[unixODBC-dev] ODBC driver regression testing

Eric Sharkey sharkey at netrics.com
Fri Feb 25 02:37:07 GMT 2005

> In 
> anycase; I realize auto-tools is the way to go in the end as it is *the* 
> standard for cross-platform code and more to the point - for unixODBC.  

I don't think there really is a standard for cross-platform building.
The auto-tools work great for cross-Unix building, and ok if you
throw in Windows/cygwin and really pretty terribly on OpenVMS/GNV.

When writing my driver, I had to completely toss the build system
that came with the template.  libtool is just a nightmare on VMS.

> For now qmake seems to work well and is easy to work with.

I'm a little biased against it having recently spent an hour
trying to figure out how to fix a broken build of MythTV packages,
only to find out that qmake builds its Makefiles differently if
the source is in a subtree of /usr.  It doesn't seem like something
a sensible system should do.

> The idea is to have test code built as if ODBC v1. Other code (in other 
> dirs) are built for v2 and v3.

Ah.  I understand now.

You can't do this, though:

  ret = SQLSetEnvAttr(henv, SQL_ATTR_ODBC_VERSION, 1, 0);

SQL_ATTR_ODBC_VERSION is an ODBC 3.0 concept.

Does anyone still use v1?


More information about the unixODBC-dev mailing list