bump bebox and prep image sizes for GCC 9.

(mrg)

new dtb set needs reflecting in instructions

(nia)

actually, disable tsan on everything but amd64.

it fails to build on mips64 and arm64 for me as well as others..

(mrg)

mark tsan files as !machine=sparc64.

(mrg)

powerpc64, sh3 and vax builds now work.
sh3 and vax now works (workarounds commited.)

(mrg)

make build and work on sh3 and vax.

(mrg)

avoid ICE on sh3.

(mrg)

make build (and run) with GCC 9.

(mrg)

merge generic changes in eg rs6000/freebsd64.h into netbsd64.h

(mrg)

Do not build libtsan on sparc64

(martin)

make everyone use gdb.old

(christos)

merge conflicts

(christos)

make(1): fix coverity annotation for CondParser_String

It had been broken since 2020-09-08, when CondLexer (now CondParser) was
added.

(rillig)

make(1): fix wrong error for undefined variables in lint mode

(rillig)

make(1): add test for wrong error in lint mode for undefined variable

(rillig)

make(1): in lint mode, allow undefined variables in dependency lines

This is needed to get past the first few seconds in a src/build.sh run.

The nest obstacle is src/tools/Makefile.gnuhost:30, where the variable
MODULE is undefined even though that file says in line 3 that MODULE is
expected to be set.  It has been saying this since 2001, but since make
didn't have the corresponding check enabled, this didn't break the
build.

(rillig)

make(1): inline character constants in var.c

This removes a level of indirection, and the macro names were quite
similar in appearance as well.  The macros may have been used by people
whose editors don't recognize C string and character literals when
navigating to the corresponding brace or parenthesis.  These editors had
already been confused before this commit since there are also string
literals with unbalanced parentheses, and there are far fewer
compensating comments such as /*{*/ or /*)*/ in the code.  In fact, the
only such comment that is left over was in for.c.

This way, there is now a single consistent way of writing these
character literals, which is without macros or comments, since that is
the simplest form.

(rillig)

make(1): describe how Parse_Error might print stack traces

In deeply nested include chains and .for loops this may be a useful
feature.  It's a little tricky to generate an intuitive stack trace,
though not impossible.  This explanation also serves as a detailed
documentation about how the .include and .for directives are
implemented.

(rillig)

hppa status

(skrll)

make(1): in debug mode, reset unused stack space to NULL

(rillig)

Switch hppa to gcc 9.

Summary for 883 test programs:
    7487 passed test cases.
    644 failed test cases.
    51 expected failed test cases.
    582 skipped test cases.

(skrll)

make(1): add test for nested .for loops with the same variable name

(rillig)

make(1): in the deptgt test, describe what happens to the targets

(rillig)

make(1): add test for the global variable 'targets' in parse.c

(rillig)

make(1): fix documentation of ParseLinkSrcArgs.specType

That code is never called for .SUFFIXES, but for .END.

(rillig)

make(1): clean up documentation and code of ParseLinkSrc

The previous documentation was focused on the implementation details,
which are already clear from the code.  Add some high-level
documentation to help readers understand how this function fits into the
overall picture.

(rillig)

make(1): add test for combining several special targets

(rillig)

swap insns for aarch64eb

(ryo)

make(1): eliminate the global variable specType

This variable didn't have an intended lifetype of "whole application",
therefore it is better implemented as a local variable that is passed
around the few functions that really need it.  It's the same pattern as
in the ModifyWord functions in var.c, only this time without a typedef.

(rillig)

make(1): extract ParseDependencyTargetWord from ParseDoDependency

(rillig)

make(1): split "general GNode *" in ParseDoDepends

(rillig)

make(1): fix type of parseKeywords.op

(rillig)

make(1): extract some code out of ParseDoDependency

With its almost 600 lines, that function is way too long.

(rillig)

make(1): inline LPAREN in parse.c

It's shorter and more readable, and the other characters don't have
named constants as well.

(rillig)

(Re)Enable -Werror=array-bounds.

(skrll)

make(1): flip conditions in ParseDoDependency

In the usual case where the character is a letter or another ordinary
character, each of the terminal conditions has to be evaluated,
therefore from the compiler's view the order doesn't matter.  For
humans, "a unless b" is easier to grasp and more common than "not b but
a", therefore switch to the common pattern.

