trampoline exclusively, thus relegating "sigcontext"-style handlers (which
have not been documented for many years now) to the dustbin of the compat
library.
COPTS.*+= -Wno-error=.*
lines for building argon2 sources, by fixing the problems at source.
Addresses Rin Okuyama's concerns on tech-userlevel/tech-crypto in
Message-ID: <f1eab199-3607-bb05-15cc-033e38721ba4@gmail.com>
command, so that the command cannot appear to be more options
(which always then fails, as there would be no arg for "-c" to
treat as the command string in that case).
For the full (LONG) explanation, see:
http://mail-index.netbsd.org/current-users/2021/10/29/msg041629.html
I don't know what an alow is, maybe some cousin of the alot.
Awaiting a nature documentary by Allie Brosh about these perhaps more
elusive beasts.
PR misc/56473, from jschauma
- sparc and sparc64 were not using version 0 sigcontext when there were
no arguments in the signal version. This was probably a bug.
- vax is using +1 the version numbers of the other archs.
- Only hppa was defining __LIBC12_SOURCE__ so it was getting a working
sigcontext before. all the other ports that supported sigcontext had
the compat code disabled.
[pointed out by thorpej, thanks!]
If we want to remove sigcontext support from userland at least now there
is less work to do so.
Move cursor to end of the currently active field and sync the cursor
location so the cursor get positioned correctly when the form window
is refreshed.
write buffer associated with the file descriptor is empty. This is
currently implemented only for sockets, and is intended primarily to
provide visibility to applications that all previously written data
has been acknowledged by the TCP layer on the receiver. Compatible
with the same filter in FreeBSD.
Release 2021e - 2021-10-21 18:41:00 -0700
Changes to code
none
Release 2021d - 2021-10-15 13:48:18 -0700
Changes to code
'zic -r' now uses "-00" time zone abbreviations for intervals
with UT offsets that are unspecified due to -r truncation.
This implements a change in draft Internet RFC 8536bis.
Release 2021c - 2021-10-01 14:21:49 -0700
Changes to code
Fix a bug in 'zic -b fat' that caused old timestamps to be
mishandled in 32-bit-only readers (problem reported by Daniel
Fischer).
Changes to documentation
Distribute the SECURITY file (problem reported by Andreas Radke).
Release 2021b - 2021-09-24 16:23:00 -0700
Changes to maintenance procedure
The new file SECURITY covers how to report security-related bugs.
Several backward-compatibility links have been moved to the
'backward' file. These links, which range from Africa/Addis_Ababa
to Pacific/Saipan, are only for compatibility with now-obsolete
guidelines suggesting an entry for every ISO 3166 code.
The intercontinental convenience links Asia/Istanbul and
Europe/Nicosia have also been moved to 'backward'.
Changes to code
zic now creates each output file or link atomically,
possibly by creating a temporary file and then renaming it.
This avoids races where a TZ setting would temporarily stop
working while zic was installing a replacement file or link.
zic -L no longer omits the POSIX TZ string in its output.
Starting with 2020a, zic -L truncated its output according to the
"Expires" directive or "#expires" comment in the leapseconds file.
The resulting TZif files omitted daylight saving transitions after
the leap second table expired, which led to far less-accurate
predictions of times after the expiry. Although future timestamps
cannot be converted accurately in the presence of leap seconds, it
is more accurate to convert near-future timestamps with a few
seconds error than with an hour error, so zic -L no longer
truncates output in this way.
Instead, when zic -L is given the "Expires" directive, it now
outputs the expiration by appending a no-change entry to the leap
second table. Although this should work well with most TZif
readers, it does not conform to Internet RFC 8536 and some pickier
clients (including tzdb 2017c through 2021a) reject it, so
"Expires" directives are currently disabled by default. To enable
them, set the EXPIRES_LINE Makefile variable. If a TZif file uses
this new feature it is marked with a new TZif version number 4,
a format intended to be documented in a successor to RFC 8536.
zic -L LEAPFILE -r @LO no longer generates an invalid TZif file
that omits leap second information for the range LO..B when LO
falls between two leap seconds A and B. Instead, it generates a
TZif version 4 file that represents the previously-missing
information.
The TZif reader now allows the leap second table to begin with a
correction other than -1 or +1, and to contain adjacent
transitions with equal corrections. This supports TZif version 4.
The TZif reader now lets leap seconds occur less than 28 days
apart. This supports possible future TZif extensions.
Fix bug that caused 'localtime' etc. to crash when TZ was
set to a all-year DST string like "EST5EDT4,0/0,J365/25" that does
not conform to POSIX but does conform to Internet RFC 8536.
Fix another bug that caused 'localtime' etc. to crash when TZ was
set to a POSIX-conforming but unusual TZ string like
"EST5EDT4,0/0,J365/0", where almost all the year is DST.
