[Kst] Plot Live data problems on MS-Windows kst2.0-b3

[Peter Kümmel]

I've added the dirfile_maker to the qmake build system and
uploaded a binary which works when you copy it into the bin folder
(it needs the runtime files from mingw)

https://sourceforge.net/projects/kst/files/dirfile_maker_new.exe


I don't see problems when I plot the generated data. But while
porting to Windows I saw no growing of the files when the usleep
is only 20 ms also using fwrite and flushing doesn't help. Only when
I open/close the file for each data point I see the generated files
continuously growing and a smooth dynamic plot.
But with a sleep of 200ms it also works with write.

Maybe your problem is the same: while having a file open kst/getdata
doesn't see any changes until Windows decides to write the data to disk,
but this sounds strange, because reading a opened file should also read
from the buffer. Maybe it's the "file-sytem 'feature'".

A would be interesting the behavior of kst & dirfile_maker is the same
on loacl disk and on a network drive.



On 25.09.2010 23:18, Peter Milne wrote:
> Hi Barth
> 
> OK, will do, but it won't be for a week .. I'll be on the road and not
> carrying ms-w.
> 
> /Peter
> 
> 
> On Sat, Sep 25, 2010 at 8:47 PM, Barth Netterfield
> <netterfield at astro.utoronto.ca> wrote:
>> Peter M,
>> This windows update bug seems very weird, and I worry that it could be some
>> strange file-sytem 'feature'.
>> In kst/tests/dirfile_maker is the source for a little app which makes a fake
>> dirfile that can be used to test real-time bugs.
>> Could you try to get this going under windows, and verify that you see the
>> same bug with this?
>> thanks!
>> cbn
>> On Sat, Sep 25, 2010 at 2:14 PM, Peter Milne <peter.milne at d-tacq.com> wrote:
>>>
>>> Hi Peter
>>>
>>> Just started to use kst2.0-b3, but this problem shows up [same on b1,
>>> sorry to be so slow to report]:
>>>
>>> So, for plotting live data from a network file share, we have the new
>>> 'Time Interval' feature.
>>>
>>> On my system, it "works", but it's far from ideal.
>>>
>>> What appears to happen:
>>> - the data plots.
>>> - the plot updates.
>>> At lease, the left hand side of the time scale [old data] advances.
>>> But the right hand side [new data] either doesn't update or updates
>>> _very_ slowly. With the result that, the timebase window on screen is
>>> continuously shrinking, and after about 5 updates, the timescale
>>> window on screen is very short and the plot is blank.
>>>
>>> A second issue, this is the "front-to-back" line effect raised here:
>>> https://bugs.kde.org/show_bug.cgi?id=249555
>>>
>>> At the time Barth pointed out that the data generator _must_ update
>>> the "reference field" last, and after changing the generator code to
>>> the appropriate sequence, the problem was resolved [at least on KST on
>>> Linux].
>>>
>>> So currently the situation is - plotting live data from a SAMBA share
>>> under KST on Linux, works perfectly, both updates, timescale and
>>> appearance of the plot, but under MS-Windows, it's less than ideal.
>>> It's not the end of the world, because it's possible to force the
>>> screen to update by pressing the Refresh button..
>>>
>>> Best regards
>>>
>>>
>>> Peter
>>>
>>> ps: b3 : the new "Manager", "Wizard" icons are cool, but the
>>> left-right sequence across the screen doesn't really reflect my
>>> workflow -  I'd prefer something like the old order of where "Wizard",
>>> generally the first thing you do, was near the left.
>>>
>>>
>>> --
>>> Peter Milne www.d-tacq.com
>>> _______________________________________________
>>> Kst mailing list
>>> Kst at kde.org
>>> https://mail.kde.org/mailman/listinfo/kst
>>
>>
>> _______________________________________________
>> Kst mailing list
>> Kst at kde.org
>> https://mail.kde.org/mailman/listinfo/kst
>>
>>
> 
> 
>