The problem is that the function used by the server to increase
the thread's priority (pthread_setschedparam) has the unintended
side-effect of changing the calling thread scheduling policy,
possibly overwriting a scheduling policy set by a sysadmin.
The solution is to rely on the pthread_setschedprio function, if
available, as it only changes the scheduling priority and does not
change the scheduling policy. This function is usually available on
Solaris and Linux, but it use won't work by default on Linux as the
the default scheduling policy only accepts a static priority 0 -- this
is acceptable for now as priority changing on Linux is broken anyway.
configure.in:
Check for the existence of the pthread_setschedprio function.
include/my_pthread.h:
Use the pthread_setschedprio function to set the thread priority
if the function is available.
The grep expression that finds a running "mysqld" program fails if the
"mysqld_safe" is running with the same PID.
Now, excise "ps" output that has the word " grep" or "mysqld_safe" in
it, to be a little more certain that the matched process is not a false
positive hit. This will fail when the path to mysqld contains either
of those two names, which should be acceptable.
Additionally, some text to search could be truncated if very long.
Expand the number of lines "ps" emits.
Merge it up from 5.0 to 5.1,
adapt to some version differences.
configure.in:
Add an "automake condition" whether InnoDB is configured
so that we can evaluate it for the (non)generation of "innochecksum".
In 5.1, the test command must differ from 5.0.
Mostly, this affected files (programs, scripts, and manual pages)
which got built during a RPM build but were not listed in the
appropriate "%files" section of the "spec" file.
This is fixed now, they are added.
To make this consistent, this patch also makes the build of "innochecksum"
(and its inclusion in a tar.gz or other package) depend on whether InnoDB
is configured in the build.
Also, some tools to create Windows packages are irrelevant in any binary
Unix package (not the sources !), and so they are deleted before packaging.
configure.in:
To prevent "innochecksum" from getting built even if InnoDB is not configured
(and then being included in such packages),
we need an "automake condition" that evaluates whether we have InnoDB.
extra/Makefile.am:
Evaluate the new automake condition about having InnoDB,
and make the build of "innochecksum" depend on it.
man/Makefile.am:
Cleanup: There are manual files which we do not need,
like those about tools for Windows builds / packaging
(irrelevant in any Unix binary package)
and about NDB tools which do not get built.
scripts/make_binary_distribution.sh:
In any Unix binary package, do not include tools for Windows builds
(and their manual pages).
This does not affect source packages, of course.
support-files/mysql.spec.sh:
There were several files (binaries, scripts, and manuals)
which got built with a RPM but never packages,
add them to the respective RPM.
1) Disabled abi_check rule for all compilers except gcc
2) restored the -dI option to retain the header information.
Makefile.am:
1) changed all-local to depend on a configure variable
that will be set based on if the compiler is gcc.
2) restored the -dI option removed earlier since now
only gcc is being used.
configure.in:
Added a check to see if gcc is being used. If it is being
used set the ABI_CHECK variable to the abi_check target.
include/mysql.h.pp:
The .pp file is changed to correspond to the option
-dI.
Another problem is that the backtrace facility wasn't being
enabled for non-Linux targets even if the target OS has the
backtrace functions. Also, the stacktrace functions inside
mysqltest were being used without proper checks for their
presence in the build.
client/mysqltest.c:
Only use stacktrace functions if they are available.
configure.in:
Check if the compiler defines __bss_start
include/my_stacktrace.h:
Enable stacktrace if system has backtrace functions.
mysys/stacktrace.c:
Use backtrace functions if the system supports it.
sql/mysqld.cc:
Only use stacktrace functions if they are available.
The problem was that when a embedded linked version of mysqltest
crashed there was no way to obtain a stack trace if no core file
is available. Another problem is that the embedded version of
libmysql was not behaving (crash) the same as the non-embedded with
respect to sending commands to a explicitly closed connection.
The solution is to generate a mysqltest's stack trace on crash
and to enable "reconnect" if the connection handle was explicitly
closed so the behavior matches the non-embedded one.
client/CMakeLists.txt:
Link mysys to mysqltest.
client/Makefile.am:
Link mysys to mysqltest.
client/mysqltest.c:
Add fatal signal handling with backtracing for Unix and Windows.
configure.in:
Add check for weak symbols support and remove a spurious word.
include/Makefile.am:
Add new header with prototype for stack tracing functions.
include/my_stacktrace.h:
Add new header with prototype for stack tracing functions.
libmysqld/CMakeLists.txt:
stack tracing is now part of mysys.
libmysqld/Makefile.am:
stack tracing is now part of mysys.
libmysqld/lib_sql.cc:
Re-connect if connection was explicitly closed. This is
done to match the behavior of the non-embeded libmysql.
mysql-test/t/sql_low_priority_updates_func.test:
Test expects parallelism between queries that cannot be
guaranteed under embedded.
mysys/CMakeLists.txt:
Add stacktrace to mysys.
mysys/Makefile.am:
Add stacktrace to mysys.
mysys/stacktrace.c:
Move stacktrace to mysys and add weak symbol for the
C++ name de-mangling function so that it can later be
overridden in C++ code. Also add my_ prefix to exported
functions.
sql/CMakeLists.txt:
stacktrace was moved to mysys.
sql/Makefile.am:
stacktrace was moved to mysys.
sql/mysqld.cc:
Add my_ prefix to mysys functions.
added a rule that use gcc to generate preprocessor output (gcc -E)
that can be then compared to a already generated output using
the diff utility.
Ran make test on the repository to verify changes.
Makefile.am:
Added a rule for checking that the abi/api has not changed.
The following steps are followed in the rule in makefile.am
1) Generate preprocessor output for the files that need to
be tested for abi/api changes. use -nostdinc to prevent
generation of preprocessor output for system headers. This
results in messages in stderr saying that these headers
were not found. Redirect the stderr output to /dev/null
to prevent seeing these messages.
2) sed the output to remove blank lines and lines that begin
with # (The header information is retained to enable easy
analysis of abi diffs at a later stage).
3) diff the generated file and the canons (.pp files already in
the repository).
4) delete the .out file that is generated.
If the diff fails, the generated file is not removed. This will
be useful for analysis of ABI differences (e.g. using a visual
diff tool).
A ABI change that causes a build to fail will always be
accompanied by new canons (.out files). The .out files that
are not removed will be replaced as the new .pp files.
e.g. If include/mysql/plugin.h has an ABI change then this
rule would leave a include/mysql/plugin.out file.
A developer with a justified API change will then do a
mv include/mysql/plugin.out include/mysql/plugin.pp to
replace the old canons with the new ones.
configure.in:
1) Removed the part of the file that was icheck related.
2) Added an entry for the configure variable DIFF
include/mysql.h.pp:
The pre-processor output canon file for include/mysql.h
include/mysql/plugin.h.pp:
The pre-processor output canon file for include/mysql/plugin.h
include/mysql_h.ic:
since the icheck target has been removed, this file need
no longer be present in the repository
sql/mysql_priv.h.pp:
The pre-processor output canon file for sql/mysql_priv.h
