Thank you for your comments. I'll review the patch carefully, but I'm not
sure any patch is required, except to my mental model. I learned a lot
trying to build psqlodbc from CVS, and your message continues that
education. Further responses inline.
At 08:25 AM 2/3/2004 +0000, Dave Page wrote:
>The changes to AC_INIT and AM_INIT_AUTOMAKE change from "new style" to
>"old style" and are therefore definitely wrong.
My research indicated the opposite. Thanks for the correction.
>The PGAC_ARG_* macros provide a more robust argument checking than the
>standard ones, so I would keep them. If there is a problem that you
>need to get these macros from the PostgreSQL source tree (only for CVS
>developers, not "to build"), then just copy over the required m4 files
>and put them in your CVS.
If this is the only reason for the PostgreSQL macros, I would question
whether it's worth the dependency. I did not recognize them, so I thought
they were, perhaps, for an older version of autoconf.
>The changes in AM_CONDITIONAL are then pointless if you keep using the
>PGAC_ARG_ macros. Also, "$with_unixodbc-no" is wrong.
I was getting shell errors to the effect of "unary operator
expected." I'll try again with the config file.
>Replacing PGAC_VAR_INT_TIMEZONE by AC_STRUCT_TIMEZONE is wrong, because
>they don't do the same thing at all. Again, just copy the m4 file if
>you must.
Another incorrect assumption, on my part, that this was some outdated construct.