- Convert to USES=mysql
- Use options helpers
- Regenerate patch with make makepatch
3.2.2, 2015.03.18 (r2780)
o The DBDriver::operator= would leak a live connection if you
assigned a disconnected DBDriver to it. Patch by Quentin Armitage.
o Plugged a potential DBDriver instance leak in Connection ctor.
Patch by Yury Alekseev.
o Explicitly #including algorithm.h in examples/multiquery.cpp for
std::max().
o Added missing "break;" statements in Transaction ctor taking an
isolation level parameter, which caused only the last one to do
the right thing. Problem noticed by Tony Whyman, fix by me.
o The bootstrap script now checks for the existence of needed build
tools before trying to use them. This lets us emit a clearer
error message than the shell will. Original patch by Adrian
Cornish.
o Fixed a build problem related to std::min/max with Visual C++ 2013
due to a change in the definitions provided by its header files.
o Tracked several changes made to Connector/C made since 3.2.1.
o Assorted documentation updates.
o Generating HTML with UTF-8 encoding, not 8859-1
3.2.1, 2013.12.10 (r2756)
o Connection copy ctor wasn't copying the "throw exceptions"
flag. The copy always had exceptions enabled.
o The resetdb example would print a bogus message about logging in
as '1'@'1' instead of showing the actual values passed on the
command line. This did not affect how it actually logged in.
o Previous release tarball was missing a few text files present in
the source repository.
o Updated build paths for VC++ and MinGW to track a file layout
change in Connector/C 6.1.2.
3.2.0, 2013.06.20 (r2747)
o References to MySQL 5.1 have been replaced with 5.6, that
being the current GA release.
o All Windows builds now assume you're using Connector/C 6.1
instead of having installed the MySQL server on your machine
just to get development files. If you are running a MySQL
server on your development machine and don't want to install
Connector/C just to make MySQL++ happy, see the README-*.txt
file for your platform to get instructions on fixing the paths
so the build system finds the development files under the MySQL
server directory.
o The generated VC++ 2008 project files now assume you want 64-bit
executables. (It is 2013 now, after all.)
Since the VC++ 2005 project files continue to be configured for
32-bit builds, the easiest way to get a 32-bit project in VC++
2008+ is to import the VC++ *2005* project files, rather than
adjust the build target settings of all 44 vc2008/*.vcproj
files within the IDE.
o Added Query::replacefrom(): like insertfrom() but it uses
REPLACE statements instead of INSERT. Patch by Adrian Cornish.
o Added support for SQL "unsigned float" columns. Patch by
Leonardo Lang.
o Added "abicheck" top-level build target, which runs the
ISPRAS ABI checker tool (http://goo.gl/e19lD) against
a predecided "ABI-stable" version of MySQL++, comparing it
against the current version. This should help avoid future
unplanned ABI breakages, as happened between 3.0.9 and 3.1.0.
We have chosen not to revert those changes that broke the ABI,
but instead have chosen to accept the 3.1.0 ABI as the new
stable ABI for the rest of the 3.x series. Running the ABI
checker is now part of the release process, so we should not
change the ABI again until 4.0!
o The Query::insert() and replace() overloads that take a pair
of iterators now work with containers that only provide forward
iterators. Idea by Adrian Cornish.
o Using libtool again on Autotools systems. It gives us
relocatable libraries, which is needed in some linking
situations. Patch by Adrian Cornish.
o VC++ release builds now have optimization enabled. This might
also affect other toolchains that create separate Debug and
Release builds, if Bakefile obeys the directive for those
systems. (Such as, perhaps, Xcode.) Initial patch by Matthew
Bradbury.
o Fixed a crash in CommandLine mechanism when the usage message is
called up for programs that have no "extra" parameter and
corresponding usage appendage. Initial patch by "Crazy Pete".
o Query::storein() now tolerates empty result sets, due either to
DBMS failure or passing a query that never returns results,
such as INSERT.
o Exposed DBDriver::result_empty() as Query::result_empty(). We
did this primarily because it was needed as a helper for the
previous fix, but it could also be useful more broadly.
o Added -Wextra to the pedantic build flags.
o Fixed the "escape" manipulator so it will work with Null<>
wrapped types. Patch by "Kemet".
o ssqls2parse now depends on libmysqlpp. (It wasn't getting
relinked when you changed the library.)
o The configure script's test for the existence of mysql_ssl_set()
got broken by a change in the mysql_loc.m4 test, causing it to
always return "no" even if the platform's MySQL C API library did
have the function. Therefore, the SslOption could never be set.
Fixes this bug: stackoverflow.com/questions/7794408
o Fixed a missing a "buried headers" ifdef wrapper for
type_info.cpp. Patch provided by Xavier Richez.
o We now export the DBDriver class from the MySQL++ DLL when
building on Windows. Client programs can now access it directly,
instead of being able to access only the fields and members
exposed through Connection and Query. Fix suggested by Xavier
Richez.
o MinGW builds no longer require that you manually create an import
library for the MySQL C API DLL from a .def file we provide,
which would inevitably get outdated. We can now link directly to
the DLL, and rely on the MinGW linker to figure out the imports.
o Split the -l flags out of LDFLAGS in the userman's example
Makefiles. They're now in LDLIBS, which causes them to be
put after -o $(EXE) by the default GNU make rules, where they
should be.
o Assorted documentation and build system improvements.
o Assorted fixes for newer GCCs.
PR: 211431 [1]
Submitted by: Anon <tfaykpgl@sharklasers.com> [1]
Approved by: maintainer timeout
6d4a383e7f