Re: More i18n

[Ludovic Courtès]

Hi,

I agree with your comments and fixes and I'll address them soom.  Just a
few notes...

Neil Jerram <address@hidden> writes:

> "SuS v2" isn't quite a reason for testing, is it?  I think you need a
> few more words here.

I meant that they may not be available everywhere.  I should certainly
make it clearer.

> That's a lot of constant definitions!  Did you consider the possibility of
> any other approach (e.g. symbols)?

I'm no longer sure it's generally good to use symbol to that end.  It
creates secondary name spaces where regular module-based access control
cannot be used.

There's nothing obliging us to use integers for those constants.  For
instance, `DAY_1' could be defined as:

  (define DAY_1 (cons 'a 'b))

Then, if a piece of code doesn't know (or cannot access) the value of
`DAY_1', it cannot forge it and consequently cannot get the effect of
`(language-information DAY_1)'.  This is the same as what happens with
other bindings.

Now, as Kevin said, the names are very C-ish...

>> +  c_result = strdup (c_result);
>> +
>> +  codeset = (char *) alloca (strlen (tmp_codeset) + 1);
>> +  strcpy (codeset, tmp_codeset);
>
> Can we avoid alloca here?  We know from another current thread that
> some systems don't seem to have it.

Sure, but `alloca ()' may be faster on most platforms and it avoids
leaks.  Isn't it just that we should update the `alloca' blob from the
Autoconf manual?

> That's as far as I've got for now.  Does anyone else have a cunning
> way of reviewing a big diff like this?  I'm finding it quite slow.  It
> would be nicer if epatch or something similar worked, but for me it
> doesn't.

Unfortunately, no.

Thanks!

Ludovic.