Re: Tcl9: source files are interpreted as utf-8 by default

saito <saitology9@gmail.com> wrote:

On 1/8/2025 2:32 PM, Rich wrote:

 

I did something like that some 15 years ago. But my case involved a
migration. I had a ton of legacy iso-8859 files on a system-wide
utf-8 Linux system. That caused me problems too, but iconv fixed it.

 
In my case, I used the \uxxxx escapes for anything that was not plain
ASCII, so all my scripts are both "basic 8859" and "utf-8" at the same
time, and having Tcl 9 source them as utf-8 won't cause an issue.  But
it sounds like Uwe directly entered the extended 8859 characters into
the scripts.  Which very well may have made perfect sense if he had
more than one or two of them per script.

 
Interesting thread.
 
Is there a way check a script file for such incompatibilities ahead
of time?

Detecting 'character encodings' reliably is a fickle business.  One
could run a UTF-8 validator (which it appears Ted posted in another
post) that validates that all the bytes in a script are valid UTF-8
encodings.  That implies it is UTF-8, and is likely mostly reliable. 
But I suppose if one wanted to do so one could come up with chimera
file that is valid UTF-8 encoding but also valid for some other
character encoding.  Then "UTF-8" encoding valid may not be the correct
interpretation for the file.

Would this work as a solution?  You build your own Tcl/Tk and add or
duplicate the source command from an earlier version that you are happy
with.  Then you start up your app as you do currenly, and once it is
loaded, you switch the source command to the new version and change the
system encoding.

That would likey work -- but would add the burden of distributing and
maintaining that custom patched version into the future, which Uwe may
not want to take on.