| Re: Non-Blocking I/O [message #14291 is a reply to message #14286] |
Sat, 13 February 1999 00:00   |
ashmall
Messages: 14
Registered: October 1998
|
Junior Member |
|
|
In article <36C4195B.A3E6CF3C@Physik.Uni-Marburg.De>, Ruediger Kupper <Ruediger.Kupper@Physik.Uni-Marburg.De> wrote:
> William Thompson wrote in response to ashmall@my-dejanews.com (Justin Ashmall):
>
>>> Just a thought, but would an FSTAT on the unit number give you any
> information
>>> as to whether there was data waiting to be read?
>>
>>> Justin
>>
>> Probably not. I've dealt with situations where we've had to read from a file
>> which was open for write by another process. As I recall, the behavior of
>> FSTAT was somewhat flakey under those conditions.
>
> Exactly. FSTAT -seems- to be just the IDL function that should do the job, but
> unfortunately it gives absolutely no hint in this case.
> FSTAT results do look like
>
I thought as much! I actually posted a message a short while back about some
trouble I was having with FSTAT and open files. I was hoping it might be
peculiar to NT...
Justin
> ** Structure FSTAT, 12 tags, length=36:
> UNIT LONG 100
> NAME STRING '/homes/kupper/IPC/fifo'
> OPEN BYTE 1
> ISATTY BYTE 0
> ISAGUI BYTE 0
> INTERACTIVE BYTE 0
> READ BYTE 1
> WRITE BYTE 0
> TRANSFER_COUNT LONG 1
> CUR_PTR LONG -1
> SIZE LONG 0
> REC_LEN LONG 0
>
> regardless of any waiting or not waiting data.
>
> Good thought Justin, anyway!
>
>> The situation we were dealing with was to read an incoming spacecraft
> telemetry
>> stream. Since there already was a process (written in C) which was archiving
>> the telemetry stream into data files, what we ended up doing was to simply
> read
>> those files while they were still being written. That way, we avoided the
>> whole pipe/fifo business. Sounds like that wouldn't help you, though.
>>
>> Our original scheme was to use a two-way socket connection between IDL and a
> C
>> process which was handling telemetry reception. IDL would send out a request
>> for data to the socket, and the C process would either respond with a
>> packet, or with a "no-data-yet" message. That way, IDL would always read
> back
>> something.
>
> Okay, so there seems to be no way around using some intermediary C-Routines
> which
> handle reception.
> IDL just doesn't support Inter Process Communication...
>
> Thank you both for your help.
>
> Best regards,
> Ruediger.
>
>
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