Fix yet another bug that caused 'localtime' etc. to mishandle slim
TZif files containing leap seconds after the last explicit
transition in the table, or when handling far-future timestamps
in slim TZif files lacking leap seconds.
Fix localtime misbehavior involving positive leap seconds.
This change affects only behavior for "right" system time,
which contains leap seconds, and only if the UT offset is
not a multiple of 60 seconds when a positive leap second occurs.
(No such timezone exists in tzdb, luckily.) Without the fix,
the timestamp was ambiguous during a positive leap second.
With the fix, any seconds occurring after a positive leap second
and within the same localtime minute are counted through 60, not
through 59; their UT offset (tm_gmtoff) is the same as before.
Here is how the fix affects timestamps in a timezone with UT
offset +01:23:45 (5025 seconds) and with a positive leap second at
1972-06-30 23:59:60 UTC (78796800):
time_t without the fix with the fix
78796800 1972-07-01 01:23:45 1972-07-01 01:23:45 (leap second)
78796801 1972-07-01 01:23:45 1972-07-01 01:23:46
...
78796815 1972-07-01 01:23:59 1972-07-01 01:23:60
78796816 1972-07-01 01:24:00 1972-07-01 01:24:00
Fix an unlikely bug that caused 'localtime' etc. to misbehave if
civil time changes a few seconds before time_t wraps around, when
leap seconds are enabled.
Fix bug in zic -r; in some cases, the dummy time type after the
last time transition disagreed with the TZ string, contrary to
Internet RFC 8563 section 3.3.
Fix a bug with 'zic -r @X' when X is a negative leap second that
has a nonnegative correction. Without the fix, the output file
was truncated so that X appeared to be a positive leap second.
Fix a similar, even-less-likely bug when truncating at a positive
leap second that has a nonpositive correction.
zic -r now reports an error if given rolling leap seconds, as this
usage has never generally worked and is evidently unused.
zic now generates a POSIX-conforming TZ string for TZif files
where all-year DST is predicted for the indefinite future.
For example, for all-year Eastern Daylight Time, zic now generates
"XXX3EDT4,0/0,J365/23" where it previously generated
"EST5EDT,0/0,J365/25" or "". (Thanks to Michael Deckers for
noting the possibility of POSIX conformance.)
zic.c no longer requires sys/wait.h (thanks to spazmodius for
noting it wasn't needed).
When reading slim TZif files, zdump no longer mishandles leap
seconds on the rare platforms where time_t counts leap seconds,
fixing a bug introduced in 2014g.
zdump -v now outputs timestamps at boundaries of what localtime
and gmtime can represent, instead of the less-useful timestamps
one day after the minimum and one day before the maximum.
(Thanks to Arthur David Olson for prototype code, and to Manuela
Friedrich for debugging help.)
zdump's -c and -t options are now consistently inclusive for the
lower time bound and exclusive for the upper. Formerly they were
inconsistent. (Confusion noted by Martin Burnicki.)
Changes to build procedure
You can now compile with -DHAVE_MALLOC_ERRNO=0 to port to
non-POSIX hosts where malloc doesn't set errno.
(Problem reported by Jan Engelhardt.)
Changes to documentation
tzfile.5 better matches a draft successor to RFC 8536
<https://datatracker.ietf.org/doc/draft-murchison-rfc8536bis/01/>.
If the parameters are unspecified:
- Set the default memory consumption based on the amount of memory
available to userspace.
The algorithm actually slows down incredibly quickly as the "memory"
parameter is increased. We want to avoid running out of memory on low
memory systems, but increase the difficulty of bruteforcing passwords
from systems with a lot of memory. At the same time, we want to avoid
problems when concurrent logins are happening.
- Run a hashing loop for one second with steadily increasing "time"
until we settle on a value for "time". We want to use as much CPU time
as reasonable for computing the password hash without making logins
inconvenient.
- Centralize vnode kevent handling in the VOP_*() wrappers, rather than
forcing each individual file system to deal with it (except VOP_RENAME(),
because VOP_RENAME() is a mess and we currently have 2 different ways
of handling it; at least it's reasonably well-centralized in the "new"
way).
- Add support for NOTE_OPEN, NOTE_CLOSE, NOTE_CLOSE_WRITE, and NOTE_READ,
compatible with the same events in FreeBSD.
- Track which kevent notifications clients are interested in receiving
to avoid doing work for events no one cares about (avoiding, e.g.
taking locks and traversing the klist to send a NOTE_WRITE when
someone is merely watching for a file to be deleted, for example).
In support of the above:
- Add support in vnode_if.sh for specifying PRE- and POST-op handlers,
to be invoked before and after vop_pre() and vop_post(), respectively.
Basic idea from FreeBSD, but implemented differently.
