> > Perhaps a compromise would be to set versioninfo on
> libpq.dll (which
> > we alerady do), and on all the EXEs, and ignore the rest of
> the DLLs.
> > It's not ideal, but it's a great deal better than nothing at all.
>
> If that is an option, why not just put versions into the
> build-time linkable DLLs, which really need a version, and
> leave it out for the rest. Clearly, we cannot put a version
> into every file anyway (headers files, etc.), so "everything
> must have a version" does not hold anyway, unless there is
> some weird rule again that certain things must have one.
As for DLLs, yes, that sounds reasonable. We also need to put it on the
EXEs. That would mean which DLLS?
libpq.dll and pgevent.dll definitly. Any of the ecpg dlls?
If we limit ourselves to these libs, are you fine with keeping the 7.5.x
version numbering there? (As said before, for libpq we don't have much
choice, and pgevent.dll has no other versioning scheme anyway, since
it's brand new and win32 only)
As said, not ideal, but good enough I think. I think the rule generally
is any EXE and DLL. But as long as it's a private DLL that nobody else
ever uses, I think it's reasonable to skip the rule there.
//Magnus
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match
