In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Brendan Eich) wrote:
> Simon Fraser wrote:
>
> > Another issue: ensure that global static data is const where possible;
> > this allows the linker to place it in a read-only section of the DLL,
> > that has file-mapping advantages (see bug
> > <http://bugzilla.mozilla.org/show_bug.cgi?id=74803>).
> >
> > Example:
> >
> > Bad: static const char* kFoobar = "Foobar";
> > Good: static const char* const kFoobar = "Foobar";
>
> Best:
>
> static const char kFoobar[] = "Foobar";
>
> Otherwise you waste four bytes (on most platforms) on a const pointer
> that serves no purpose, and may confuse sizeof users who wish to compute
> the string length at compile time (see bug
> http://bugzilla.mozilla.org/show_bug.cgi?id=72658).
>
> So I would say Good, above, is Bad, still.
I agree.
> > Bad:
> >
> > static PRUint16 g_ufMappingTable[] = {
> > #include "8859-2.uf"
> > };
> >
> > Good:
> >
> > static PRUint16 const g_ufMappingTable[] = {
> > #include "8859-2.uf"
> > };
> >
>
> But remember that const and * and contravariant (ding!), so it seems
> better style (and more consistent with your earlier style) to put the
> const above _before_ the PRUint16 type, not after (and eerily near the
> g_ufMappingTable declarator). If the table were an array of pointers to
> PRUint16, you'd probably say 'static const PRUint16 *const
> g_ufMappingTable[] = {...}' rather than 'static PRUint16 const * const
> g_ufMappingTable[] = {...}'. The latter would be inferior,
> stylistically, while still legal in C/C++.
Ooops, you're right. I was too hooked on the 'const char* const' thing.
I updated bug 80329 accordinly.
Simon
--
Simon Fraser Entomologist
[EMAIL PROTECTED] http://people.netscape.com/sfraser/
