tools: Add test for nonblocking connect(2) performance measurement

test: Sync with FreeBSD.

* Add examples and non-standard primaries to the manual page.
* Reduce code size of ops table.
* Put some static keywords in the source code.
* Accept == as an alias of = wh

test: Sync with FreeBSD.

* Add examples and non-standard primaries to the manual page.
* Reduce code size of ops table.
* Put some static keywords in the source code.
* Accept == as an alias of = which is a popular GNU extension.
* Add more regression tests.

-We have eaccess(2) now, so use it.

Obtained-from: FreeBSD

show more ...

sh: Sync with FreeBSD

* Recognize "--" and explicitly reject options in wait builtin.
* Mention possible ambiguities with $(( and ((.
* Write as much into the heredoc pipe as possible, to avoid fork

sh: Sync with FreeBSD

* Recognize "--" and explicitly reject options in wait builtin.
* Mention possible ambiguities with $(( and ((.
* Write as much into the heredoc pipe as possible, to avoid forking.
* Don't modify exit status when break/continue/return passes !.
* Document a few expansions for the $PS1 and $PS2 environmental
variables.
* Don't consider jobs -s/-p as reporting the status of jobs.
* Improve error handling in read builtin.
* Use O_CLOEXEC and F_DUPFD_CLOEXEC instead of separate fcntl() call.
* Remove linked list of stack marks.
* Allow multiple operands in wait builtin.
* Return status 127 for unknown jobs in wait builtin.
* Do not read from stdin if an error occurs during -i -c cmd.
* Remove mkinit.

Obtained-from: FreeBSD

show more ...

tools: Add bench tools for accept(2) and connect(2)

The main reasons behind making these tools instead of using netperf TCP_CC
test are:
- Netperf TCP_CC creates new server for each testing client
-

tools: Add bench tools for accept(2) and connect(2)

The main reasons behind making these tools instead of using netperf TCP_CC
test are:
- Netperf TCP_CC creates new server for each testing client
- Netperf TCP_CC creates new listen socket on different TCP port for each
testing client
- Netperf TCP_CC created listen socket is in blocking mode
- Netperf TCP_CC closes accepted sockets immediately

All of the above listed items make netperf TCP_CC do not reflect real life
TCP server usage.

Tools in accept_connect:
- "connect_client"
Fork certain amount of processes to perform blocking connect(2)/close(2)
for the certain amount of time.
- "accpet_server"
Fork certain amount of processes to perform blocking accept(2) on the same
addr/port bound socket. close(2) on the accepted socket is called
immediately. If -r is specified, each processes will create their own
listen socket on the same port/addr w/ SO_REUSEPORT sockopt.
- "kq_accept_server"
Fork certain amount of processes to perform non-blocking accept(2) on the
same addr/port bound socket w/ help of kqueue(2). close(2) on the accepted
socket is called after the other side closed the connection. If -r is
specified, each processes will create their own listen socket on the same
port/addr w/ SO_REUSEPORT sockopt.

kq_accept_server closely mimics the real life TCP server.

show more ...

add .Mt to man pages in bin/ games/ sys/ tools/

netrate/netperf: Record min/max results and calculate Jain's fairness index

netrate/netperf: %.2f should be accurate enough

tools/netrate: Add simple tools to calculated multiple netperf results

netperf itself must be installed through dports.

Add FreeBSD's <tgmath.h> tests.

Add FreeBSD's libm tests.

Give /usr/local/... precedence over /usr/pkg/... in various paths.