libaltq, cbq: add two options to interface

- no-control: don't create a control class automatically
- no-tbr: don't install TBR

(ozaki-r)

altq, cbq: allow the root class to be default

(ozaki-r)

altq, cbq: allow no control class

(ozaki-r)

Via Jess Thrysoee, from Adrian Bunk: Fix libedit build on Linux/Alpha
Alpha is the only Linux architecture that has SIGINFO:
    https://sources.debian.org/src/manpages/5.10-1/man7/signal.7/#L522

But even on Alpha Ctrl-T is not supported, and therefore no VSTATUS:
    https://sources.debian.org/src/manpages/5.10-1/man3/termios.3/#L603-L608

For consistency check both signal existence and character existence

(christos)

The second REQ_RESET USB request by reset_chip() makes LCR, PRE and DIV
registers to have default values 8N1 and 19200 just as the first req does.
Nuke reset_chip() and rundandant set_dte_rate() to amend.
While here added some narrative about prescale/divisor calculation.

(nisimura)

Hook up ffsext_strategy to fifos. Pointed out by dholland@

(christos)

unset IFF_PROMISC at bpf_detach()

Doing "d->bd_promisc = 0" is that bpf_detach() does not call
ifpromisc(ifp, 0). Currently, there is no reason for
this behavior so that it is removed.
In addition to the change, the workaround for it in vlan(4)
is also removed.

(yamaguchi)

Added a test about clearing IFF_PROMISC at vlan_unconfig

This test is related to PR/49196

(yamaguchi)

find libgbm in libgbm.old.

(mrg)

Make an mbuf writable before un-tagging

(yamaguchi)

tests: add tests for ALTQ CBQ

(ozaki-r)

Rump-ify ALTQ (librumpnet_altq.so)

(ozaki-r)

shmif: support ALTQ

(ozaki-r)

Disable the mutex_spin_{enter,exit}() fast-path stubs while I investigate
some an odd IPL-related panic that seems to be related (that I can fairly
reliabily reproduce on an LCA45).

(thorpej)

lint: fix wording of message for incomplete struct (2021-07-04)

A struct member is not an argument.  In err.c 1.124 from 2021-07-04, the
wording had been changed, without noticing that the message was used
twice: once for an argument, once for a struct member.

(rillig)

tests/lint: add tests for empty and incomplete structs

(rillig)

tests/lint: test binary integer literals and underscores

(rillig)

tests/lint: skip test for floating point overflow on alpha

On alpha and a few other platforms (see t_integration.sh), 'long double'
has 128 bit, which under IEEE 754 rules means a decimal exponent of up
to 4932.

(rillig)

tests/lint: take archsubdir from usr.bin/xlint/Makefile.inc

This fixes the tests on the various ARM platforms where the platform
name does not correspond to MACHINE_ARCH, such as earmv7hf.

(rillig)

Swap encryption no longer experimental or default-off.

(nia)

Fix non-MULTIPROCESSOR build (e.g. INSTALL kernel).

(thorpej)

PR install/56303: when using xz files enable swap during set extraction
if the machine does not have more than 256MB of RAM.

(martin)

lagg: fix typo for ALTQ

(ozaki-r)

altq, cbq: calculate a delay period based on a borrowed class

A delay period was calculated based on a suspended class even if the
class can borrow bandwidth from another class.  In that case a delay
period should be calculated based on a borrowed class.  Otherwise
a borrowing class can't borrow bandwidth correctly.

(ozaki-r)

altq, cbq: apply the do-while idiom to TV_* macros

(ozaki-r)

altq, cbq: don't increment the xmit counter of the root class twice wrongly

(ozaki-r)

Fix a comment

(skrll)

Fix a comment

(skrll)

move the i18n modules into libX11 directly.

fixes at least alacritty, and saves memory too.

(mrg)

also look for libgbm in ${OLD_SUFFIX}.

(mrg)

As with membar_producer() and membar_sync(), initialize the fast-path
lock stubs with "unop" isns where memory barriers need to be, and if
we detect a multiprocessor system, patch those locations with "mb" insns.

(thorpej)

lint: order type declarations in grammar, group by C99 section

No functional change.

(rillig)

lint: reorder grammar rules in the same way as in C99

The code coverage before and after this change is exactly the same,
except of course for cgram.y and cgram.c.

