Bug #13621 Does not compile on m68k: /usr/include/asm/atomic.h: 'local_irq_save' was not d

Description:
Hello

Please don't include kernel headers, their maintainers do not care about applications...
We had a similar problem in Juli but this one is very slightly different:

http://buildd.debian.org/fetch.php?&pkg=mysql-dfsg-4.1&ver=4.1.14-5&arch=m68k&stamp=112784...

if g++ -DDEFAULT_BASEDIR=\"/usr\" -DDATADIR="\"/var/lib/mysql\"" -DDEFAULT_CHARSET_HOME="\"/usr\"" -DSHAREDIR="\"/usr/share/mysql\"" -DHAVE_CONFIG_H -I. -I. -I.. -I../include -I../include -I.    -O3 -DDBUG_OFF -DBIG_JOINS=1 -felide-constructors -fno-rtti -O2   -fno-implicit-templates -fno-exceptions -fno-rtti  -MT my_new.o -MD -MP -MF ".deps/my_new.Tpo" -c -o my_new.o my_new.cc; \
then mv -f ".deps/my_new.Tpo" ".deps/my_new.Po"; else rm -f ".deps/my_new.Tpo"; exit 1; fi
/usr/include/asm/atomic.h: In function 'int atomic_add_return(int, atomic_t*)':
/usr/include/asm/atomic.h:91: error: 'local_irq_save' was not declared in this scope
/usr/include/asm/atomic.h:95: error: 'local_irq_restore' was not declared in this scope
/usr/include/asm/atomic.h: In function 'int atomic_sub_return(int, atomic_t*)':
/usr/include/asm/atomic.h:105: error: 'local_irq_save' was not declared in this scope
/usr/include/asm/atomic.h:109: error: 'local_irq_restore' was not declared in this scope
make[3]: *** [my_new.o] Error 1
make[3]: Leaving directory `/build/buildd/mysql-dfsg-4.1-4.1.14/mysys'

How to repeat:
Compile on the m68k architecture with a recent gcc-4.0 suite.
See above URL for the complete build log with configure statement.

Suggested fix:
Don't include any kernel header.

Maybe you find ideas how to fix it in the following bug reports:
http://bugs.mysql.com/bug.php?id=7851
http://bugs.mysql.com/bug.php?id=11845

A possible will be uploaded to this bug report now.

fix

Attachment: 14_configure__gcc-atomic.h.diff (text/plain), 755 bytes.

Christian, could you upload the config.log (and, preferrably, the output of sh -x ./configure <your usual configure options>) ?

configure tries to detect whether atomic.h is usable and should not define HAVE_ATOMIC_ADD otherwise.

Hello

The bug also comes up with MySQL 5.0.15. It seems to be dependend on the GCC version.

The test program from configure.in, I compiled with "gcc -Wall -o t t.c", is:
   #include <asm/atomic.h>
   int main()
   {
          atomic_t v;

          atomic_set(&v, 23);
          atomic_add(5, &v);
          return atomic_read(&v) == 28 ? 0 : -1;
  }

Results:

gcc-2.95: OK
gcc-3.3: 
   /usr/include/asm/atomic.h: In Funktion »atomic_add«:
  /usr/include/asm/atomic.h:21: error: inconsistent operand constraints in an `asm'

The relevant lines of atomic.h are:
  static __inline__ void atomic_add(int i, atomic_t *v)
  {
        __asm__ __volatile__("addl %1,%0" : "=m" (*v) : "id" (i), "0" (*v));
  }

Be aware that the kernel headers are obviously not the same on every architecture and kernel version so yours might look slightly different. This one is from 2.6.13-rc3.

I'm not sure if the primary problem is in MySQL or gcc vs. the kernel.

bye,

-christian-

Forget my last mail, I did the compile tests in the wrong chroot :)
Attached you find config.log and the build log.

bye,

-christian-

config.log

Attachment: config.log.bz2 (application/octet-stream, text), 27.75 KiB.

build output

Attachment: log.bz2 (application/octet-stream, text), 15.70 KiB.

I think the thing to do here is to make sure our tests for the atomic_*() functions are done using C++ by adding the appropriate AC_LANG_{SAVE,CPLUSPLUS,RESTORE} magic.

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/internals/32152

From the error message it does not look like c/c++ is an issue.
Jim are you sure your fix will solve the problem ?
Christian did the fix helped you ?

Is BUG#15244 a duplicate of this one? Also see http://lists.mysql.com/packagers/264 for a related discussion.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

Christian,

Please, check the patch proposed and inform about the results.

Hello

Sorry, I can't verify it currently as I don't have access to a reasonable fast m68k machine
but I will report if the patch does not work after the next 5.0.x release, ok?

bye,

-christian-

OK. I'll set status "Need Feedback" until that.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

-1) early version is affected, too
 0) it is not related to m68k
 1) the patch helps

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/4736

pushed to 5.1.10, 5.0.21, 4.1.19 and 4.0.27

It was a problem with configure and including atomic operations header file (atomic.h) on some linux distributions (like altlinux, and some other)

Noted in 4.0.27, 4.1.19, 5.0.21, 5.1.10 changelogs.

Avoid trying to include <asm/atomic.h> when it doesn't work in C++
code. (Bug #13621)