Custom Query (541 matches)


Show under each result:

Results (1 - 3 of 541)

1 2 3 4 5 6 7 8 9 10 11
Ticket Resolution Summary Owner Reporter
#680 fixed "Crush earthworm" tport_getmsg; keyget not at FIRST_BYTE somebody stefan

from EW forum:

Recently I upgrade my earthworm to EW7.9 on windows 7. Since then earthworm get in crush after some days of regular functioning. Before to crush I have this warning:

ERROR: tport_copyfrom; keyget not at FIRST_BYTE, Region 1000

ERROR: tport_getmsg; keyget not at FIRST_BYTE, Region 1000

ERROR: tport_getmsg; keyget not at FIRST_BYTE, Region 1000

ERROR: tport_putmsg; keyold not at FIRST_BYTE, Region 1000

ERROR: tport_putmsg; keyold not at FIRST_BYTE, Region 1000

ERROR: tport_putmsg; keyold not at FIRST_BYTE, Region 1000

After this, the software freeze and module waveserverv  dies with naqs2ew, pickew, diskmgr, binder, slink2ew, heli_ew.

Somebody have some ideas to solve this problem? (I attached the screen capture of ew).


and a 2nd user:

As Andrea wrote some months ago,

We decide to upgrade our Earthworm System to Version V.9, 64 bits (Stock distribution, not compiled) on Windows 7 (a machine with 8 cores)

We observe a similar behavior.

After some days working well, the system freezes,

In the main window appears the following message:  tport_getmsg; keyget not at FIRST_BYTE, Region 1000

When we click this window, the system closes.

We check the logs and there is no record about this error.

Do you have some idea, about it?

Best Regards.

#629 fixed *nix Makefiles mix up compiler flags with OS flags somebody baker

I test Sun Studio builds on Linux by editing ew_linux.bash to leave off the WARNINGS and define CC and FC using the absolute paths to the Sun Studio cc and f77. The builds fail because the Sun Studio compilers do not recognize several -W options, such as -Wall and -Wunused. For example:

/opt/oracle/solstudio/bin/cc -m32 -g -Dlinux -D__i386 -D_LINUX -D_INTEL -D_USE_SCHED -D_USE_PTHREADS -D_USE_TERMIOS -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -I/home/baker/Software/Earthworm/earthworm-7.9-7077/include -I../include -O2 -Wall -m32 -g -Dlinux -D__i386 -D_LINUX -D_INTEL -D_USE_SCHED -D_USE_PTHREADS -D_USE_TERMIOS -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -I/home/baker/Software/Earthworm/earthworm-7.9-7077/include  -c -o st_addtodelta.o st_addtodelta.c
cc: -W option with unknown program all
make[4]: *** [st_addtodelta.o] Error 1

The faulty -W options are not in GLOBALFLAGS or CFLAGS from ew_linux.bash; they are coming from hard-coded compiler options in Makefiles (cd to the src directory):

$ find . -iname makefile\* -not -name \*.nt -exec grep -H -o -w "[-]W[^l][^ ]*" {} ';'
./display/webswave/makefile.unix:-W -Wall
./seismic_processing/Hypo71PC/makefile.unix:-W -Wall

The Makefiles incorrectly assume that gcc is used for Linux builds. The compiler flags should be defined in ew_linux.bash. Or, the Makefiles should be made more intelligent to figure out what compiler is being used, if the flags are that important.

My solution was to remove any -W flags from all the non-Windows NT Makefiles (cd to the src directory):

$ find . -iname makefile\* -not -name \*.nt -exec grep -q -w "[-]W[^l][^ ]*" {} ';' -exec sed -i -r -e "s/[-]W[^l][^ ]*//g" {} ';'
$ find . -iname makefile\* -not -name \*.nt -exec grep -H -o -w "[-]W[^l][^ ]*" {} ';'

That changed the output for one compile:

/opt/earthworm/earthworm-7.9-7213/src/data_sources/srusb2ew/USBXCH/Lib/SrUsbXch.c:661:41: warning: multi-character character constant
/opt/earthworm/earthworm-7.9-7213/src/data_sources/srusb2ew/USBXCH/Lib/SrUsbXch.c:903:27: warning: multi-character character constant

It turns out this code compares a 4-character binary signature to a multi-character constant:

#if defined( SROS_MSDOS)
#define SRDAT_HDR_ID                    (0x52535253L)    // = "SRSR"
#define SRDAT_HDR_ID                    ((long)('RSRS')) // = "SRSR"

The code is not portable. The C90 ( standard says:

The value of an integer character constant containing more than one character, ... is implementation-defined.

'RSRS' can be four packed 8-bit characters, as the code assumes, or only one 32-bit character. The -Wno-multichar in src/data_sources/srusb2ew/makefile.unix obscures this bug. In addition, the comments in the include file indicate that the code relies on non-portable assumptions about the alignment of fields in a struct.

I recommend removing all the hard-coded -W compiler options from the *nix Makefiles.

#10 fixed .d files for export_scnl somebody stefan
Here's an export_scnl I have used successfully, don't remember if I modified export_scn.d to make it; there isn't one in the distribution as you say.

