Matt Dargavel wrote:
>
> It looks like ulong is defined in my_global.h, and according to the MySQL
> docs http://dev.mysql.com/doc/refman/4.1/en/windows-client-compiling.html
> you need to include my_global.h before mysql.h on Windows. Unfortunately,
> if you do that you get a lot of errors about types etc already being
> defined. That's about as far as I've got with the issue...
Yes, this is a mess.
Problem 1: my_global.h isn't C++-aware. Aside from the min/max stuff
you've discovered, there are additional crocks in here, like "#define
bool BOOL" in the Windows case. Ptui!
Problem 2: If you pull in my_global.h, you also get my_config.h, which
is an autoconf config.h file, but these aren't supposed to be used
outside the project's build system. Two autoconf config.h files will
conflict, as with MySQL++'s config.h.
Problem 3: Once you work around all that, you find that they've changed
some linkage setting in MySQL 5.0. It feels like an RTL difference to
me....we used to get this a lot back in the 1.7.x days when MySQL++
shipped building against one of the less common RTL versions by default,
so it'd conflict with other stuff.
The solution to the first two is to ignore the MySQL manual. It'll be
simpler to emulate what we need from my_global.h than to work around the
messes it creates. (We're already doing this, such as with the longlong
typedefs.) I've checked in several such changes...the ulong and min/max
problems are fixed now, for whatever good it may do.
Despite knowing what the solution to #3 probably is, I've run out of
willpower. I don't use MySQL++ on Windows myself, and the reasonable
way to attack the problem is to dig into the MySQL build system, which I
really don't want to do.
Here, Matt, have this nice hot potato.