No functional change.

(rillig)

Revert 1.4, that just trades one problem for a different one
(splitting the message on % instead of white space, which affects
less messages, but makes a bigger mess of them when it happens).

The real problem is that the expansion of the message was unquoted,
which allowed the field splitting to happen at all (which was mitigated
by the way printf rescans its format string for each arg when there are
more args than conversions in the format) otherwise it would have been
a much bigger mess (both times).

Just add quotes where quotes are needed, no more splitting, all good.

(kre)

mknative-gmp for mipsn64* and GMP 6.2.1.

(mrg)

add mipsn64 support, so native-gmp works, and likely, build.sh tools
on a mipsn64 host.

(mrg)

tests/lint: test trying to define a ()-less function

(rillig)

Printing out our model name is not considered an error message.

(thorpej)

Optimized fast-paths for mutex_spin_enter() and mutex_spin_exit().

(thorpej)

Add Jasper Lake support.

(msaitoh)

Regen.

(msaitoh)

Add Intel Jasper Lake devices and Reealtek RTL8821CE.

(msaitoh)

Add 0x96(Elkhart Lake) and 0x9c(Jasper Lake).
Not listed in SDM but listed in those spec update documents.

(msaitoh)

look for libglapi in ${OLD_SUFFIX}.  noted by gson.

(mrg)

mknative-gmp for GMP 6.2.1 and most platforms.

(mrg)

regen for GMP 6.2.1 and i386 and arm64.  update docs so i don't forget
that i mostly got 'native-gmp' working last year.

(mrg)

regen gmp parts for amd64 and gmp 6.2.1.

(mrg)

s/Precision Time Management/Precision Time Measurement/

(msaitoh)

define a X86_32_ASAN_BIT_OFFSET macro that defaults differently on
netbsd/i386 than other x86-32 targets.

fixes PR#56280.

(mrg)

tests/lint: rework ATF test driver

To skip tests that work only on particular platforms, a simple 'eval'
from the shell does not work since it does not exit on failure.  Fix
this by storing the commands in a local variable first.

Remove configuration knobs 'lint1-only-if-arch' and 'lint1-skip-if-arch'
since they are unused.

When skipping a test, actually mark it as skipped in the ATF statistics.

(rillig)

switch Mesa 19 to use MesaLib.old subdir.

(mrg)

note that new GMP may require some cleaning.

(mrg)

GMP is updated to 6.2.1 now.

(mrg)

fix a couple of issues found while testing new drm.

- actually check the return value of radeon_bo_open() in a three places
- for create_pixmap_for_fbcon(), release the bo when jumping to
  'out_free_fb'.  fixes memory leak.

(mrg)

merge GMP 6.2.1.

(mrg)

lint: reorder grammar rules for expressions according to C18

No functional change.

(rillig)

copy the Mesa 19.7.1 build framework into .old.

(mrg)

xorg/lib stuff to enable mesa.old, prepare for new mesa

- rename OLD_PREFIX to OLD_SUFFIX, since it is
- include mesa-which.mk and use ${OLD_SUFFIX} in several places
- remove mesa < 18 support

(mrg)

remove mesa < 18 support

(mrg)

merge mesa 19.1.7 into MesaLib.old.

(mrg)

lint: make _Generic a primary-expression

C11 says so, and unless the _Generic expression was wrapped in
parentheses, it was not possible before to use it as a function call
expression.

(rillig)

tests/lint: demonstrate that _Generic must be a primary_expression

(rillig)

lint: rename grammar rule 'term' to 'cast_expression'

No functional change.

(rillig)

note that MPC is now at 1.2.1, and that GMP has 6.2.1 available.

(mrg)

lint: reorganize grammar rules for unary expressions

According to GCC, __builtin_offsetof is a primary-expression.

According to GCC, __real__, __imag__ and __extension__ are
unary-operator, just like '!' and '~'.

According to C11, _Alignof is a unary-operator.

(rillig)

lint: remove grammar conflict for __real__ and __imag__

The rules were redundant since a term can be parenthesized by itself,
there is no need for an extra rule.  The rules for '__real__(term)' and
'__imag__(term)' were never reduced since these cases were handled by
the rules '__real__ term' and '__imag__ term' a few lines above.

No functional change.

