Mailing List Archive: DCOracle2 on AIX5.2 -- issues

DCOracle2 on AIX5.2 -- issues

Mar 13, 2006, 10:32 AM

Post #1 of 7 (4752 views)

I have Python 2.4.2 installed on AIX 5.2. With some trouble, I have been
able to compile and install dco2.so. The main problem was that the header
files desired by setup.py were not found in oracle_home, but elsewhere in
my Oracle installation -- they appear to be part of Oracle 8, not Oracle 9,
at least here at AEP. oci.h was not found in oracle_9_home/rdbms/public,
but in oracle_8_home/rdbms/demos. nzt.h and ociextp.h were not found in
oracle_9_home/rdbms/public, but in oracle_8_home/rdbms/public. There
exists no oracle_9_home/lib/libclntsh.so here, but instead,
oracle_9_home/lib32/libclntsh.a.

At any rate, the installed "worked," insofar as dco2.so was compiled and
installed in .../site-packages. I have been able to import dco2 without
complaint from python. I have not tried to use it yet.

However, during compilation of cdo2.o, I did encounter these symptoms:

cc_r -DNDEBUG -O -I/oracle/product/9.2.0.6/rdbms/demo
-I/oracle/product/9.2.0.6/network/public -I/oracle/product/9.2.0.6
/plsql/public -I/oracle/product/9.2.0.6/rdbms/public
-I/oracle/product/8.1.7.2/rdbms/demo -I/oracle/product/8.1.7.2/rdbms/public
-I/app/sandbox/s625662/installed/include/python2.4 -c src/dco2.c -o
build/temp.aix-5.2-2.4/src/dco2.o -brtl -bexpall -DORACLE8i
"src/dco2.c", line 5004.38: 1506-280 (W) Function argument assignment
between types "unsigned long*" and "long*" is not allowed.
"src/dco2.c", line 6066.25: 1506-068 (W) Operation between types "short*"
and "unsigned short*" is not allowed.
"src/dco2.c", line 6109.25: 1506-068 (W) Operation between types "short*"
and "unsigned short*" is not allowed.

If anyone could suggest a possible way to correct these problems, I would
appreciate it. Note that the compiler is NOT gcc, but AIX's native xlc,
here called with "cc_r." This was the same compiler that we used to build
Python itself -- without encountering any such complaints.

Also I would appreciate anyone's comment on the significance lack of
correspondence between the file structure expected by setup.py and that in
my Oracle home.

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

dieter at handshake

Mar 14, 2006, 11:54 AM

Post #2 of 7 (4578 views)

Permalink

mfmorss@aep.com wrote at 2006-3-13 13:32 -0500:
> ...
>However, during compilation of cdo2.o, I did encounter these symptoms:
>
>cc_r -DNDEBUG -O -I/oracle/product/9.2.0.6/rdbms/demo
>-I/oracle/product/9.2.0.6/network/public -I/oracle/product/9.2.0.6
>/plsql/public -I/oracle/product/9.2.0.6/rdbms/public
>-I/oracle/product/8.1.7.2/rdbms/demo -I/oracle/product/8.1.7.2/rdbms/public
>-I/app/sandbox/s625662/installed/include/python2.4 -c src/dco2.c -o
>build/temp.aix-5.2-2.4/src/dco2.o -brtl -bexpall -DORACLE8i
>"src/dco2.c", line 5004.38: 1506-280 (W) Function argument assignment
>between types "unsigned long*" and "long*" is not allowed.
>"src/dco2.c", line 6066.25: 1506-068 (W) Operation between types "short*"
>and "unsigned short*" is not allowed.
>"src/dco2.c", line 6109.25: 1506-068 (W) Operation between types "short*"
>and "unsigned short*" is not allowed.

Obviously, your compiler does not like thing (to be precise:
the implicit conversion between "short" and "unsigned short")
that the author did not expect to hurt compilers (and which
most compilers accept).

Your options:

* use a different (less strict) compiler

* fix the code (such that even your strict compiler feels happy).

--
Dieter
_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

matt at bane

Mar 15, 2006, 6:16 AM

Post #3 of 7 (4559 views)

Permalink

It's probably the TRACE macro which is generating a lot of these
messages, if you don't need to be able to do runtime tracing (for
diagnostics) then you can tweak the #define TRACE(condition,args) to
be a no-op macro -- just delete the body of the macro.

On Mar 14, 2006, at 2:54 PM, Dieter Maurer wrote:

> mfmorss@aep.com wrote at 2006-3-13 13:32 -0500:
>> ...
>> However, during compilation of cdo2.o, I did encounter these
>> symptoms:
>>
>> cc_r -DNDEBUG -O -I/oracle/product/9.2.0.6/rdbms/demo
>> -I/oracle/product/9.2.0.6/network/public -I/oracle/product/9.2.0.6
>> /plsql/public -I/oracle/product/9.2.0.6/rdbms/public
>> -I/oracle/product/8.1.7.2/rdbms/demo -I/oracle/product/8.1.7.2/
>> rdbms/public
>> -I/app/sandbox/s625662/installed/include/python2.4 -c src/dco2.c -o
>> build/temp.aix-5.2-2.4/src/dco2.o -brtl -bexpall -DORACLE8i
>> "src/dco2.c", line 5004.38: 1506-280 (W) Function argument assignment
>> between types "unsigned long*" and "long*" is not allowed.
>> "src/dco2.c", line 6066.25: 1506-068 (W) Operation between types
>> "short*"
>> and "unsigned short*" is not allowed.
>> "src/dco2.c", line 6109.25: 1506-068 (W) Operation between types
>> "short*"
>> and "unsigned short*" is not allowed.
>
> Obviously, your compiler does not like thing (to be precise:
> the implicit conversion between "short" and "unsigned short")
> that the author did not expect to hurt compilers (and which
> most compilers accept).
>
> Your options:
>
> * use a different (less strict) compiler
>
> * fix the code (such that even your strict compiler feels happy).
>
> --
> Dieter
> _______________________________________________
> Zope-DB mailing list
> Zope-DB@zope.org
> http://mail.zope.org/mailman/listinfo/zope-db

