FreeTDS Development Group

[Mark Schaal <mark AT champ.tstonramp.com>]

> 
> All,
> 
> In attempting to write a binding from a previously written database layer to
> TDS i've run into a few issues.
> 
> The first issue I have is with error reporting, currently error reporting is
> being passed to an "upper layer library" via a calling the global function
> pointers:
> 
> int (*g_tds_msg_handler)();
> int (*g_tds_err_handler)();
> 
> This dosen't lead well to using TDS in a multithreaded/multiple path of
> execution programming environment because all errors must be handled 
> through one
> function.  My solution is to keep the existing global error function 
> pointers and yet
> add a set of pointers to the TDSSOCKET structure to allow custom functions 
> to be
> called per TDS connection.
> 
> By adding the individual function pointers to each socket connection it will
> allow the customized handling of errors much easier then the global function
> pointers.  Does anybody have any thoughts about this?

The message handles has always been one of the confusion parts of freetds
for me.

The ctlib spec has a function ct_diag() that allows message handling on
a per-connection basis, so we would something like your suggestion if
someone needed to implement that.  I'm not sure that you need multiple
function pointers.  ct_diag() is defined to replace both callbacks.

Mind you, no one is promising that freetds is thread-safe to being
with.  I don't have expertise to even make a guess on that.

> 
> Also currently the prototype of tds_free_login() is not included in tds.h, 
> but
> since it is only calling free on the pointer to the login structure 
> shouldn't it
> be implemented as a macro?
> 
> Also couldn't the macros such as tds_is_error, tds_is_message be 
> implemented as
> macros in the form of:
> 
> #define TDS_TYPE_TEST(type, socket)
>       (tds_peek(socket) == type) ? 1 : 0
> 
> #define TDS_IS_ERROR(socket) \
>       TDS_TYPE_TEST(TDS_ERR_TOKEN, socket)
> 
> Implementing these as macros would reduce the function call overhead since 
> they
> are used quite frequently. 
> 
> So in closing, I'm looking for feedback on any of these proposed changes.  
> I'll
> wait a little while for some then start working up a patch against the 
> current
> sources.
> 
> Thanks,
> 
> Rusty


I don't think most of the tds_is_* functions are used very frequently.  


Mark
-- 
Mark J. Schaal                                  Phone:      (909) 620-7724
TST On Ramp                                     Fax:        (909) 620-8174
System Administrator                            E-Mail: mark AT tstonramp.com