(rillig)

tests/lint: test precedence of __real__ and cast_expression

(rillig)

lint: add quotes around placeholders in message 23 and 232

(rillig)

lint: support __attribute__((hot))

The corresponding attribute 'cold' was already added in cgram.y 1.84
from 2016-12-29.

(rillig)

lint: fix shift/reduce conflict for dangling else

The following line no longer occurs in the yacc output:
257: shift/reduce conflict (shift 427, reduce 270) on T_ELSE

No functional change.

(rillig)

tests/lint: test dangling else

(rillig)

lint: resolve shift/reduce conflict in notype_direct_decl

When a notype_direct_decl was followed by a type_attribute_list, and the
next token was another type_attribute, the parser could either continue
the current type_attribute_list or start a new one.  Either way has the
same effect since type_attribute_list has no associated action.

This reduces the conflicts by 4, one for each of T_ALIGNAS, T_ATTRIBUTE,
T_NORETURN, T_PACKED.  There are several other conflicts involving these
4 tokens, but they are harder to fix.

No functional change.

(rillig)

lint: use separate tokens for logical not and bitwise complement

The token T_UNARY was misleading since it only captured 2 of the 6
operators that C99 calls unary-operator.  Make the grammar easier to
understand by explicitly listing these 2 operators.

No functional change.

(rillig)

lint: remove redundant %prec from the grammar

Now that the precedence of the unary operators has been modeled via
separate rules, the precedence is no longer needed.

No change to the generated parser, with both yacc or Bison.

(rillig)

lint: remove shift/reduce conflict for sizeof

According to the grammar, the expression 'sizeof(int)x' was ambiguous.
Since 'sizeof(int)' is a valid term, that could have been reduced,
which would have resulted in a parse error when trying to parse 'x'.

Now 'sizeof' takes a unary_expression instead of a term.

No functional change.

(rillig)

lint: resolve shift/reduce conflicts for unary expressions

The grammar rule 'term' was ambiguous since both the prefix and postfix
increment operators were listed with the same precedence.  The
expression '++x++' was parsed as '++ (x++)', as expected, since
conflicts resolve towards shift.

Resolve these conflicts by structuring the grammar as in C99, with the
GCC extension of statement-expressions.  The resolved conflicts are:

134: shift/reduce conflict (shift 161, reduce 347) on T_LBRACK
134: shift/reduce conflict (shift 162, reduce 347) on T_LPAREN
134: shift/reduce conflict (shift 163, reduce 347) on T_POINT
134: shift/reduce conflict (shift 164, reduce 347) on T_ARROW
134: shift/reduce conflict (shift 165, reduce 347) on T_INCDEC
state 134
term : term . T_INCDEC  (335)
term : term . T_LBRACK expr T_RBRACK  (341)
term : term . T_LPAREN T_RPAREN  (342)
term : term . T_LPAREN argument_expression_list T_RPAREN  (343)
term : term . point_or_arrow T_NAME  (344)
term : T_EXTENSION term .  (347)

No functional change.

(rillig)

Make sure dd_size is init'd, even when nothing has been read (so it
will be 0).  Reported on tech-userlevel by Mouse 20210711T044753+0000

(kre)

lint: fix bug when parsing unused variable (since 2021-07-10)

Partially revert to cgram.y 1.248 from 2021-06-29.

This fixes the parse error for variables whose declaration starts with
__attribute__((unused)).  In the many refactorings of the last days this
bug has slipped in, and since there were several refactorings in that
area, there may be have been further bugs that are not caught by the
current test suite.  Revert for now and maybe apply them later again
when there are more tests.

Things kept from the current version are:

The names of most of the rules, as they correspond more closely to C99
and do not affect the behavior in any way.

In type_direct_decl, the replacement of type_attribute_list with
type_attribute since that nonterminal is already part of a repetition
(saves 4 conflicts).

In block_item, the order of the rules corresponds to C99.  This has no
influence on the generated parser, except for the rule numbers, which
are informative.

The merge of the duplicate code for struct_tag, enum_tag and
enum_constant, as they all contained exactly the same code.

(rillig)

tests/lint: when accepting test results, skip crashes

(rillig)

Mention that emulated disks by HPDisk and HPDrive work.

Also add more HP-IB disk models per recent changes.

(tsutsui)

HP9000/360 can have up to 16 MB RAM, not 48 MB.