(rillig)

-Wno-error=array-bounds (missed from previous commit)

(skrll)

-Wno-error=array-bounds

(skrll)

make(1): split complicated condition in ParseDoDependency

(rillig)

Enable TSan tests for GCC and >32bit address space environments

(kamil)

make(1): split confusing condition in ParseDoDependency

A while loop is easier to understand than a do-while loop.

An if statement of the form if-something-then-do-something is easier to
understand than if-complicated-condition-then-continue-doing-something.

(rillig)

nd: Name l3addr union of llentry and use in-place of nd_addr.

Probably makes more sense and makes nd.h less messy.

(roy)

Consistently, always link sanitizer runtime with libm and libpthread

This reflects the behavior from LLVM and avoids issues with dlsym(3) APIs.

(kamil)

make(1): clean up documentation of ParseDoDependency

The documentation was too low-level in parts, essentially repeating the
code in prose.  Instead, add more examples and high-level context
information since that is something that the code cannot do.

(rillig)

Undo local patch adding GNU-stack

This is unneeded on NetBSD.

(kamil)

Avoid using dummy tsan_rtl_unimpl.cc

Instead of providing broken support, error during build to catch the
problem quickly.

(kamil)

Catch unsupported LONG_JMP_SP_ENV_SLOT

A variation of this patch is pending upstream: https://reviews.llvm.org/D87602

(kamil)

sprinkle LE32TOH to fetch instructions on aarch64eb

(ryo)

PID_MAX is just an initial value (soft maximum). Don't use it for CTASSERT.
defined __HAVE_CPU_MAXPROC to use function cpu_maxproc().

pointed out by mrg@, thanks.

(ryo)

Switch from register_t to unsigned long int

Removes dependency on _KERNTYPES.

(kamil)

Fix typo in usage

(kim)

Trailing whitespace.

(skrll)

make(1): remove obsolete comments from tests

(rillig)

make(1): don't require nested variables to be defined in conditions

This code only applies to lint mode (-dL) for now.  After a test phase
and a thorough check for possible side effects, it will be activated in
normal mode, too.  Having this code in lint mode is required to run
src/build.sh, which relies a lot on using variables with undefined
nested variables in conditions.

In the default mode, any errors about nested undefined variables are not
printed since in Var_Subst, oldVars is true.  Therefore, it is not urgent
to properly handle these nested variables correctly there.

(rillig)

Sync with external/gpl3/gdb/dist/gdb by removing unused files:

- rust-exp.c is generated file.
- x86bsd-nat.[ch] were replaced by x86-bsd-nat.[ch] when 8.0.1 was merged.

(rin)

make(1): add test for undefined expressions in conditions in lint mode

(rillig)

make(1): add tests for numbers in conditions

(rillig)

apply some parens to make macros safe.  fixes GCC 9 sanitizer issues.

(mrg)

wg: Add altq hooks.

While here, remove the IFQ_CLASSIFY bottleneck (takes the ifq lock,
so it would serialize all transmission to all peers on a single wg(4)
interface).

altq can be disabled at compile-time or at run-time; even if included
at comple-time the run-time impact should be negligible if disabled.

(riastradh)

fixed armv7-eb build issue; as not gcc related.

(mrg)

merge conflicts

(christos)

Revert previous to make iscsi kernel headers optional again, as
required by kamil.

Now, sanitizer in GCC9 has been fixed differently for MKISCSI=no.

(rin)

switch m68k and sh3 to binutils 2.34.  testing sun3 in tme
and landisk in gxemul shows them both to work as well as
binutils 2.31.

that's binutils 2.34 for everyone now.

(mrg)

fix earmv7eb builds -- add the dtb set items.

(mrg)

fix syntax errors in previous.

(mrg)

switch riscv and ia64 to GCC 9.
switch riscv to binutils 2.34.

(mrg)

riscv64 vs xorg-server is solved (non-gcc specific issue.)

(mrg)

remove additional -DHAVE_DIX_CONFIG_H.

add ${X11FLAGS.DIX} ${X11INCS.DIX} to record.  fixes build on riscv64.

(mrg)

Make Arm MD ACPI code big endian friendly.

(jmcneill)

note some platforms switched and some are ready to switch.

(mrg)

regen for riscv32/riscv64 target emulation list.

(mrg)

merge riscv*-netbsd* target support.

