MySQL Bugs: #8599: Compile failure in handler.cc / ha

Bug #8599	Compile failure in handler.cc / ha_federated.h
Submitted:	18 Feb 2005 12:10	Modified:	5 Apr 2005 21:23
Reporter:	Joerg Bruehe	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: Compiling	Severity:	S1 (Critical)
Version:	5.0.3-pre	OS:	HP/UX (HP-UX)
Assigned to:	Bugs System	CPU Architecture:	Any

Description:
Compile problem on HP-UX prevents build there:

Error 221: "ha_federated.h", line 75 # Member 'create_where_from_key' not declared in class ha_federated ["ha_federated.h", line 59].
      bool ha_federated::create_where_from_key(String *to, KEY *key_info, const byte *key
                         ^^^^^^^^^^^^^^^^^^^^^                                           
gmake[4]: *** [handler.o] Error 2

This happens on hp3750, hp3750-64bit, hpita2-64bit, hpux11, hpux11-64bit

Latest changeset:
  1.1852 05/02/17 18:07:14 jimw@mysql.com +5 -0
  Clean up merge from 4.1, including making several enum fields
  in the grant tables case-insensitive and updating test results.

In the 64 bit compiles, these warnings follow (all refer to HA_FILE_BASED):
Warning 863: "handler.cc", line 975 # Result of operator << is widened from int to unsigned long.
      if (lower_case_table_names == 2 && !(file->table_flags() & HA_FILE_BASED))
                                                                  ^^^^^^^^^^^^^  
Warning 863: "handler.cc", line 1571 # Result of operator << is widened from int to unsigned long.
          !(table.file->table_flags() & HA_FILE_BASED))
                                        ^^^^^^^^^^^^^  
Warning 863: "handler.cc", line 1636 # Result of operator << is widened from int to unsigned long.
            !(table.file->table_flags() & HA_FILE_BASED))
                                          ^^^^^^^^^^^^^  

How to repeat:
Build on any of these HP-UX machines.

I'm trying to build on hpux11 and hp3750, using the build options from logfile:
 such as on hp3750:

 CC=cc CXX=aCC CFLAGS="-DBIG_TABLES +DAportable" CXXFLAGS="-DBIG_TABLES -DAPortable" ./configure --prefix=/usr/local/mysql --localstatedir=/usr/local/mysql/data -libexecdir=/usr/local/mysql/bin --with-extra_charsets=complex --with-server-suffix="-max" --enable-thread-safe-client --enable-local-infile --disable-shared --with-berkely-db --with-low-memory --with-readline --with-embedded-server --with-innodb

Then, when I run 'make', I get:

Making all in Docs
                cd . && \
                        /bin/sh /home/mysqldev/patg/mysql-5.0/missing --run
                        makeinfo -I . --no-headers --no-split --output
                        manual.txt
                        makeinfo: missing file argument.
                        Try `makeinfo --help' for more information.
                        *** Error exit code 1

                        Stop.
                        *** Error exit code 1

                        Stop.
                        *** Error exit code 1

                        Stop.

Is this something simple I'm missing? I want to try to see if my code fixes work, but can't get past this in either case. If someone could help me get past this, I can see if my pending patches work. 

Thanks much ;)

Ok, I got past the problems with makeinfo by just nixing 'Docs' from the SUBDIRS list in the main Makefile, and the compile proceeds as far as:

cc -O -DDBUG_OFF -DBIG_TABLES +DAportable -DHPUX11 -DSNPRINTF_RETURN_TRUNC -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -DHAVE_BROKEN_GETPASS -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_BROKEN_INLINE -o mysql_client_test mysql_client_test.o  ../libmysql/.libs/libmysqlclient.a /home/mysqldev/patg/mysql-5.0/zlib/.libs/libz.a -lpthread -lpthread ../mysys/libmysys.a ../zlib/.libs/libz.a -lpthread -lcrypt -lnsl -lm -lpthread
        source='insert_test.c' object='insert_test.o' libtool=no \
        depfile='.deps/insert_test.Po' tmpdepfile='.deps/insert_test.TPo' \
        depmode=hp /bin/sh ../depcomp \
        cc -DUNDEF_THREADS_HACK -I. -I. -I.. -I../include      -O -DDBUG_OFF -DBIG_TABLES +DAportable   -DHPUX11  -DSNPRINTF_RETURN_TRUNC -DHAVE_BROKEN_PREAD -DDONT_USE_FINITE -DHAVE_BROKEN_GETPASS -DNO_FCNTL_NONBLOCK -DDO_NOT_REMOVE_THREAD_WRAPPERS -DHAVE_BROKEN_PTHREAD_COND_TIMEDWAIT -DHAVE_BROKEN_INLINE -c `test -f 'insert_test.c' || echo './'`insert_test.c
cc: "../include/mysql.h", line 502: error 1000: Unexpected symbol: "*".
cc: "../include/mysql.h", line 502: error 1573: Type of "to" is undefined due to an illegal declaration.
*** Error exit code 1