- Add support in vnode_if.sh for specifying CONTEXT fields in the
vop_*_args structures. These context fields are used to convey information
between the file system VOP function and the VOP wrapper, but do not
occupy an argument slot in the VOP_*() call itself. These context fields
are initialized and subsequently interpreted by PRE- and POST-op handlers.
- Version VOP_REMOVE(), uses the a context field for the file system to report
back the resulting link count of the target vnode. Return this in tmpfs,
udf, nfs, chfs, ext2fs, lfs, and ufs.
NetBSD 9.99.92.
Fix a misinterpretation of the role of field fore and back. Now field
pad is rendered in the back attribute always instead of only being used
for the non-current field.
Don't move the cursor when getch is called if input is not going to
be echoed. Lynx uses the cursor to mark the current selection in a
pop-up, previously the cursor was being relocated when getch was called
which broke the Lynx item marking.
This changes argon2 from a separately built library into a private
dependency of libcrypt and removes the argon2(1) utility (we already
have pwhash(1)). Giving libcrypt more library dependencies
complicates things (e.g. libcrypt is a dependency of openssl).
pthreads support gets disabled in argon2 for similar reasons.
For testing argon2, we rely on the libcrypt test suite.
There are too many minor variations regarding padding and exact alphabet
to safely use the implementation in libc or an existing implementation
in libcrypt.
Allow the version number to be unspecified as in the argon2 upstream
test suite, properly defaulting to a version if the v= block is
entirely missing, and treating the remaining block as parameters.
Fix a null pointer derefence when the encoded password is unspecified
in the settings string.
process that that calls getaddrinfo(3) will end up cacheing the
kqueue(2) file descriptor in its res_state structure. If that
process fork(2)s and calls getaddrinfo(3) again might end up closing
that cached file descriptor which can end up pointing to a different
file object than the kqueue(2) original one. To fix this, associate
the kqueue(2) file descriptor with the process id that created it,
and don't close(2) it if it is being closed from a different process.
An alternative fix would be to attach the resolver to a fork(2) hook
to cleanup the res_state, but handling it internally in the resolver
is less intrusive. This was discovered by Dima Veselov when using
the FreeRADIUS package.
10^9 is 1e9 (it's "e" that spells 10 here), not 10e9. The target
audience of this man page is not likely to be very fluent in floating
point, so avoid significand/exponent spelling and use the spelling
that it is familiar with.
Fix build failure when built as libhack, due to missing fileio.h.
IMO, this should be cleaner than generating fileio.h for {get,put}win(3),
that are useless for install media.
For __newwin() and __resizewin(), the line hash was calculated as if
HAVE_WCHAR is disabled.
Fix this bug by refactoring __hash_line() function, which calculates
the line hash by an appropriate method.
This is the last remaining instance of lint warning 215, which is about
implicitly defined functions. Since C99, these implicit declarations
have not been allowed anymore.
Since 2021-02-28, lint does not warn about constant conditions if they
involve sizeof.
The fallthrough annotations have never been necessary for case labels
that directly follow each other.
The lint comments CONSTCOND and LINTED were not necessary. It is
simpler to just specify what to free. GCC optimizes free(NULL) to be a
no-op.
No functional change.
Since that macro can expand to an empty token list, it adds its own
semicolon as needed. Removing the extra semicolon fixes the lint
warnings about empty declarations. These empty declarations are a GCC
extension.
Alpha has no hardware-assisted fused multiply-add, so the fallback
C implementation must be used. A basic smoke test shows the
implementation works for
"x * x - 1.0f" vs "fmaf(x, x, -1.0f)" with "volatile float x = 1.0008f;"
not return 0, the synthetic writes from funopen/fmemopen could. This avoids
infinite loops in >= test19 in fmemopen, but the tests still fail, perhaps
because they assume the previous behavior, where flush does not adjust the
stdio pointers on error.
In namespace.h 1.102 from 2005-07-16, the #ifndef was added to fix the
broken lint2 pass. Somewhere between 2005 and now, lint has been
changed so that it does not need this workaround anymore.
After usr.bin/xlint/lint1/tree.c 1.303 from 2021-06-30, the #ifndef in
namespace.h caused the only warning about an implicit function
declaration in the whole NetBSD tree, in lib/libc/gen/sysconf.c:149,
where _getpagesize was not defined, but only getpagesize.
verbose mechanism with MIIVERBOSE. This reduces some duplicated code
and allows us to once again permit auto-unload of MIIVERBOSE.