(mrg)

make(1): rename local variables in PrintLocation

For PrintLocation, there is no "current" filename or line number,
therefore the "c" in the variable names was confusing.

(rillig)

make(1): in Var_Parse, replace bmake_strldup with bmake_strsedup

(rillig)

make(1): add test for dynamic variable with modifiers in Var_Parse

(rillig)

switch vax and ia64 to binutils 2.34.  reverse the list and leave
the remaining m68k (untested), riscv (riscv64 ld is missing emulations
for 32 bit targets), and sh3 (untested).

(mrg)

make(1): fix documentation of VarParseErrors

It does not matter whether lint mode is enabled or not, Var_Parse must
reliably return whether an error message has been printed or not.

In the current phase where proper error handling is implemented only in
lint mode, this leads to code that looks a bit bloated since a few extra
branches are added, but that's ok.  Eventually the SILENT constants will
all be removed, and then the number of different cases will shrink
again.

(rillig)

make(1): in lint mode, report undefined variables in conditions

(rillig)

make(1): explain why the dep-var test behaves differently in lint mode

(rillig)

make(1): suppress wrong "Malformed conditional" for undefined variables

This only has an effect in lint mode right now.

(rillig)

make(1): in lint mode, improve error handling for undefined variables

It's a first step for improving the error message that make prints.

(rillig)

make(1): make documentation of VarParseErrors more precise

(rillig)

make(1): prepare Var_Parse for proper error handling and reporting

Right now, Var_Parse swallows many errors during parsing and evaluation.
Ideally, these errors should propagate from the deeply nested
expressions where they occur up to the top-level expressions.  When such
an error occurs, the depending expressions should not be evaluated any
further.  They may still be parsed, but side effects should be
minimized.

The goal is to prevent incomplete expressions like the "xy}" in
moderrs.exp:106 from being evaluated and eventually passed to the shell
for execution.  This expression is a left-over from a parse error in the
mod-t-parse target in moderrs.mk:154.

This commit is a first step in analyzing and verifying the current state
of affairs. The modelling in VarParseErrors already looks complicated
but is expected to closely match reality.

(rillig)

wg: Fix detach logic.

Not tested but this should be less of a rake to step on if anyone
made an unloadable wg module.

(riastradh)

wg: Use RUN_ONCE to defer workqueue_create until after configure.

Should really fix workqueue(9) so workqueue_create can be done before
CPUs have been detected in configure, but this will serve as a stop-
gap measure.

(riastradh)

wg: Add missing kpreempt_disable/enable around pktq_enqueue.

(riastradh)

make(1): fix tests for .INCLUDES and .LIBS

Found by an early draft of a refactoring of Var_Parse to properly report
errors.

(rillig)

make(1): shorten debug output of ApplyModifiers

Having the words "eflags" and "vflags" in the debug output was too
repetitive.  That they are flags is made obvious by the '|' separator,
and the flags have clearly distinguishable names (VARE_* vs. VAR_*),
which lowers the chance for confusion.

(rillig)

Ticket #1084

(martin)

Pull up following revision(s) (requested by tsutsui in ticket #1084):

crypto/external/bsd/openssl/dist/crypto/engine/eng_devcrypto.c: revision 1.8

Restore a local change for PR/54740 lost during openssl 1.1.1e merge.
syslogd(8) complains "Could not open /dev/crypto: Device not configured"
again when pseudo-device crypto(4) is not configured in a kernel.

http://cvsweb.netbsd.org/bsdweb.cgi/src/crypto/external/bsd/openssl/dist/crypto/engine/eng_devcrypto.c#rev1.5
http://cvsweb.netbsd.org/bsdweb.cgi/src/crypto/external/bsd/openssl/dist/crypto/engine/eng_devcrypto.c#rev1.7

Should be pullued up to netbsd-9.

(martin)

make(1): fix position of MAKE_ATTR_UNUSED in inline functions

The attribute needs to be before the return type, otherwise GCC 5
complains that Hash_GetValue is defined but not used, when compiling
with USER_CPPFLAGS=-Dinline=.  The other functions don't get any
warnings.  It's probably because Hash_GetValue is the only inline
function that returns a pointer.

(rillig)

make(1): clean up RCSID blocks

These blocks mostly consisted of redundant structure, following the same
#ifndef pattern over and over, with only minimal variation.

It's easier to maintain if the common structure is only written once and
encapsulated in a macro.

