[lm-sensors] sensors.conf syntax: anyone use escape sequences?

[Mark M. Hoffman]

Hi Jean:

(resurrecting an old thread here)

* Jean Delvare <khali at linux-fr.org> [2006-11-17 11:26:16 +0100]:
> Hi Mark,
> 
> On Sun, 12 Nov 2006 09:01:42 -0500, Mark M. Hoffman wrote:
> > This is a question about sensors.conf syntax:
> > 
> > For a quoted string, libsensors currently accepts C language escape sequences
> > like '\n' for newline, etc.  It also accepts octal escapes, like '\015', but
> > for some reason it doesn't accept hex escapes, like '\x20'.
> > 
> > Question: is anyone actually using these escapes?
> > 
> > I wouldn't mind getting rid of them.
> 
> I've never seen them used. I agree that octal escapes don't sound very
> useful, assuming libsensors accepts all character values in quoted
> strings, and not only the 32-127 range. Same applies to C language
> escape sequences, except \n, as I guess there is no other way to
> include a newline in a string?

You raise a good point here.  The existing scanner accepts any characters from
the 'alnum' class for unquoted identifiers.  This corresponds to isalnum() of
the C standard library.  That function changes behavior depending on locale.

The end-result is that the behavior of the scanner depends on the locale
setting at _compile-time_ instead of at run-time.  (That's when flex checks
isalnum() to build its static tables.)  That's just nasty.

Right now, the code does accept everything between quotes, which is good.  For
the unquoted behavior, I think it should accept strictly ASCII (32-127) in
order to avoid having the run-time behavior influenced by the compile-time
locale setting.  Any better ideas?

Regards,

-- 
Mark M. Hoffman
mhoffman at lightlink.com