Backport to 5.5
Current MySQL builds, even on Pushbuild, are not reproducible; they return
different results depending on which directory they are built from (and
Pushbuild uses several different directories). This is because absolute paths
leak into debug information, and even worse, __FILE__. The latter moves code
around enough that we've actually seen sysbench changes on the order of 4% in
some tests.
CMake seemingly insists on using absolute paths, but we can insert our own
layer between CMake and GCC to relativize all paths. Also give the right flags
to get debug information reproducible and turn off build stamping. This makes
the mysqld build 100% bit-for-bit reproducible between runs on my machine,
even when run from different directories.
Problem Statement
=========
Fix various issues when building MySQL with Visual Studio 2015.
Fix:
=======
- Visual Studio 2015 adds support for timespec. Add check and
related code to use this and only use our replacement if
timespec is not defined.
- Rename lfind/lsearch to my* to avoid redefinition problems.
- Set default value for TMPDIR to "" on Windows as P_tmpdir
no longer exists.
- using VS definition of snprintf if available
- tzname are now renamed to _tzname.
Bug#17959689: MAKE GCC AND CLANG GIVE CONSISTENT COMPILATION WARNINGS
Bug#18313717: ENABLE -WERROR IN MAINTANER MODE WHEN COMPILING WITH CLANG
Bug#18510941: REMOVE CMAKE WORKAROUNDS FOR OLDER VERSIONS OF OS X/XCODE
Backport from mysql-5.6 to mysql-5.5
Bug#17894997 CMAKE WARNING WRT INTERFACE_LINK_LIBRARIES
Bug#17905155 CMAKE WARNING WHEN GENERATING MAKEFILE
Bug#71089 CMake warning when generating Makefile
Use old policy for LINK_INTERFACE_LIBRARIES.
Bug#68338 RFE: make tmpdir a build-time configurable option
Post-push fix: 'cmake -LH | grep TMP' showed TMPDIR as a BOOL option,
which was a bit confusing: show it as a PATH instead.
Bug#68338 RFE: make tmpdir a build-time configurable option
Background: Some distributions use tmpfs for mounting /tmp by
default, which has some advantages, but brings also new
issues. Fedora started using tmpfs on /tmp in version 18 for
example. If not configured otherwise in my.cnf, MySQL uses
system's constant P_tmpdir expanded to /tmp on Linux. This can
introduce some problems with limited space in /tmp and also some
data loss in case of replication slave [1].
In case distributions would like to use /var/tmp, which should be
better for MySQL purposes, then we have to patch the source or
change tmpdir option in my.cnf, which is however not updated in
case it has already existed.
Thus, it would be useful to be able to specify default tmpdir
path using a configure option, while using P_tmpdir in case it is
not defined explicitly.
Based on a contribution from Honza Horak
Ignore --with-client-ldflags
it's not supported by the cmake scripts anyways.
Ignore --with-mysqld-ldflags
it's only used with --with-mysqld-ldflags=-static
and that doesn't work.
from the MySQL 5.5 source tree.
Contrary to the comment, the spec file for these ULN RPMs
was missing in the previous changeset (blame ".bzrignore").
This change now brings said spec file, and it adds cmake
handling for the new files.
Still for internal tool tests only, not yet ready for publishing.
This patch inserts an 'ENABLE_GCOV' option for enabling gcov compilation
on Linux machines. It modifies the CMakeLists.txt setting this option
to 'OFF' by default.
Note: The option requires a debug build. For example,
-DCMAKE_BUILD_TYPE:string="Debug"
With this change, there will be new files "INFO_SRC"
and "INFO_BIN", which describe the source and the
binaries.
They will be contained in all packages:
- in "tar.gz" and derived packages, in "docs/",
- in RPMs, in "/usr/share/doc/packages/MySQL-server".
"INFO_SRC" is also part of a source tarball.
It gives the version as exact as possible, preferably
by calling "bzr version-info" on the source tree.
If that is not possible, it just contains the three
level version number.
"INFO_BIN" contains some info when and where the
binaries were built, the options given to the compiler,
and the flags controlling the included features.
The tests (test "mysql" in the main suite) are extended
to verify the existence of both "INFO_SRC" and "INFO_BIN",
as well as some of the expected contents.
option for specific compilers
Reorganize the maintainer mode cmake code to allow options
for specific compilers. For now, enable -Wcheck for ICC,
but do not turn warnings into errors.
options/settings
1. Changed the default value for socket on Windows to the windows
default
2. Removed hard-coded trailing slashes from innodb_data_home_dir
and innodb_log_group_name_dir.
3. Added extra backslashes to the innodb directory example
4. Made the tempdir platform "dependent"
5. Fixed the comments in the .ini files
6. Removed the tmpdir from the templates and the scripts
The reason for the bug is that :
- we use system checks in cmake/os/mysql_release.cmake
- we include cmake/os/mysql_release.cmake using CMAKE_USER_MAKE_RULES_OVERRIDE
- this (having system checks based on TRY_COMPILE inside file pointed by
CMAKE_USER_MAKE_RULES_OVERRIDE does not work with cmake 2.8.3,
and according to Kitware was never meant to work, it just happened to work by accident
until 2.8.2 release (though, it seems not to work wiith 2.6.0 either)
Related CMake bug discussing the situation:
http://public.kitware.com/Bug/view.php?id=11469
The fix is to use INCLUDE instead of CMAKE_USER_MAKE_RULES_OVERRIDE as suggested
by Kitware. The downside is that compile flags are not in cache, but this is pure cosmetics.
The functionality is the same, flags are used for compiling are correct using INCLUDE.
Problem: with "make package" , many small packages are
generated, one per CMake COMPONENT, instead of expected single
package. This is due to the new (in cmake 2.8.3) component-based
install for archive( e.g ZIP,TGZ ) CPack generators.
See http://public.kitware.com/Bug/view.php?id=11452 for discussion.
Fix: use CPACK_MONOLITHIC_INSTALL=1 to enforce single package.
Reset this variable temporarily to 0 for MSI creation (MSI needs
COMPONENTs)
A simple version of the patch : add WITH_PIC to debug compilation.
Engines that are not dependent on internal stuff (like THD structure)
will not be recompiled for embedded in debug mode.
Examples of such engines are : archive, federated, innodb. Engines
with dependencies on internals (myisam,heap,myisammrg) will still be
fully recompiled.