To avoid "defined but unused" warnings from GCC in the case where
MAKE_NATIVE is not defined, I had to add volatile.  Adding
MAKE_ATTR_UNUSED alone would not preserve the rcsid variable in the
resulting binary.

(rillig)

arp test: Use the ndp cache expiration test in place of the old one

As the logic is the same.
While here, GC some variables and comment out a redundant sleep.

(roy)

make(1): fix parameter type of CondEvalExpression

(rillig)

make(1): clean up API for evaluating conditions

There was no need to make struct If publicly visible.

There was no need to have parameters in the public API that were passed
the same constants all the time.

The former function names had not been distinctive.

(rillig)

Elf64_Phdr::p_offset is 64 bits, not 32

(jmcneill)

make(1): clean up comments for ParseErrorInternal and ParseMessage

(rillig)

make(1): fix comment for ParseVErrorInternal

(rillig)

make(1): reduce complexity of PrintLocation

Analyzing a printf statement that is split into several small pieces is
harder than necessary in this case.  Joing the printf statements into
simple strings.  While here, remove the needless casts to int and add a
few empty lines for visual guidance.

(rillig)

make(1): add recently added tests to distrib file list

(rillig)

Tickets #1077 - #1083

(martin)

Pull up following revision(s) (requested by riastradh in ticket #1083):

sys/miscfs/genfs/genfs_rename.c: revision 1.5
tests/fs/vfs/t_renamerace.c: revision 1.37
tests/fs/vfs/t_renamerace.c: revision 1.38

tests/fs/vfs/t_renamerace: Test a screw case hannken@ found.

genfs_rename: Fix deadlocks in cross-directory cyclic rename.

Reproducer:
A: for (;;) { mkdir("c", 0600); mkdir("c/d", 0600); mkdir("c/d/e", 0600);
    rmdir("c/d/e"); rmdir("c/d"); }
B: for (;;) { mkdir("c", 0600); mkdir("c/d", 0600); mkdir("c/d/e", 0600);
    rename("c", "c/d/e"); }
C: for (;;) { mkdir("c", 0600); mkdir("c/d", 0600); mkdir("c/d/e", 0600);
    rename("c/d/e", "c"); }

Deadlock:
- A holds c and wants to lock d; and either
- B holds . and d and wants to lock c, or
- C holds . and d and wants to lock c.

The problem with these is that genfs_rename_enter_separate in B or C
tried lock order .->d->c->e (in A/B, fdvp->tdvp->fvp->tvp; in A/C,
tdvp->fdvp->tvp->fvp) which violates the ancestor->descendant order
.->c->d->e.

The resolution is to change B to do fdvp->fvp->tdvp->tvp and C to do
tdvp->tvp->fdvp->fvp.  But there's an edge case: tvp and fvp might be
the same (hard links), and we can't detect that until after we've
looked them both up -- and in some file systems (I'm looking at you,
ufs), there is no mere lookup operation, only lookup-and-lock, so we
can't even hold the lock on one of tvp or fvp when we look up the
other one if there's a chance they might be the same.

Fortunately the cases
(a) tvp = fvp
(b) tvp or fvp is a directory
are mutually exclusive as long as directories cannot be hard-linked.

In case (a) we can just defer locking {tvp, fvp} until the end, because
it can't possibly have {fdvp or fvp, tdvp or tvp} as descendants.  In
case (b) we can just lock them in the order fdvp->fvp->tdvp->tvp or
tdvp->tvp->fdvp->fvp if the first one of {fvp, tvp} is a directory,
because it can't possibly coincide with the second one of {fvp, tvp}.

With this change, we can now prove that the locking order is consistent
with the ancestor->descendant partial ordering.  Where two nodes are
incommensurate under that partial ordering, they are only ever locked
by rename and there is only ever one rename at a time.

