A registry (.reg) file looks a little removed from a regular text, no doubt a lot to do with us of unicode, but also other strange bytes. Is there a bit of Fortran code which will write a .reg file in the required format?
You can write all ask plain text, binary records have a type deliminator and then comma separated hex values. A simple reg file write could be 10 lines of code. Look at the wiki page.
Ok, I've written a .reg file which writes an identical file to the one that is exported from regedit, with exception of an extra 0D0A at the end of the file. I tried ADVANCE=NO in the format statement, but it still writes these extra 2 bytes.
Any idea how to suppress them?
Both files have CR LF (0D00 0A00) twice at the end in double byte character format (Bigendian) but in the one you write an additional CRLF (0D0A) is output in single byte format at the end. However I an not trusting anything I see in a text editor I would want to read type bytes as a stream byte by byte and check that.
Well I am not sure what is happening here. When I write a text file from Fortran I get an ANSI/UTF-8 file with single byte characters. Your file is double byte encoded and then has a EOF in single byte format. But the Fortran runtime as I see it only supports ANSI for formatted (text) files. Do you gave an example code a few lines showing the open, write and close so we can test/replicate? Someone else may be able to see the problem but I think I would need to plod through step by step through what you are doing.
Here is the code, with the middle portion stripped out, but you can get the idea:
open(unit=2,file='aqreg.reg',status='unknown') write(2,'(a)',advance='no') & char(z'ff')//char(z'fe')//char(z'57')//char(z'00')//char(z'69')//char(z'00')//char(z'6e')//char(z'00')//char(z'64')//char(z'00')//char(z'6f')//char(z'00')//char(z'77')//char(z'00')//char(z'73')//char(z'00')//& char(z'0d')//char(z'00')//char(z'0a')//char(z'00')//char(z'0d')//char(z'00')//char(z'0a')//char(z'00') close(unit=2)
The most recent file is not remotely something that the registry editor would accept for an import.
I cannot reproduce the extra CR-LF using a recent compilers.
(Use of a hardcoded unit number below 10 makes me nervous - prefer a value obtained using the NEWUNIT specifier, but that does not appear to be the issue here.)
The registry editor will import "ansi" encoded files (as would be written by normal/boring Fortran formatted write statements), so none of this should be necessary. When checking the results of an import, be mindful of registry redirection - are you looking in the right place?
(From a Fortran language point of view, the easiest way to write UTF-16LE encoded files is via an appropriate ENCODING specifier in the OPEN statement. But Fortran processor support for different encodings is processor dependent, and the Intel Fortran runtime does not support UTF-16LE encoding.)