Per Service Information Manual HP 9000 Series 300 Computers Models 360/370.

(tsutsui)

Fix silent freeze on probing sti(4) framebuffer on 712/60.  PR/52162

Ok'ed by skrll@.
Should be pulled up to netbsd-9 and netbsd-8.

(tsutsui)

tests/lint: analyze yesterday's bug for parsing declarations

(rillig)

Add Device and drive info of 2202A, 7908A, 7911A, and 7941A.

Geometries and description info are taken from hpdrive.ini.sample
in HPDrive.  Briefly tested on HPDisk.

(tsutsui)

tests/lint: parse error for unused variable (since 2021-07-10)

Since cgram.y 1.294 from 2021-07-10.

(rillig)

tests/lint: suppress ShellCheck warnings for intentional violations

The undeclared variables are generated by the AWK program.
The variable 'flags' must be split into words at exactly this point.

(rillig)

We need IFS to be set to % for the rest of the script so be explicit
about it. Fixes message truncation for non-English in sysinst.

(cjep)

Final note on OpenBSD cross-buildling. Remaining issues were due to
default limits on the build host.

(cjep)

Optimized fast-paths for rw_enter() / rw_tryenter() / rw_exit().

(thorpej)

Pull in pmap_subr.s (missed in prior commit).

(thorpej)

Define ALPHA_PGBYTES (missed in prior commit).

(thorpej)

remove dead mesa7 dist.

(mrg)

remove dead mesa7 support.

(mrg)

also look in dist/../include.

(mrg)

avoid 'visibility' attribute on static functions.

(mrg)

add something

(dholland)

merge xterm 368 and libdrm 2.4.107.

(mrg)

xorg-server is now 1.20.12.

(mrg)

merge xorg-server 1.20.12 and xkeyboard-config 2.33.

(mrg)

bump minor for for libXfixes 6.0.0's new function

(mrg)

merge libX11 1.7.2, and libXfixes 6.0.0.

(mrg)

lint: fix parsing of __attribute__ for member (since 2021-07-10)

Since cgram.y 1.280 from 2021-07-10, lint could not parse struct members
that have multiple __attribute__ in front of their type.

(rillig)

remove an old #if 0'd section.

(mrg)

tests/lint: demonstrate parse error for __attribute__ member

(rillig)

lint: inline grammar rules declmod and qualifier_or_storage_class

The rule declmod had a confusing name since declmods was not exactly a
list of declmod.

Inlining the rules reduces the abstraction level.  There are still some
shift/reduce conflicts in that area, so make the rules as simple as
possible, in order to resolve these conflicts.

No functional change.

(rillig)

lint: merge duplicate code for struct and enum declarations

No functional change.

(rillig)

lint: move type_attribute back into the main section

Type attributes are a standard feature since C11 and thus do not belong
in the GCC-specific section.  Only the GCC __attribute__ belongs there.

No functional change.

(rillig)

lint: merge duplicate code for struct and enum tags

This is a tricky area.  Inlining identifier_sym did not work because of
the side-effects of getsym.  In situations where two identifiers of
different kinds follow each other, such as 'enum id1 id2', the timing is
crucial, and inlining identifier_sym would have led to an internal error
in getsym.  The symbol type has to be reset from FTAG to FVFT before
reading the next token.

No functional change.

(rillig)

Move the optimized pmap_zero_page() from pmap.c to a new pmap_subr.s,
and optimize it a teeny bit little more.  Provide an optimized (for 21164,
anyway) pmap_copy_page() as well.

(thorpej)

tests/lint: test error handling in enum-specifier

(rillig)

lint: rename grammar rules to align with C99

No functional change.

(rillig)

lint: merge duplicate code for typespec and notype_typespec

No functional change.

(rillig)

tests/lint: test lexing of integer suffixes

(rillig)

tests/lint: add another example for storage class in declaration

(rillig)

tests/lint: test declarations

(rillig)

lint: rename func_arg_list to align with C99

No functional change.

(rillig)

lint: rename y_sb to y_name

That buffer is only used for names of identifiers, so be more specific.

No functional change.

(rillig)

lint: rename clrtyp/deftyp to begin_type/end_type

The abbreviations clr/def did not make it obvious that these two
functions or grammar rules form pairs.

No functional change.

