Am 24.06.2022 um 13:19 schrieb Klaus Hasenbach:
Thank you, Sven!
Actually I ran it on Windows 2022 Server with US-language (points
instead commas).
But I tried it also on my lokal machine with the same US-language
settings and it is just the same.
The error does not occure always.
Please try to run the script twice or three times than it is more
likeli to produce the error.
OK, I got the error now. Here's my preliminary theory, based on the
following error message from gretl together with the contents of my
dotdir (C:/...blabla..../AppData/Roaming/gretl):
File exists
*** error in function dbnomics_bundle_get_data, line 33
outfile --buffer=tracestr
Apparently on Windows the "outfile --buffer" variant is working via
tempfiles. I'm seeing a lot of files in the dotdir on the pattern of
"outfile.3419N1", and indeed most of them contain the text "trying
append..." which comes from the outfile block.
Then gretl complains "file exists", and so I'm guessing that the
mechanism for inventing unique temp file names is flawed and doesn't
really ensure unique names. Hence gretl chokes.
Carrying on with the speculation, why doesn't it happen on Linux? I see
two possibilities: Either it's not using tempfiles at all for the
"--buffer" option, (and is directly working via memory instead), or the
mechanism to produce unique filenames works better there.
The bug will surely be quickly fixed, but the only workaround I can see
right now is to retrieve less variables in one go, and/or manually clean
up the dotdir in between.
cheers
sven