Allocate TLS 1.3 receive buffers lazily
The receive buffer is not used for records created for sending. Avoid
allocating a maximum-sized buffer in tls13_record_new() and instead
allocate a header-sized buffer when tls13_record_recv() is first called.
The buffer will grow as needed once the record length is known.
This avoids an unnecessary allocation for outgoing records and reduces
the initial allocation size for incoming records.
ok tb jsing
init(8): remove support for the window= field in /etc/ttys
This feature was introduced in 4.3BSD, and announced in "Bug Fixes
and Changes in 4.3BSD", dated April 15, 1986. The feature provided
a way to start a window system associated with a tty before
starting getty(8).
However the implementation has a few negatives and these days
graphical environments are better started in other ways.
In particular:
- the window command is run with root privileges
- init(8) does not track or supervise the process as part of the
tty session.
- reloading /etc/ttys by sending SIGHUP to init, can leave a stale
window command cached incorrectly
- repeated reloads leak memory
Rather than fix all these issues, remove the feature instead.
[5 lines not shown]
llvm/sparc64: match BFD -N alignment
BFD aligns nonpaged PT_LOAD segments to their maximum contained section
alignment. Do the same for SPARCV9 instead of retaining the 1 MiB
maximum page alignment.
This removes roughly 1 MiB of leading padding from ofwboot without
changing its load address or memory layout.
OK: deraadt@, kettenis@
llvm/sparc64: match BFD -N alignment
BFD aligns nonpaged PT_LOAD segments to their maximum contained section
alignment. Do the same for SPARCV9 instead of retaining the 1 MiB
maximum page alignment.
This removes roughly 1 MiB of leading padding from ofwboot without
changing its load address or memory layout.
OK: deraadt@, kettenis@
Overhaul printing of whitespace before tbl(7) in man(7).
This is the last major piece i intend to put in for making mandoc(1)
compatible with groff-1.23.
In spite of the change in man_term.c 1.194 (August 2022)
which removed unconditional vertical space before .TS,
GNU eqn(1) still prints vertical space in some cases
where the layout contains vertical lines.
rtorrent: add default fallback cache line size
This type of hard failure is just dumb as it forced many patches for
both Linux and non-Linux systems out there. Fix the build on riscv64
and most probably on mips64, sparc64 etc
seamonkey: unbreak build with libcxx22, discussed with landry
hunspell_fopen_hooks.h contains a preprocessor hack to replace the standard
fopen() with one supporting readahead. This used to work because libcxx's
fstream header used a naked fopen() call. libcxx22 changed to std::fopen(),
so the build breaks because there is no std::hunspell_fopen_readahead():
/usr/include/c++/v1/fstream:716:20: error: no member named 'hunspell_fopen_readahead' in namespace 'std'; did you mean simply 'hunspell_fopen_readahead'?
716 | return __do_open(std::fopen(__s, __mdstr), __mode);
geo/josm: update to svn r19613, from MAINTAINER Holger Mikolon
original diff from Christoph Liebender moving to jdk25, but this has
memory issues at runtime as noted by MAINTAINER so stay with jdk17.