Note: previous to this, there are multitudes of this error:

Warning 740: "item_subselect.h", line 151 # Unsafe cast between pointers/references to incomplete classes: casting
    'Item_cache *' to 'Item *'.
      Item* el(uint i) { return my_reinterpret_cast(Item*)(row[i]); }
                                                           ^^^^^^    
Warning 749: "sql_lex.h", line 393 # The cast from 'st_select_lex_node *' to 'st_select_lex *' is performed as a
    'reinterpret_cast'. This operation is non-portable and potentially unsafe.
        return my_reinterpret_cast(st_select_lex*)(slave);

That was fixed yesterday, before the build.

The problem is that HP compiler does not like

      bool ha_federated::create_where_from_key(String *to,
                                                      KEY *key_info, const byte *key

in the class definition. It should be

      bool create_where_from_key(String *to,
                                                      KEY *key_info, const byte *key

(this happened many times already - pytical copy-paste error :)

The previous error is due to a change I've made to mysql.h that makes cli_fetch_lengths useable in my handler:

unsigned long * STDCALL mysql_fetch_lengths(MYSQL_RES *result);
// this line (this comment is not in the code)
void            STDCALL cli_fetch_lengths(ulong *to, MYSQL_ROW column,
                              unsigned int field_count);
MYSQL_FIELD *   STDCALL mysql_fetch_field(MYSQL_RES *result);

So, I've changed the line to:

 void       STDCALL cli_fetch_lengths(unsigned long *to, MYSQL_ROW column,
                              unsigned int field_count);

And it gets past that point, but then fails at:

/usr/bin/sed \
          -e 's!@''testdir''@!/usr/local/mysql/mysql-test!g' \
          -e 's!@''bindir''@!/usr/local/mysql/bin!g' \
          -e 's!@''scriptdir''@!/usr/local/mysql/bin!g' \
          -e 's!@''prefix''@!/usr/local/mysql!g' \
          -e 's!@''datadir''@!/usr/local/mysql/share!g' \
          -e 's!@''localstatedir''@!/usr/local/mysql/data!g' \
          -e 's!@''libexecdir''@!/usr/local/mysql/bin!g' \
          -e 's!@''PERL''@!/opt/perl/bin/perl!' \
          -e 's!@''VERSION''@!5.0.3-alpha!' \
          -e 's!@''MYSQL_TCP_PORT''@!3306!' \
          -e 's!@''MYSQL_BASE_VERSION''@!5.0!' \
          -e 's!@''MYSQL_UNIX_ADDR''@!/tmp/mysql.sock!' \
          -e 's!@''MYSQL_TCP_PORT''@!3306!' \
          -e 's!@''MYSQL_NO_DASH_VERSION''@!5.0.3!' \
          -e 's!@''MYSQL_SERVER_SUFFIX''@!-max!' \
          install_test_db.sh > install_test_db-t
        /usr/bin/chmod +x install_test_db-t
        /usr/bin/mv install_test_db-t install_test_db
Make: Don't know how to make std_data/client-key.pem.  Stop.
*** Error exit code 1

just typing 'make' again got the code to complete the compile (???), and federated compiles and links with the change to mysql.h and serg's recommended fix to ha_federated.h, but when I run the tests, a warning is given that the tables are defaulting to MyISAM. I need to <try> to compile with --debug and run the tests in debug mode. The original problem for this bug is fixed, but I'm not sure if that means this bug report should be closed. I will consult someone in eng or dev about this.

Fixed - this was due to a test result that had the wrong values.