BSD Commit Log Search

   When a .Bd block is broken by an .It macro and hence implicitly ended,
   restore the ROFF_NOFILL global parser state to what is was before the block.
   This fixes an assertion failure in the HTML formatter triggered
   by the nonsensical sequence of macros .Bl .Bd -unfilled .It
   that kristaps@ found in the pam_smartcard(8) manual on macOS.

   Parse terminal replies with keys and apply them only when finished, so
   that a preceding Escape can be handled correctly. GitHub issue 5676 from
   Joan Fabrégat.

   Use the .Cm macro for keywords used as arguments to commands and options,
   mostly replacing dubious use of .Ic, but a few others as well;
   nicm@ agrees with the general direction.

   enable uncm(4) in the installer on amd64

   This gives us another way to do firmware-less installs over the network.

   typo + whitespace

   Add manual page for uncm(4).

   From Jan Schreiber

   add uncm(4), a driver for USB CDC NCM (i.e. USB phone tethering)

   Patch by Jan Schreiber

   ok deraadt@ brynet@

   Tested by Jan and myself with Android phones.
   Committed via uncm0

   Add table for destroy-unattached.

   riscv64 vector support: actually restore vcsr/vstart

   The asm statements used to restore said csr mentioned the value to
   restore as an output operand, not an input operand. This hurts as we
   write what happens to be in the allocated register (here "a1") to
   the csr and then struct vreg.

   Issue spotted in build failures due to libgcrypt (mail/notmuch) and to
   libjpeg (games/neverball-data). The diff fixes the testsuite for
   both ports.

   ok jsing@ kettenis@

   Do not allow drags started on status line to trigger pane key bindings,
   GitHub issue 5692.

   isakmpd: Bound the HASH coverage in cfg_verify_hash()

   The verifier assumed the HASH payload is the first payload and
   derived the covered length from the message length, reading past
   the end of the message when another payload precedes HASH.  Compute
   the covered range from the HASH payload's actual position and reject
   it if it extends past the message end.

   from markus@, ok me deraadt@

   Fix redraw and selection position when changing selection mode.

   Remove 'output from previous run as dimmed text' feature

   The dimmed text does not display correctly on xterm with white
   background and the console.

   Requested by / OK naddy@

   Improve text about status lines, GitHub issue 5677.

   Redraw selections after scrolling to line boundaries, GitHub issue 5689
   from bugreporter23 at gmail dot com.

   Close rtrx socket in child processes since it is opened up early.

   OK tb@ rcovelli@

   vmd(8): better handling of virtual switch descriptions

   in vm.conf, switch definitions can now include the keyword 'description':

   switch myswitch {
        description "foo"
        interface veb0
   }

   will cause vmd to change the host veb0 interface description to "foo".

   switch myswitch {
        description
        interface veb0
   }

   eg "description" by itself without a provided description,
   will simulate the 8.0 and previous behavior; vmd will change the host
   veb0 interface description to "switch%d-%s", where %d is the numerical

    [17 lines not shown]

   Do not allow screen deletion while a VT switch is in progress. If done
   asynchronously, depending on the display driver implementation of the
   switching primitives, this could open a race with a risk of memory
   corruption.

   Problem found and fix provided by Acts1631.
   from miod@

   this is errata/7.9/034_wsdisplay.patch.sig

   Do not allow screen deletion while a VT switch is in progress. If done
   asynchronously, depending on the display driver implementation of the
   switching primitives, this could open a race with a risk of memory
   corruption.

   Problem found and fix provided by Acts1631.
   from miod@

   this is errata/7.8/070_wsdisplay.patch.sig

   zlib regress: improve comment to state what it tests (gzwrite.c r1.7)

   sys/qcgpio: fix GPIO wakeup masking during suspend

   The suspend path disables pins marked for wakeup while leaving other
   GPIO interrupts enabled. Here, I reverse the condition to preserve
   wakeup sources and mask ordinary interrupts; restore the saved
   configuration on resume.

   OK: kettenis@

   zlib regress: unit test that segfaults without gzwrite.c 1.7

   based on https://gist.github.com/thesmartshadow/e0b9481792afb7c31e86fee1ff084490

   Fix iwi(4) build with IWI_DEBUG set. Use %td modifier for pointers
   difference.

   From Nazarenko Mykyta.

   Do not allow screen deletion while a VT switch is in progress. If done
   asynchronously, depending on the display driver implementation of the
   switching primitives, this could open a race with a risk of memory
   corruption.

   Problem found and fix provided by Acts1631.

   Fix mouse drag selection on the top and bottom lines in vi mode, GitHub
   issue 5667 from David Schweikert.

   zlib: fix buffer overwrite in non-blocking gzwrite

   Upstream commit df84af25

   use SOCK_NONBLOCK instead of fcntl O_NONBLOCK
   ok claudio

   use accept4() with SOCK_NONBLOCK instead of seperate fcntl with O_NONBLOCK
   ok bcook

   use SOCK_NONBLOCK instead of fcntl O_NONBLOCK; ok djm

   Hack up mt7925_mcu_hw_scan() to support 5Ghz channels as well.

   A better solution is needed when more channels are needed.
   The HW can scan 64 channels in one command which is enough for now.
   Also the probe req template need work but again this can wait.