groff treats '-' (the character you get when you hit the "minus" key on
the keyboard) as "hyphen", not as "minus". Thus it renders incorrectly
in some locales, and line-wraps strangely.
groff treats the two-character sequence "\-" as "minus", and the
four-character sequence "\(hy" as "hyphen".
Details here: https://lists.debian.org/debian-devel/2003/03/msg01481.html
This commit replaces every instance of "-" in our manpages where the
meaning is "minus" with "\-", so it works right.
This fixes many lintian warnings.
After the removal of heuristic overrun detection, this param's value
was never changed from 0.
As noted clearly in the documentation, this parameter was for debugging
and subject to removal at any time, so we don't need to wait to remove
it.
Signed-off-by: Jeff Epler <jepler@unpythonic.net>
The factoring out of the 7i43/7i90/etc EPP driver code never progressed
beyond a good intention.
Thanks to Andy Pugh for the bug report.
Signed-off-by: Sebastian Kuzminsky <seb@highlab.com>
.. by tuning the dpll so that it can handle the +-500ppm
changes in the linux CLOCK_MONOTONIC rate that can be caused
by software like ntpd and ntpdate.
Signed-off-by: Jeff Epler <jepler@unpythonic.net>
.. by a new read-request function. This is useful on systems
with more than one hm2_eth card, because the transmission of the
second read request can overlap with the first card processing
and responding to its read request.
This also splits the multiple personalities of "queue_read",
so that the length argument to "queue_read" is no longer
overloaded to select whether to queue / actually send / actually
receive the requested reads.
It also fixes the new functions so that they return success /
failure codes correctly; before, some send/recv failures were
logged via LL_PRINT but not by return value.
It also moves rtapi_delay so that the delay is only taken after a
failed recv(). With multiple cards, this typically removes
the delay for all ethernet cards after the first.
In my hal-test harness, these measures together lowered the typical
thread1.time by >40% with two cards.
Signed-off-by: Jeff Epler <jepler@unpythonic.net>
.. including on a single gigabit ethernet interface with a
gigabit switch between the PC and multiple hostmot2 cards.
Signed-off-by: Jeff Epler <jepler@unpythonic.net>
The encoder comp has no parameters, and the min-velocity-estimate pin
is really called min-speed-estimate (because it's direction-insensitive).
Signed-off-by: Sebastian Kuzminsky <seb@highlab.com>
note: default clock-mode (0) behavior is identical to
prior behavior
Credit to: Slavko Kocjancic
Author's info:
15. November 2014 addition by eslavko
streamer.N.cur-depth, streamer.N.empty and streamer.N.underruns are
now (correctly) updated even if streamer.N.enabled is set to false.
in addition there are two new pins to enable clocking streamer.
by default values the streamer behaviour is as before update.
the new pin is
streamer.N.clock bit input
it's clock input with actions defined by clock_mode pin.
streamer.N.clock-mode s32 input
define behaviour of clock pin.
value 0 (default) freerun at every loop
value 1 means clock by falling edge
value 2 means clock by rising edge
value 3 meand clock by any edge
Signed-off-by: Dewey Garrett <dgarrett@panix.com>