Proof:
- For same-directory renames, genfs_rename_enter_common locks the
  directory first and then the children.  The order
  directory->child[i] is consistent with ancestor->descendant and
  child[0]/child[1] are incommensurate.
- For cross-directory renames:
  . While a rename is in progress and the fs-wide rename lock is held,
    directories can be created or removed but not changed, so the
    outcome of gro_genealogy -- which, given fdvp and tdvp, returns
    the node N relating fdvp/N/.../tdvp or null if there is none --
    can only transition from finding N to not finding N, if one of
    the directories is removed while any of the vnodes are unlocked.
    Merely creating directories cannot change the ancestry of tdvp,
    and concurrent renames are not possible.
    Thus, if a gro_genealogy determined the operation to have the
    form fdvp/N/.../tdvp, then it might cease to have that form, but
    only because tdvp was removed which will harmlessly cause the
    rename to fail later on.  Similarly, if gro_genealogy determined
    the operation _not_ to have the form fdvp/N/.../tdvp then it
    can't begin to have that form until after the rename has
    completed.
    The lock order is,
    => for fdvp/.../tdvp:
      1. lock fdvp
      2. lookup(/lock/unlock) fvp (consistent with fdvp->fvp)
      3. lock fvp if a directory (consistent with fdvp->fvp)
      4. lock tdvp (consistent with fdvp->tdvp and possibly fvp->tdvp)
      5. lookup(/lock/unlock) tvp (consistent with tdvp->tvp)
      6. lock fvp if a nondirectory (fvp->t* or fvp->fdvp is impossible)
      7. lock tvp if not fvp (tvp->f* is impossible unless tvp=fvp)
    => for incommensurate fdvp & tdvp, or for tdvp/.../fdvp:
      1. lock tdvp
      2. lookup(/lock/unlock) tvp (consistent with tdvp->tvp)
      3. lock tvp if a directory (consistent with tdvp->tvp)
      4. lock fdvp (either incommensurate with tdvp and/or tvp, or
          consistent with tdvp(->tvp)->fdvp)
      5. lookup(/lock/unlock) fvp (consistent with fdvp->fvp)
      6. lock tvp if a nondirectory (tvp->f* or tvp->tdvp is impossible)
      7. lock fvp if not tvp (fvp->t* is impossible unless fvp=tvp)

Deadlocks found by hannken@; resolution worked out with dholland@.

XXX I think we could improve concurrency somewhat -- with a likely
big win for applications like tar and rsync that create many files
with temporary names and then rename them to the permanent one in the
same directory -- by making vfs_renamelock a reader/writer lock: any
number of same-directory renames, or exactly one cross-directory
rename, at any one time.

(martin)

Pull up following revision(s) (requested by riastradh in ticket #1082):

share/man/man9/atomic_loadstore.9: revision 1.6
share/man/man9/ipi.9: revision 1.5
lib/libc/atomic/membar_ops.3: revision 1.6

Spell out acronyms in title for clarity.

Update membar_ops(3) man page with examples and relation to C11.

Add exhortation to always always always document how membars come in
pairs for synchronization between two CPUs when you use them.

atomic_load/store_* appeared in NetBSD 9, not 10.

Pullup preceded release of 9.0.

(martin)

Pull up following revision(s) (requested by kardel in ticket #1081):

sys/netinet/tcp_input.c: revision 1.420

PR/kern 55567
fix the data-only fast path. RCV.UP and SND.WL1 could be left behind
on long sequences of data only packets. pull them along to avoid relative
sequence wraps.
consistent with FreeBSD

addresses second failure mode of PR/kern 55567.
pullup to netbsd-8
pullup to netbsd-9

(martin)

Tickets #1603 and #1604

(martin)

Pull up following revision(s) (requested by kardel in ticket #1604):

sys/netinet/tcp_input.c: revision 1.420

PR/kern 55567
fix the data-only fast path. RCV.UP and SND.WL1 could be left behind
on long sequences of data only packets. pull them along to avoid relative
sequence wraps.
consistent with FreeBSD

addresses second failure mode of PR/kern 55567.
pullup to netbsd-8
pullup to netbsd-9

(martin)

Pull up following revision(s) (requested by kim in ticket #1603):

etc/rc.d/motd: revision 1.10
etc/rc.d/motd: revision 1.11
share/man/man5/rc.conf.5: revision 1.186
share/man/man5/rc.conf.5: revision 1.187
etc/defaults/rc.conf: revision 1.159

Add optional release info in /etc/motd

My personal preferencese for /etc/rc.conf:

    update_motd_release=YES
    motd_release_tag='Binaries: '

This provides an explanation to users about the second version in motd.

Document update_motd_release and motd_release_tag

New sentence, new line.

Make a ": " suffix a fixed part of the release info tag

This results in correct updates to /etc/motd even when the value of
motd_release_tag is changed (a likely event).

