Re: bk commit into 5.1 tree (kostja:1.2614) BUG#25535

[Konstantin Osipov]

* Sergei Golubchik <serg@stripped> [07/12/11 13:38]:

> > The header files are different, and it uses MY_INIT() rather
> > than my_init().  I seem to be using the same example for the
> > (now-in-progress) fourth edition.  Is the current code okay?
> 
> try to use mysql_library_init() instead of MY_INIT().
> And if load_defaults() will be in mysql.h then you won't need my_sys.h
> and my_global.h

OK, I'd leave it to Paul.

Paul, there are these alternatives now:

 * I do nothing with this bug. Then the example in the third
 * edition of your book stays valid and canonical.

 * I push my patch. The canonical example will look like this:

  #include <my_sys.h>
  #include <mysql.h>
  ... (same as before)

  Rationale: load_defaults is not part of the client
  library/connector but part of mysys (MySQL library for
  system utilities and wrappers, which we happen to ship with the
  connector). In effect this makes "officially" public other mysys
  functions (free_defaults, modify_defaults_file). We do not
  promise ABI compatibility for these, but we say they are there
  and could be used.

 * I follow Serg's suggestion. The example in your book will look
   like 

  #include <mysql.h>

  ... (same as before)

  Rationale: load_defaults is an essential part of the client
  library, that users should know about and be able to use.
  We do not publicise the rest of mysys library.
 
Note, that the current example has a small memory leak: the
memory used for load_defeaults is never freed (for that one has to
use free_defaults). Since it is not a recurring leak it's not a
problem, it will only be visible at exit, when running the program
under valgrind.

Serg, please correct me if I did not summarize the options correctly.

-- 
-- Konstantin Osipov              Software Developer, Moscow, Russia
-- MySQL AB, www.mysql.com   The best DATABASE COMPANY in the GALAXY