FreeTDS Development Group

[christos AT zoulas.com (Christos Zoulas)]

On Dec 4,  4:42pm, Frediano.Ziglio AT vodafone.com ("ZIGLIO, Frediano, VF-IT") 
wrote:
-- Subject: Re: [freetds] cast to void?

| > 
| > Christos Zoulas wrote:
| > > On Dec 3,  7:29pm, daniel.junkmail AT gmail.com ("Daniel Corbe") wrote:
| > > -- Subject: Re: [freetds] cast to void?
| > > 
| > > | Most likely to squash a warning on some random crappy compiler. 
| > > | Believe it or not not all compilers follow the standard 
| > religiously.
| > > 
| > > I've yet to see a compiler that gives a warning on that...
| > 
| > I'm sorry about the noise.  
| > 
| > It's my own code and my own friendly GNU C compiler.  I 
| > thought Frediano
| > had changed it.  (Sorry, Freddy!)
| > 
| >     memset((void*)pnullrep->bindval, ' ', bindlen);
| > 
| > bindval is defined as 
| >     
| >     const BYTE * 
| > 
| > and the compiler issues a warning if you pass a const pointer 
| > to memset,
| > as well it should.  This is the one place where the data pointed to by
| > bindval is set to something; everywhere else it's read-only.  
| > 
| > Next time, I think I'll compile first and post second.  
| > 
| 
| Mmm.... why this cast is needed?? If we found you have to allocate and
| initialize bindval we should have a no-const pointer (returned by
| malloc) around. Personally I like the distinction made in C++ that use
| const_cast instead of C style casts... but we are in C.
| 
| I also don't understand why we need to allocate some memory in order to
| return NULLs... dbgetnull is called only by _set_null_value. Perhaps
| _set_null_value should do all the job without allocation stuff?? But...
| bindval should not be always initialized?? to_free is used only to pass
| data from dbgetnull to _set_null_value, perhaps merging dbgetnull and
| _set_null_value we can remove this field.
| I noted also that we defined a new NULLREP structure with a field
| (capability) that we don't use at all.
| 
| Why we need a initialize_default_null_representations if
| default_null_representations can be initialized statically (and be
| const)?
| 
| Is correct dbsetnull for VARYBINBIND/VARYCHARBIND?? Perhaps we should
| copy the entire structure?? I think we don't even support these stranges
| types...
| 
| I have a bit of headache... I can do only suppositions...

I agree with Freddy, we should not be doing const-castaways unless it is
absolutely necessary. It is better to declare the pointer non const. Mind
you, if you turn on -Wcast-qual on gcc it will produce a warning on the
const cast-away casts... Perhaps we should just turn on more compiler
warnings to prevent this in the future.

christos