Add safe quoting to outputting the read kernel version.

Thanks to kre@ for the feedback.

(martin)

Pull up following revision(s) (requested by kim in ticket #1080):

etc/rc.d/motd: revision 1.10
etc/rc.d/motd: revision 1.11
share/man/man5/rc.conf.5: revision 1.186
share/man/man5/rc.conf.5: revision 1.187
etc/defaults/rc.conf: revision 1.159

Add optional release info in /etc/motd

My personal preferencese for /etc/rc.conf:

    update_motd_release=YES
    motd_release_tag='Binaries: '

This provides an explanation to users about the second version in motd.

Document update_motd_release and motd_release_tag

New sentence, new line.

Make a ": " suffix a fixed part of the release info tag

This results in correct updates to /etc/motd even when the value of
motd_release_tag is changed (a likely event).

Add safe quoting to outputting the read kernel version.

Thanks to kre@ for the feedback.

(martin)

Pull up following revision(s) (requested by kim in ticket #1079):

usr.bin/sockstat/sockstat.c: revision 1.22
usr.bin/sockstat/Makefile: revision 1.5
usr.bin/sockstat/Makefile: revision 1.6

RR/54971: scole_mail: sockstat command output incorrect for normal
user As the PR hints, it is not enough to have kern.expose_address=1
to see kernel addresses, we also need to have the PK_KMEM bit set
which we achieve by installing sockstat setgid kmem and opening
and closing /dev/mem. (/usr/src/sys/kern/kern_proc.c:311). It is
unfortunate that we need to give the program more privilege, to
prove to the kernel that we have the privilege to see the data.

turn on fortify since it is setgid (from scole_mail)

(martin)

Pull up following revision(s) (requested by maxv in ticket #1078):

sys/dev/nvmm/x86/nvmm_x86_svm.c: revision 1.73
sys/dev/nvmm/x86/nvmm_x86_vmx.c: revision 1.73
sys/dev/nvmm/x86/nvmm_x86_vmx.c: revision 1.74

nvmm-x86-vmx: improve the handling of CR4
- Filter out certain features we don't want the guest to enable. This is
  for general correctness, and future-proofness.
- Flush the guest TLB when certain flags change.

nvmm-x86: improve the handling of RFLAGS.RF
- When injecting certain exceptions, set RF. For us to have an up-to-date
  view of RFLAGS, we commit the state before the event.
- When advancing RIP, clear RF.

(martin)

Pull up following revision(s) (requested by maxv in ticket #1077):

sys/dev/nvmm/x86/nvmm_x86_vmx.c: revision 1.68
sys/dev/nvmm/x86/nvmm_x86_svm.c: revision 1.74
sys/dev/nvmm/x86/nvmm_x86.c: revision 1.16

Improve emulation of MSR_IA32_ARCH_CAPABILITIES: publish only the *_NO
bits. Initially they were the only ones there, but Intel then added other
bits we aren't interested in, and they must be filtered out.

nvmm-x86-svm: improve the handling of MSR_EFER

Intercept reads of it as well, just to mask EFER_SVME, which the guest
doesn't need to see.

nvmm-x86: improve the CPUID emulation

- Mask DTES64, DS_CPL, CID, SDBG, xTPR, PN.
- B10, B20 and IA64 do not exist, so just remove them.

(martin)

nd needs arp or inet6. inet is not enough.

(roy)

inet: Fix build without ARP

(roy)

Elf64_Phdr::p_flags is 32 bits, not 64

(jmcneill)

make(1): test another attempt at resuming from fatal errors

(rillig)

make(1): inline strchr call in IsInclude

(rillig)

Initialize endc unconditionally, gcc complains and it is not obvious
whether this is a false positive.

(martin)

make(1): clean up the documentation for Parse_File

(rillig)

make(1): fix explanations in test for exporting variables

(rillig)

make(1): fix line number in expected test result for .include

(rillig)

make(1): add tests for the various .include directives

(rillig)

make(1): fix sync-mi helper target, at least for adding tests

(rillig)

Add missing dkbad.h include and define _KERNTYPES so we get all internal
types (like register_t) that we need for ptrace.
Fixes the build on sparc64.

(martin)

make(1): rename ApplyModifier_Exclam

The names of all other ApplyModifier functions already describe the
effect instead of the spelling.

