Good day everybody,
since I'm new to the list, just a quick introduction: My name is Jakob, I
have been developing the OpenWebRX project for about a year now, and have
included a bunch of new features, among them decoding of most WSJT-X modes.
For this purpose, I have been including the jt9 / wsprd command line
decoders, feeding them with wave files, which is working great.
Another feature is the ability to automatically run decoders in the
background, either following the users' frequency choice, or automatically
tuning to selected frequencies when idle. This also works fine and allows
feeding back received data into spotting networks (currently only
pskreporter, but that's a different story).
It has been brought to my attention that certain systems will generate
large amounts of stale files in their temporary directories. (see:
https://github.com/jketterl/openwebrx/issues/130)
Debugging has shown that there are multiple sources of such files, with the
"jt9" command line decoder being one of them, which is why I am writing to
this list today.
On some setups, the jt9 decoder, when called without arguments, will leave
behind exactly one file containing fftw wisdom data in the cwd. The
filename seems to be garbled, randomly composed of many special or
unprintable characters, certainly not human-readable. OpenWebRX calls "jt9"
without any arguments simply to detect if the decoder is installed, and the
corresponding modes should be enabled.
The "wsprd" tool does not seem to be affected by this, but I have found the
same problem to be present in JS8Call, which is not surprising after having
looked at the code. (I am aware that JS8Call is a different project, but
I'm hoping that by addressing this issue here, a potential fix would
propagate there eventually.)
Now, unfortunately, I don't have any experience in Fortran, but I did have
a look at the source (version 2.2.2 for line number reference), and this
seems to be a very simple issue of an uninitialized variable. Looking at
"lib/jt9.f90", I can see that the variable "wisfile" is declared in line 22
and receives its value in line 168. However, the code jumps to "999" when
the usage instructions have been output (no arguments on the commandline),
which is in line 334. The variable is then used in line 343, with the only
assignment having been skipped.
Given the simple nature of the problem, I did give it a try and came up
with a minimal fix for it, see the provided attachment. Once again, I don't
have any Fortran experience, so I guess that'll be about as much as I can
do right now. I have been able to compile and run it, and I can at least
report that it fixes the issue for me.
I'd be very happy if this patch was considered, or if the underlying
problem could be mended in another way.
73s
Jakob DD5JFK
P.S. I previously attempted to send this message to the list, but it got
caught up in moderation because I was using the wrong sender address. I
have canceled the previous message, but if there's still any duplicate
showing up somewhere, I apologize.
diff -ur wsjtx-orig/lib/jt9.f90 wsjtx/lib/jt9.f90
--- wsjtx-orig/lib/jt9.f90 2020-07-21 21:10:43.561474123 +0200
+++ wsjtx/lib/jt9.f90 2020-07-25 20:39:36.413518153 +0200
@@ -77,6 +77,7 @@
data npatience/1/,nthreads/1/
nsubmode = 0
+ wisfile=C_NULL_CHAR
do
call getopt('hs:e:a:b:r:m:p:d:f:w:t:98654qTL:S:H:c:G:x:g:X:', &
@@ -340,7 +341,7 @@
! 58('-')/6i8)
! Save wisdom and free memory
- iret=fftwf_export_wisdom_to_filename(wisfile)
+ if (len(wisfile) > 0) iret=fftwf_export_wisdom_to_filename(wisfile)
call four2a(a,-1,1,1,1)
call filbig(a,-1,1,0.0,0,0,0,0,0) !used for FFT plans
call fftwf_cleanup_threads()
_______________________________________________
wsjt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