Change details:
* Update dev/devlist2h.awk to accomodate miidevs, including generation
of MII_STR_oui_model definitions and use of oui and model rather than
vendor and product. This also changes the compressed data in the
xxxdevs_data.h files to uint32_t (since mii oui's are up to 6 hex
digits long)
* Update a couple of phy drivers to use new calls to get verbose data
* Regen all of the xxxdevs{,_data}.h files (separate commit, coming
very soon)
* Update mii/mii_verbose.[ch] and mii/mii_physubr.c to use the various
DEV_VERBOSE_xxx macros
* Update the pci, usb, and hdaudio code as needed, to #include the
xxxdevs.h files (in order to get the proper printf format strings)
* Since dev/dev_verbose.c now uses non-literal printf format strings,
(to deal with the vendor/product vs oui/model issue), we need to
make sure it gets compiled with -Wno-error=format-nonliteral, even
in userland's libpci and librumpdev!
* Bump kernel version for the change in module interfaces
Welcome to 9.99.86!
XXX It might be useful in the future to extend the MII_STR_oui_model
XXX definitions to PCI as well (and perhaps USB and HDAUDIO). This
XXX would allow for a single centralized location for the products'
XXX descriptions, rather than being dispersed among individual
XXX drivers' xxx_match tables.
Fix ABI mismatch for armhf runtime routines for floating-point arithmetics;
For hard-float arm variants, provide
(1) generic runtime routines with correct calling convention, and
(2) EABI runtime routines at the same time.
I've confirmed that no binary changes for kernels.
LGTM by skrll
Use wdwitch to determine the width of a wide character on the screen not
the number from mbrtowc which is the number of bytes in the character.
Thanks to Michael Forney for spotting this.
Correct addstr behaviour so it truncates the string in the case where
a string is added on the bottom line of a window where scrolling is
disabled as per the SUSv2 specification.
lp->hash is not initialized at this point. Since the hash is
calculated in chunks using __hash_more(), it needs to be initialized
to 0 first (just as in doupdate()).
Detected with valgrind while running python's test suite when
debugging an unrelated issue.
ok uwe@
The longjmp() function shall not cause setjmp() to return 0; if val
is 0, setjmp() shall return 1.
Fixes the _longjmp_zero and longjmp_zero test cases in the t_setjmp
test on alpha.
Fix typos (2nd acst should have been acdt), 0550 for ist should be 0530
(5.5 hours is not 5 hours and 50 minutes...).
Comment out the zp4 zp5 and zp6 zone names. They are supposedly supported
by the source (they're in the table) but cannot work, as the parsedate
lexer doesn't allow a "word" to start with an alpha and also contain
digits. Maybe (just maybe) that could be fixed sometime, but since these
have never worked, and no-one has ever seemed to miss them, and they're the
only words which are of that form, for now, just stop pretending they work.
- Define various new AFMT_*. These are not returned as supported
formats by SNDCTL_DSP_GETFMTS, because it would be very silly to
have Vorbis in the kernel.
- Implement PLAYTGT and RECSRC. For each NetBSD audio device
we only return one playback and recording source, "primary".
- Return preferred channel configuration in capabilities.
Either DSP_CH_STEREO, DSP_CH_MONO, or DSP_CH_MULTI
depending on the current hardware format.
- SNDCTL_DSP_HALT_* simply flushes the audio device.
on arm64eb resuming vi(1) would often crash. in makech(), the 'csp'
variable is either set to current window data, or a local stack
variable's address '&blank'. the window data has many lines of info
stored, and 'csp++' is used per line here. unfortunately, a case
existed where 'csp++' operated on csp initialised from '&blank' which
eventually crashes when, on my display with 160 columns and 'csp + 155'
exceeds the mapped stack and crashes.
match the '!_cursesi_screen->curwin' conditional that initialises csp,
and avoid csp++ here. assert() that csp != &blank in both places that
modify csp.
thanks to jdc@ and mlelstv@.
XXX: possibly also should avoid the putch() here as well.
Darwin in libc compilation
We build the regex code in tools without nls. We don't include any nls headers
for that, but on Darwin wint_t gets defined, so we end up with a compilation
error. The cleaner fix would have been to always use regex_foo_t types, but
the minimal fix is to only do this for the tools build, using cpp.
memset() and emitting calls to memset() where it seems code that looks
like what memset() does, which ends up recursing and blowing the stack.
this makes mips userland with GCC 10 work.
signals. This introduces a new kernel and userland NVMM version indicating
this support.
Patch by Kamil Rytarowski <kamil@netbsd.org> and committed on his request.
This is the missing libnvmm part I forgot to include in the origional commit.
Before it was based on the heuristic that we were not supplied an
attempted_completion_function, which worked well because programs
that supplied that function were not shells and did not want/understand
shell quoting. Recently though Piotr Stefaniak wanted to enhance command
completion for the Bourne Shell and this could benefit quoting the returned
command. This function adds an extra flags argument that controls that quoting.
some drivers (not hdaudio(4), but uaudio(4), eap(4), sb(4), various other
old cards) will return error if a AUDIO_MIXER_VALUE is requested and the
number of channels is not specified as input. this is not documented as
well as it should be, unfortunately.
andvar