Forum OpenACS Q&A: Installation of ACS 4.2 and Oracle 9i on Linux

I've been trying to install ACS 4.2 and Oracle 9i on a Red Hat 7.2
box and being a newbie to ACS and Oracle I've finally reached a
problem that I can't find an answer too.

I've managed to install the Kernel and Core modules but when the
install tries to create the site-wide administrator I get an "ORA-
03113: end-of-file on communication channel" error.

Here are is a partial copy of the error and part of the Oracle trace
file.  Any pointers on where to go next would be very much
appreciated.

FIRST PART OF ERROR:

ora8.c:3569:ora_tcl_command: error in `OCIStmtExecute ()': ORA-03113:
end-of-file on communication channel

SQL:
    begin
        :1 := acs.add_user(user_id => :user_id,
             email => :email,
             url => :url,
             first_names => :first_names,
             last_name => :last_name,
             password => :hashed_password,
                    salt => :salt,
                    password_question => :password_question,
                    password_answer => :password_answer,
                    creation_ip => :peeraddr,
                    email_verified_p => :email_verified_p,
                    member_state => :member_state);
        end;

    while executing
"ns_ora exec_plsql_bind nsdb0 {
...etc.

ORA TRACE FILE:

*** 2002-01-08 18:41:52.589
*** SESSION ID:(7.33) 2002-01-08 18:41:52.570
Exception signal: 11 (SIGSEGV), code: 1 (Address not mapped to
object), addr: 0x90, PC: [0x894994b, kkeidc()+299]
*** 2002-01-08 18:41:53.158
ksedmp: internal or fatal error
ORA-07445: exception encountered: core dump [kkeidc()+299] [SIGSEGV]
[Address not mapped to object] [0x90] [] []
Current SQL statement for this session:
select /*+ FIRST_ROWS*/ constraint_id, constraint_name
      from rel_constraints_violated_one
      where rel_id = :b1
        and rownum = 1
----- PL/SQL Call Stack -----
  object      line  object
  handle    number  name
0x569fd9ac        72  package body INTRANET.REL_CONSTRAINT
0x569fd144        9  INTRANET.MEMBERSHIP_RELS_IN_TR
0x56a2a03c        26  package body INTRANET.MEMBERSHIP_REL
0x56bd5d40        35  package body INTRANET.ACS
0x56b0e38c        3  anonymous block
----- Call Stack Trace -----
calling              call    entry                argument values in
hex
location            type    point                (? means dubious
value)
-------------------- -------- -------------------- -------------------
---------
Exception signal: 11 (SIGSEGV), code: 1 (Address not mapped to
object), addr: 0x0, PC: [0x40480760, strcpy()+64]
*** 2002-01-08 18:41:53.607
ksedmp: internal or fatal error
ORA-07445: exception encountered: core dump [strcpy()+64] [SIGSEGV]
[Address not mapped to object] [0x0] [] []
ORA-07445: exception encountered: core dump [kkeidc()+299] [SIGSEGV]
[Address not mapped to object] [0x90] [] []

Thanks,

Rob

So there's no compelling need to use 9i at the moment.  The one problem I know about is that "delete" is now a reserved word and we have a ton of object types that define a procedure with that name, and a ton of references to them.  So we decided to hold off.

However ... I know at least one person, after experimenting with editing the package definitions for the kernel, who managed to get the kernel installed (he didn't have time to edit all the references not to mention all the other packages in the toolkit).

So ... something else is going on.

What's your stacksize parameter set to in your nsd.tcl file?  The driver needs a lot of stack space ...

(hope I don't come across as being too annoying, in my evil past I was a professional compiler writer)

"update" wasn't allowed even in 8i, and aD knew it (there's a comment in one of the packages saying that "'edit' is used as a name because Oracle won't let us use 'update'"

So it's a bit distressing to hear Jon mention that arguments against the use of "delete" and other reserved words that were incorrectly being allowed was ignored.

Oh well...we'll change 'em to "remove" or something.

On a similar topic in the ArsDigita bboard, Hal Heisler suggested: "I changed references in the Makefile for the driver from -lcore8 to -lcore9 etc.. It built fine after that."

(See: http://www.arsdigita.com/bboard/q-and-a-fetch-msg?msg%5fid=000iiI&topic%5fid=21&topic=web%2fdb).

I hope this helps a bit.

set shmsys:shminfo_shmmax=4294967295

set shmsys:shminfo_shmmin=1

set shmsys:shminfo_shmmni=100

set shmsys:shminfo_shmseg=10

set semsys:seminfo_semmni=100 set semsys:seminfo_semmsl=100

set semsys:seminfo_semmns=220 set semsys:seminfo_semopm=100

set semsys:seminfo_semvmx=32767

Folks should take a look at Julian's note, as he points out that he thinks the queries are bogus to begin with.

and similar for the rel_constraints_violated_two view.

This discussion thread has revealed that bad design decisions (using reserved words as identifiers) and sloppy coding (NULL=NULL) may work for a while but they are really disasters waiting to happen!

Keep up the good work!

where rspm.segment_id(+) = constrained_rels.required_rel_segment
and constrained_rels.container_id is null
and rspm.party_id is null;

where rspm.segment_id(+) = constrained_rels.required_rel_segment
and constrained_rels.party_id is null
and rspm.party_id is null;