(rillig)

Add v_spec_ctrl, avx512_fp16, sme, sev and sev_es. Tested by nonaka@.

(msaitoh)

0x6a and 0x6c are 3rd gen Xeon Scalable (Ice Lake).

(msaitoh)

lint: extract clrtyp from noclass_declspecs

Finally each clrtyp pairs up with its corresponding deftyp, which makes
the grammar much easier to understand.  The many grammar rules that
contain these actions now all look the same.

No functional change.

(rillig)

Add some definitions from Intel SDM:

- CPUID leaf 7:0 %ecx bit 13 TME_EN (Total Memory Encryption)
- CPUID leaf 7:0 %edx bit 18 PCONFIG (Platform CONFIGuration)

(msaitoh)

lint: extract clrtyp from declaration_specifiers

This makes the grammar more uniform and thereby easier to understand.
The actions clrtyp and deftyp now come in pairs almost everywhere.

No functional change.

(rillig)

lint: split grammar rule clrtyp_declmods

By extracting clrtyp from that rule, clrtyp and deftyp pair up nicely in
most of the grammar rules.  These two actions delimit the common part of
a declaration that declares multiple identifiers.

No functional change.

(rillig)

lint: inline grammar rule clrtyp_typespec

No functional change.

(rillig)

lint: merge duplicate code from grammar rule clrtyp_typespec

No functional change.

(rillig)

lint: swap order in grammar rule clrtyp_typespec

It should not make a difference whether the rule says 'T_TYPENAME
clrtyp' or 'clrtyp T_TYPENAME'.  The latter order is more consistent
with the name of the rule, though.

According to the code, these two tokens take completely distinct
actions.  The code coverage outside of cgram.y and cgram.c stays exactly
the same, at least for lint's test suite.

No functional change intended.

(rillig)

lint: rename grammar rule declmods to clrtyp_declmods

It wasn't obvious from the name 'declmods' that this rule starts a new
type declaration.

No functional change.

(rillig)

Use intmax_t to printf an off_t

(martin)

lint: flush stdout before an internal error

Previously, the internal error may have appeared too early in the
output.  Provoking internal errors and assertion failures is still easy,
just run afl-fuzz for a few seconds.

(rillig)

lint: fix assertion failure in declare_argument

(rillig)

lint: do not allow struct{const;}

In traditional C, a struct member was defined syntactically as
'type-specifier struct-declarator-list', the concept of a type-qualifier
was not known back then.

C90 invented the type-qualifier 'const' and relaxed the syntactic
requirement for struct member declarations by allowing 'const x'. Having
only a type-qualifier without an actual type may be regarded as an
"incomplete type", which would be forbidden by C90 and later.

Anyway, this doesn't occur in practice anyway, so there is no need for
lint to try to parse it.  This removes a bit of dead code, since a
type-qualifier-list can never have type struct or union.

(rillig)

lint: remove noclass_declmods from grammar

That rule was hard to understand since it contained clrtyp but not the
corresponding deftyp.  It's easier to read when a grammar rule contains
both clrtyp and deftyp in the same line.

C99 does not mention the term 'decl-modifier' anywhere, and in fact the
only thing that this grammar rule allowed was a type-qualifier-list.
The new name better reflect this.

No functional change.

(rillig)

lint: fix assertion failure for malformed member declaration

(rillig)

lint: merge duplicate code in grammar for 'for' loops

No functional change.

(rillig)

lint: clean up grammar for do-while statements

No functional change.

(rillig)

lint: merge duplicate code in grammar for 'while' statements

No functional change.

(rillig)

tests/lint: add more tests for covering the grammar

(rillig)

tests/lint: when adding a new test, set an invalid option

This forces the test author to think about whether any custom options
are really needed.  While here, mention the other possible
customizations as well.

(rillig)

tests/lint: add code coverage for grammar rule parameter_declaration

(rillig)

lint: add code coverage for grammar rule direct_notype_param_decl

(rillig)

add more info to ease future debugging.

(christos)

space before star

(christos)

s/ifdef _ARM_ARCH_6/if defined(_ARM_ARCH_6)/ for consistency.  NFCI.

(skrll)

tests/lint: move test for __attribute__ out of msg_124.c

That test case didn't belong there since there was no chance of getting
an 'illegal pointer combination' by applying an operator.

(rillig)