but noted it in two jhalfs builds.
I've already sent a patch to libc-alpha to fix them. See link in the
added comment. The patch message explains why they depend on tmpfs or
not.
GCC-15 has bumped the C++ standard to C++14, and the required GCC
version to 5.4 because 5.3 and earlier may generate some wrong code with
C++14 and constexpr (that GCC-15 code base uses).
LC_ALL=C.UTF-8 removes a spurious test failure at run-builtins line 365.
And for some tests failing due to missing locale it turns unreadable
characters into some incorrect but readable characters.
Also update known failures: the known ulimit -c failure has a different
line number; the removal of ja_JP.SJIS and the newly introduced
zh_TW.BIG5 tests also introduce several new failures.
Update to iana-etc-20250807.
Update to vim-9.1.1829.
Update to ncurses-6.5-20250809.
Update to Python-3.13.7 (Security Update).
Update to linux-6.16.1.
Update to iproute2-6.16.0.
Update to systemd-257.8.
Add a modification to glibc from upstream
to solve an incompatibility with valgrind.
A frequently asked question is "why not using -i.orig instead of -i for
${some_random_sed_command_in_book}" but I really expect people to learn
from this example and do it on their own if they really need...
XML::Parser.
It shows up as a warning while building XML::Parser, but the
functionality of the package doesn't seem to be impacted for our use
case at least.
The upstream kernel has dropped it because real 32-bit hardware almost
never has RAM > 4 GiB. The users running 32-bit software on 64-bit
hardware should really switch to MLFS now.
Regenerating the XML files also makes the prompt for X2APIC consistent
with the 6.15 kernel.