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

mfmorss at aep

Mar 15, 2006, 6:17 AM

Post #4 of 7 (4551 views)

Permalink

>Obviously, your compiler does not like thing (to be precise:
>the implicit conversion between "short" and "unsigned short")
>that the author did not expect to hurt compilers (and which
>most compilers accept).
>
>Your options:
>
> * use a different (less strict) compiler
>
> * fix the code (such that even your strict compiler feels happy).

Thanks for that observation. However, we are not interested in DCOracle2
for a research project on a $1,000 Linux box, but for critical, daily
industrial production on a $500,000 AIX server. For reliability's sake, we
really have to trust our native AIX compiler (xlc) which, indeed, was able
to compile Python 2.4 itself with no trouble. It would seem that if Python
is able to compile on AIX, DCOracle2 should be also. And gcc would be a
big piece of software to maintain just to get DCOracle2 to compile (and
should we also recompile Python?).

If you fix someone's C code, you really have to wonder if you did it
correctly, and you also have to wonder if the next release will contain
code that is even more difficult to fix. In short, you can hardly trust
software for industrial use if you have to patch it by hand.

I posted originally about these C issues because I assumed it would be
advantageous to Zope if its various parts could be installed whereever
Python could, and that the developers would therefore take some interest in
my troubles. If they don't, we will probably just look for an alternative
Python interface to Oracle. That would be too bad because DCOracle2 does
seem to work in spite of the compiler complaints, and it seems quite nice
to use.

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

jens at dataflake

Mar 15, 2006, 7:02 AM

Post #5 of 7 (4567 views)

Permalink

On 15 Mar 2006, at 14:17, mfmorss@aep.com wrote:
> I posted originally about these C issues because I assumed it would be
> advantageous to Zope if its various parts could be installed whereever
> Python could, and that the developers would therefore take some
> interest in
> my troubles. If they don't, we will probably just look for an
> alternative
> Python interface to Oracle. That would be too bad because
> DCOracle2 does
> seem to work in spite of the compiler complaints, and it seems
> quite nice
> to use.

You need to keep in mind that the developers who worked on DCOracle2
over the years probably have no access to any system running AIX. AIX
is considered too oddballish to worry about by most Zope developers.
The people who do run AIX need to step up to the plate and help out,
there is no other way.

jens

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

mfmorss at aep

Mar 15, 2006, 7:04 AM

Post #6 of 7 (4577 views)

Permalink

OK fine, we may do that.

Mark F. Morss
Principal Analyst, Market Risk
American Electric Power

Jens Vagelpohl
<jens@dataflake.o
Sent by: zope-db list mailling
zope-db-bounces+m <zope-db@zope.org>
fmorss=aep.com@zo cc
pe.org
Subject
Re: [Zope-DB] DCOracle2 on AIX5.2
03/15/2006 10:02 -- issues
AM

On 15 Mar 2006, at 14:17, mfmorss@aep.com wrote:
> I posted originally about these C issues because I assumed it would be
> advantageous to Zope if its various parts could be installed whereever
> Python could, and that the developers would therefore take some
> interest in
> my troubles. If they don't, we will probably just look for an
> alternative
> Python interface to Oracle. That would be too bad because
> DCOracle2 does
> seem to work in spite of the compiler complaints, and it seems
> quite nice
> to use.

You need to keep in mind that the developers who worked on DCOracle2
over the years probably have no access to any system running AIX. AIX
is considered too oddballish to worry about by most Zope developers.
The people who do run AIX need to step up to the plate and help out,
there is no other way.

jens

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db

Re: DCOracle2 on AIX5.2 -- issues [ In reply to ]

dieter at handshake

Mar 15, 2006, 12:02 PM

Post #7 of 7 (4570 views)

Permalink

mfmorss@aep.com wrote at 2006-3-15 09:17 -0500:
> ...
>If you fix someone's C code, you really have to wonder if you did it
>correctly, and you also have to wonder if the next release will contain
>code that is even more difficult to fix. In short, you can hardly trust
>software for industrial use if you have to patch it by hand.

"DCOracle2" currently has no maintainer.
You probably should switch to "cx_Oracle" which has a maintainer...

>I posted originally about these C issues because I assumed it would be
>advantageous to Zope if its various parts could be installed whereever
>Python could, and that the developers would therefore take some interest in
>my troubles. If they don't, we will probably just look for an alternative
>Python interface to Oracle. That would be too bad because DCOracle2 does
>seem to work in spite of the compiler complaints, and it seems quite nice
>to use.

Then an additional option is to ignore the compiler warnings...

--
Dieter
_______________________________________________
Zope-DB mailing list
Zope-DB@zope.org
http://mail.zope.org/mailman/listinfo/zope-db