#  export_scnl configuration file
#  Exports messages from a given list of station/channel/network codes
#  Knows how to decipher TYPE_TRACEBUF2, TYPE_PICK2K, TYPE_CODA2K msgs
# Import/Export are the long-distance message transfer
# modules. They operate as one-on-one pairs. There are two variants of
# export: export_generic, which ships messages with specified logos, and
# export_scn which ships only trace data, but with specified SCN names.
# Import/Export send heartbeats to each other, as well as into their
# local earthworm systems. If the heartbeat from the distant partner is
# not received whithin the expected time period (RcvAliveInt) the link is
# terminated, and an attempt to reconnect is initiated. If things go
# seriously wrong, the heartbeat into the local earthworm system in
# stopped. The expectation is that "restartMe" has been set in the .desc
# file, and we'll be killed and restarted.
# All socket operations are performed with a timeout. This is noramlly
# defaulted, but can be set in this file (SocketTimeout).
# Export maintains a circular FIFO buffer of messages to be shipped. The
# size of this buffer (RingSize) controls the maximum latency of the
# data.
#Configuration notes:
# "restartMe" should be stated in our desc file.
# The period of our local heartbeat (HeartBeatInt) must be safely smaller
# (faster) than our advertised period in our .desc file (tsec:).
# Otherwise we'll get continually restarted for no good reason.
# Note that tsec:0 implies no heartbeats expected, and so we'll never get
# restarted.
# The rate at which we send heartbeats to our distant partner should be
# considerably faster than the rate at which our partner expects them.
# Otherwise, a heartbeat delayed in transmission will cause our partner to
# conclude that the link is broken, and cause them to break the link and
# reinitialize. Which will cause us to do the same.
# For export, the ServerIPAdr is  the address of the port to be used in
# the exporting machine.  This is to specify a network card case the
# exporting machine has several network cards.
# If SocketTimeout is specified, it should be at least as large as the
# expected period of heartbeats from our distant partner.
 MyModuleId     MOD_EXPORT_SCNL      # module id for this prograM
 RingName       WAVE_RING           # transport ring to use for input/output
 HeartBeatInt   30                  # EW internal heartbeat interval (sec)
#                                     Should be >= RcvAliveInt
 LogFile        1                   # If 0, don't write logfile; if 1, do.
                                    # if 2, write to module log but not to
                                    # stderr/stdout
# Logos of messages to export to client systems
#              Installation       Module       Message Type

# ServerIPAdr and ServerPort specify the local IP address and port number
# on which this module will be ready to accept a client connection.
 ServerIPAdr   # Export messages via this ethernet card
 ServerPort     16105          # Well-known port number to export msgs on

 MaxMsgSize       30000          # maximum size (bytes) for input/output msgs
 RingSize      100           # max# messages in outgoing circular msg buffer

 SendAliveText   "alive"       # string sent to client as heartbeat
 SendAliveInt      30          # seconds between alive msgs sent to client.
                               #  0=> no heartbeat
 RcvAliveText    "alive"       # text of client's heartbeat (we get this)
 RcvAliveInt       1500         # seconds between client's heartbeats to us.
                               #  0 => no heartbeat
 MaxLatency         60         # Optional command; Packets with delay time
                               #  (i.e., time between present and packet
                               #  start time) greater than this amount will be
                               #  filtered out.  Time is in minutes.  0 -> time
                               #  is not checked.  System time must be accurate.
                               #  Only works on TYPE_TRACEBUF messages.

# Socket commands
# The commands below are optional.
# SocketTimeout defaults to RcvAliveInt + 3
#  If set to -1, all socket calls will block (no timeout).
#  SocketTimeout has no effect in export, unless it is set to -1,
#  because there is no code in export to handle socket timeouts.
#  If set to -1, the program may run slightly more efficiently because
#  timeout checking code will not execute.
# SocketTimeout 200000  # Timeout length in milliseconds for socket calls

# SocketDebug defaults to 0
 SocketDebug   0       # if 1, socket_ew debug statements are logged
                       # if 0, socket_ew debug is turned off

# List of station/channel/network codes to export.
# Use any combination of Send_scn (to send data as-is)
# and Send_scn_remap (to change the SCN on the fly) commands.
# Use * as a wildcard for any field. A wildcard in the
# "map to" fields of Send_scn_remap means that field will
# not be renamed.
#               send SCN     map to SCN
#Send_scnl_remap        * * * --     * * * 01
#Send_scnl    * * NT --
Send_scnl        NP1 HHE ZZ 01               # send this specific channel
#Send_scn        JPS *   NC               # send all components of JPS NC
#Send_scn_remap  JGR VHZ NC   *   EHZ *   # change component code only
#Send_scn_remap  CAL *   NC   ALM *   *   # send all component of CAL, but
                                         # change the site code to ALM

Philip Crotwell wrote:
> Hi
> There is an export_scnl in earthworm v7.3 bin and the default
> startstop_linux.d lists a export_scnl module, but there is not example
> export_scnl.d param file in the v7.3 params directory and the docs on
> the web don't list it in the export section.
> Is there either an example .d and .desc file for export_scnl? Or are
> there any docs about the difference between export_scn.d and
> export_scnl.d?
> thanks,
> Philip

1 2 3 4 5 6 7 8 9 10 11
Note: See TracQuery for help on using queries.