(rillig)

make(1): inline call to strchr in ValidShortVarname

It's a pity that neither GCC 5 nor GCC 10 nor Clang 9 inline this code
themselves, even though it would be easy to do.

Clang 9 at least replaces strchr with memchr, but that is still too
complicated for a simple "is this character one of these" question.

For a repeated "if (varname != ...)" instead of the switch, GCC 10
generates really boring and inefficient code, even though it is easy to
see that the order of the comparisons doesn't matter.

(rillig)

make(1): in lint mode, complain about erroneous $$

Since 2008-12-21, make has silently ignored strange variable names in
constructs like '$$', '$}', '$' followed by nothing.  Ignoring these
bugs in makefiles instead of reporting them is not a good idea.

To improve the situation, make complains about these errors now, but
only in lint mode (-dL).  This preserves existing behavior while still
allowing to validate existing makefiles that they don't depend on this
bug.

If the test phase goes well, these error messages may be enabled
unconditionally.

https://mail-index.netbsd.org/pkgsrc-users/2020/09/12/msg032229.html

(rillig)

Bump date for previous.

(wiz)

Document keyboard mode ioctls and let wsconsctl manage it.

(mlelstv)

make(1): add test for empty variable name in :@var@...@ modifier

(rillig)

make(1): reduce indentation in Compat_RunCommand

The while (1) had been there since the initial import on 1993-03-21, and
in all that time there had never been a good reason for having it.

(rillig)

make(1): remove #endif comments for very short blocks

(rillig)

make(1): remove redundant VARARGS comments

These had been necessary at a time when the functions were declared with
"unknown parameter types".

On 1994-03-05, conditional support for function prototypes was added,
for those compilers that already supported this feature.

On 2002-06-15, the prototypes were made mandatory, and since then the
VARARGS comments had been completely redundant, or worse, simply wrong
(for ParseVErrorInternal).

(rillig)

Allow components to be specified by wedge name.

(mlelstv)

make(1): consense documentation for Fatal

(rillig)

make(1): improve implementation comment in Var_Parse

(rillig)

make(1): fix comments in test for the :M variable modifier

One mistake per sentence is pretty much, I must have been quite
unconcentrated.

The other commits from around that time are fine though.

(rillig)

Support audio descriptor for fstat(1).
sys/dev/audio/*.h: export only what we need for fstat.

(isaki)

Currently restrict GCC TSan files to gcc=9

(kamil)

gcc(1): Install TSan for 64bit CPUs.

(kamil)

Sync netbsd_syscall_hooks.h with LLVM r. 74760bb00fb9b78a

Update the syscall definitionss from a snapshot 2.5 years old to the
current one.

(kamil)

Build and install GCC TSan for 64-bit CPUs

(kamil)

make(1): fix assertion failure in Dir_Destroy in -DCLEANUP mode

When the openDirectories path list is cleaned up, each path from it is
first dequeued and then freed via Dir_Destroy.  At this point, the path
is no longer in openDirectories, which triggered an assertion in
Lst_Remove, called by Dir_Destroy.

(rillig)

Always link TSan runtime with librt and libm

With these changes, TSan/amd64 works.

(kamil)

Disable GetNamedMappingFd for NetBSD

Analogous logic is in LLVM rev. 74760bb00fb9b78a2fe122.

Removes undefined symbol linkage to shm_unlink and shm_open.

(kamil)

make(1): add test for escaping dollars in the :M variable modifier

(rillig)

Ignore unused functions in tsan_interceptors.cc

(kamil)

Sync chunks of tsan_interceptors.cc with LLVM rev. 74760bb00fb9b78a2fe122

Sync tsan_interceptors.cc with upstream file tsan_interceptors_posix.cpp.

Changes:
- define fileno_unlocked() for NetBSD
- use defined __sanitizer_FILE for NetBSD
- handle symbol indirection for vfork and nanosleep
- delete sem_* API, currently moved to common code with sanitizers
- undef SANITIZER_INTERCEPT_PTHREAD_SIGMASK to fix build
- remove locally introduced SANITIZER_NETBSD

With these changes, TSan for GCC builds.

(kamil)

make(1): fix return type of ApplyModifier_Localtime

(rillig)

Remove dead code block (from a local patch)

(kamil)

make(1): move buffer from Var_Parse to ParseVarname

There's no reason to keep the buffer in memory after the variable name
has been parsed.

(rillig)