Re: Proposed patch to clean up signed-ness warnings

> > For me, your patche seems to be a retrogression. In my understanding,
> > the reason why PostgreSQL uses "char *" in many places is just it was
> > designed in the old days when ASCII was the only charset in the world.
>
> Are you proposing that we change all the "char *" to "unsigned char *"?

No, I suggest we change all "char *" to "unsigned char *" only where
it points a string which could hold non ASCII character strings. I
thought we learned the danger of 1) comparing chars with signed bit
on, 2) passing chars with sign bit on to functions which expect int
etc...

> I looked at that briefly but it seems like a huge loss, both in
> notational ugliness and in the amount of code that would have to be
> touched.

If you are just care the amount of effort, why don't you leave as it
is and use pre v4 gcc? :-)

> Also, it would force us to add a bunch of explicit casts to
> avoid warnings with standard library functions like strlen().

Counter examples could be easily found in isalpha(), toupper() etc.

> To me the
> bottom line is that 99% of the code only needs to know that a character
> string is a character string. As this patch demonstrates, there is only
> a tiny fraction that needs to have the "unsigned" declaration. I don't
> think we should allow that fraction to dictate a notational burden for
> all the rest.

To support multiple charsets/collataions, I think we need to change
the way to represent character strings from the unstructured "char *"
to more intelligent structure (I know it's hard to implement that
without significant performance loss, but I know we should do it in
the future).

So "unsigned char*" is not enough for the goal anyway, I'm not against
your patches.
--
SRA OSS, Inc. Japan
Tatsuo Ishii