As far as I'm concerned, v2.3 is essentially feature-complete. The only
other thing I think I might do is this Wishlist item:
o Define HAVE_MYSQL_SSL_SET by default on Windows? Make similar
decisions for other autoconf-only macros.
That, and a bit more work on the documentation.
Here's the current ChangeLog:
o Added Query::for_each() and Query::store_if() methods
proposed by Joel Fielder, and added examples for each.
o It's now possible to store BLOB data in an SSQLS. It's not
foolproof, so added a section to the user manual (5.9) to
document the method. Also, changed examples/cgi_jpeg to
use this new mechanism, instead of the ugly "raw row data"
method it used to use.
o Revamped Connection::set_option() handling. These options
used to be queued up, and applied only just before actually
establishing the connection. This made error reporting
less helpful because the diagnostic was separated from
the cause. Plus, the error messages were misleading to
begin with. Now, set_option() takes effect immediately if
the connection is not yet up (excepting one special option
that can actually be set after the connection is up) and
issues better diagnostics when it detects errors.
o Connection::connect() used to set a few options in such a
way that the user couldn't override them. Now it's smart
enough to set them with the desired default values only
when we see that the user hasn't given them other values.
o SQLString can now be initialized from a mysqlpp::null,
giving a "NULL" string. This is useful for template queries.
Patch by Michael Hanselmann.
o resetdb error message about mixing library and header version
numbers is now more explicit.
o Changed BadConversion exception's "what" message text to be
more like the other exceptions. The inconsistency lead
one to incorrectly copy-paste code from another exception
handler, expecting it to behave the same way. Now it does.
o Added Row::raw_size(), as a shortcut for Row::at().size().
o ssqls-pretty now detects when it's being run from within
the MySQL++ distribution tree and gives a different -I
flag to the compiler, so that it picks up the distribution
headers instead of those that may be on the system already.
o Fixed a build bug on older Solaris versions where the
test for the C API library was erroneously failing, stopping
the configuration process.
o Simplified mysql_shutdown() level argument detection.
Already had to do a version number ifdef check for the
Windows case, so there's really no point to doing it with
autoconf on Unixy platforms. Moved version number check
into lib/connection.cpp, and nuked the separate autoconf
and Windows tests.
o Removed dependency of sql_types.h on myset.h and (indirectly)
datetime.h. Now we only define sql_* typedef aliases for
those MySQL++ types if the headers are included before
o Fixed a typo in one of the storein_sequence() template
overloads, which is apparently rarely (or never?) used,
because no one reported the compiler error you'd get if
o Fixed a few more embedded null handling problems.
o ColData used to keep two copies of all data it held.
Now it keeps just one.
o Fixed install.bat script to track the unified Bakefile change
and the lack of separate debug and release builds under MinGW.
o Yet another STLport + Query memory leak fix.
o Squished a warning in newer GCCs having to do with identifier
shadowing. Patch by Jonathan Wakely.
o Fixed a null-termination bug in Query::parse(). If you
somehow constructed a query string without a terminating null
character, then tried to parse it as a template query, it
could walk off the end of the string. Patch by Worster Chen.
o Removed MYSQLPP_EXPORT tag from FieldNames and FieldTypes
class declarations, as this can cause problems in programs
that use vector<string> in VC++. It has to do with multiply
defined templates, since these classes derive from that
template, and VC++ can't resolve the conflict without help.
Since these classes aren't actually used outside the library,
this shouldn't cause a problem. Patch by Nils Woetzel.
o Partial fix to Doxygen PDF build on RHEL4 and 5. Doesn't yet
run automatically, but it can be coaxed on by hand now. Proper
fix is pending a response from the Doxygen folk.
o Lots of documentation improvements.
This is probably the biggest release since v2.0, so the opportunities
for bugs are huge. Now's the time to compile, comment, and complain.
I'll hold off the release if there's a problem, but lacking that, it'll
get packaged soon.
|• v2.3 release imminent||Warren Young||27 Jun|