Michael Widenius wrote:
> The build scripts, especially the debug ones, doesn't use the
> --full-debug option but instead uses instead exact defines to get a
> finer control. As long as this is the case we should NOT remove the
> defines from these scripts.
at least BUILD/compile-pentium-debug-max does ...
> Hartmut> And add some configure options for things like EXTRA_DEBUG instead
> Hartmut> of controlling these via manually adding -D options to CFLAGS to
> Hartmut> controll these only?
>
> May be more work than what's it's worth and will make the build script
> harder to read as you get an extra mapping between the source and the
> script.
the problem i actually ran into when building UDFs/plugins was that
for the things that were set by extending CFLAGS and friends only
it is hard to determine whether the flag was set when the server
was compiled or not, at least for those plugin types that should
only need public header files and not the full server code.
I ran into issues with that with either SAFEMALLOC or SAFE_MUTEX
(or both), so right now i have the following snippet in my code
generator for plugins:
// TODO configure should take care of this
#ifdef DBUG_ON
#define SAFEMALLOC
#define PEDANTIC_SAFEMALLOC
#define SAFE_MUTEX
#endif
simply assuming that all debug servers were built with these
options.
Thinking again just having AC_DEFINE entries for these (and
EXTRA_DEBUG) would be sufficient though so that these settings
could just be taken from the public my_config.h instead of
guessing them.
'Real' configure options would still be nice so that the
existence of these options is documented in "configure --help"
output though ...
--
Hartmut Holzgraefe, MySQL Regional Support Manager, EMEA
Don't Miss the 2009 MySQL User's Conference: http://mysqlconf.com
Sun Microsystems GmbH, Sonnenallee 1, 85551 Kirchheim-Heimstetten
Amtsgericht Muenchen: HRB161028
Geschaeftsfuehrer: Thomas Schroeder, Wolfgang Engels, Dr. Roland Boemer
Vorsitzender des Aufsichtsrates: Martin Haering
