Thread: RPM RH9.0 conflict with unixODBC

RPM RH9.0 conflict with unixODBC

From
Gaetano Mendola
Date:
This is what I got installing the RPM for 9.0:

file /usr/include/sqltypes.h from install of
postgresql-devel-7.4-0.3PGDG conflicts with file from package
unixODBC-devel-2.2.3-6



BTW I think that something is wrong with the file size
between two distributions:

this are the rpm for 9.0:

3768418 Nov 24 10:07 postgresql-7.4-0.3PGDG.i386.rpm
2660177 Nov 24 10:07 postgresql-contrib-7.4-0.3PGDG.i386.rpm
2030011 Nov 24 10:08 postgresql-devel-7.4-0.3PGDG.i386.rpm
1185543 Nov 24 10:07 postgresql-docs-7.4-0.3PGDG.i386.rpm
  948473 Nov 24 10:07 postgresql-jdbc-7.4-0.3PGDG.i386.rpm
  760377 Nov 24 10:07 postgresql-libs-7.4-0.3PGDG.i386.rpm
  660158 Nov 24 10:07 postgresql-pl-7.4-0.3PGDG.i386.rpm
   98019 Nov 24 10:07 postgresql-python-7.4-0.3PGDG.i386.rpm
9864477 Nov 24 10:06 postgresql-server-7.4-0.3PGDG.i386.rpm
  158713 Nov 24 10:07 postgresql-tcl-7.4-0.3PGDG.i386.rpm
1053201 Nov 24 10:05 postgresql-test-7.4-0.3PGDG.i386.rpm


and this for rhas-2.1:

2080872 Nov 24 10:08 postgresql-7.4-0.3PGDG.i386.rpm
  487992 Nov 24 10:08 postgresql-contrib-7.4-0.3PGDG.i386.rpm
  674440 Nov 24 10:08 postgresql-devel-7.4-0.3PGDG.i386.rpm
1126355 Nov 24 10:09 postgresql-docs-7.4-0.3PGDG.i386.rpm
  948208 Nov 24 10:08 postgresql-jdbc-7.4-0.3PGDG.i386.rpm
  157377 Nov 24 10:08 postgresql-libs-7.4-0.3PGDG.i386.rpm
  248003 Nov 24 10:08 postgresql-pl-7.4-0.3PGDG.i386.rpm
   43065 Nov 24 10:08 postgresql-python-7.4-0.3PGDG.i386.rpm
3127278 Nov 24 10:08 postgresql-server-7.4-0.3PGDG.i386.rpm
   28270 Nov 24 10:09 postgresql-tcl-7.4-0.3PGDG.i386.rpm
1048176 Nov 24 10:09 postgresql-test-7.4-0.3PGDG.i386.rpm


how you can see the sizes are completely different.
Am I missing something ?

Regards
Gaetano Mendola




Re: RPM RH9.0 conflict with unixODBC

From
Lamar Owen
Date:
On Monday 24 November 2003 03:05 pm, Gaetano Mendola wrote:
> This is what I got installing the RPM for 9.0:

> file /usr/include/sqltypes.h from install of
> postgresql-devel-7.4-0.3PGDG conflicts with file from package
> unixODBC-devel-2.2.3-6

Namespace collision, I guess.  Makes me wonder about popping all those devel
files in /usr/include; I'll have to research.  I can confirm that the
conflict exists in Fedora Core; I don't typically have unixODBC-devel
installed on my build boxes.

> BTW I think that something is wrong with the file size
> between two distributions:

> this are the rpm for 9.0:

> 3768418 Nov 24 10:07 postgresql-7.4-0.3PGDG.i386.rpm

> and this for rhas-2.1:

> 2080872 Nov 24 10:08 postgresql-7.4-0.3PGDG.i386.rpm

> how you can see the sizes are completely different.
> Am I missing something ?

Compiler version and flags.  Although that is a substantial difference.  The
rhas-2.1 compiler is the Red Hat gcc 2.96, and the RH9 compiler is gcc 3.
Apparently gcc 3 generates large code.  Sander, can you verify this, since
you built both of these?
--
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu


Re: RPM RH9.0 conflict with unixODBC

From
"Sander Steffann"
Date:
Hi,

> > this are the rpm for 9.0:
> > 3768418 Nov 24 10:07 postgresql-7.4-0.3PGDG.i386.rpm
>
> > and this for rhas-2.1:
> > 2080872 Nov 24 10:08 postgresql-7.4-0.3PGDG.i386.rpm
>
> > how you can see the sizes are completely different.
> > Am I missing something ?
>
> Compiler version and flags.  Although that is a substantial difference.
The
> rhas-2.1 compiler is the Red Hat gcc 2.96, and the RH9 compiler is gcc 3.
> Apparently gcc 3 generates large code.  Sander, can you verify this, since
> you built both of these?

The file sizes are correct. I can't explain the huge difference though...
They both contain the same files, and are built from the same specfile. The
only difference between the build commands is that the RedHat 9 RPMs are
built with --define 'build89 1', and the RHEL RPMs are built with --define
'build7x 1' and --define 'plperl 0'. The exclusion of plperl shouldn't make
a difference in this RPM, only in the postgresql-pl RPM...

In short: both filesizes are correct, and I can't explain it :(
Sander.


Re: RPM RH9.0 conflict with unixODBC

From
Gaetano Mendola
Date:
Sander Steffann wrote:
> Hi,
>
>
>>>this are the rpm for 9.0:
>>>3768418 Nov 24 10:07 postgresql-7.4-0.3PGDG.i386.rpm
>>
>>>and this for rhas-2.1:
>>>2080872 Nov 24 10:08 postgresql-7.4-0.3PGDG.i386.rpm
>>
>>>how you can see the sizes are completely different.
>>>Am I missing something ?
>>
>>Compiler version and flags.  Although that is a substantial difference.
>
> The
>
>>rhas-2.1 compiler is the Red Hat gcc 2.96, and the RH9 compiler is gcc 3.
>>Apparently gcc 3 generates large code.  Sander, can you verify this, since
>>you built both of these?
>
>
> The file sizes are correct. I can't explain the huge difference though...
> They both contain the same files, and are built from the same specfile. The
> only difference between the build commands is that the RedHat 9 RPMs are
> built with --define 'build89 1', and the RHEL RPMs are built with --define
> 'build7x 1' and --define 'plperl 0'. The exclusion of plperl shouldn't make
> a difference in this RPM, only in the postgresql-pl RPM...
>
> In short: both filesizes are correct, and I can't explain it :(
> Sander.

However 6MB of difference for postgresql-server-7.4-0.3PGDG.i386.rpm
and     2.2 MB for postgresql-contrib-7.4-0.3PGDG.i386.rpm
are too much !


I'd like to have time to dig on it...


Regards
Gaetano Mendola



Re: RPM RH9.0 conflict with unixODBC

From
Steve Wampler
Date:
On Tue, 2003-11-25 at 12:39, Gaetano Mendola wrote:
> Sander Steffann wrote:
...
> > In short: both filesizes are correct, and I can't explain it :(
> > Sander.
>
> However 6MB of difference for postgresql-server-7.4-0.3PGDG.i386.rpm
> and     2.2 MB for postgresql-contrib-7.4-0.3PGDG.i386.rpm
> are too much !

If you do a 'file' command on the executables does one set show
up stripped while the other doesn't.  The sizes of individual
files are consistent with that difference (roughly).

If they are the same, and 'ldd' shows the same set of shared
libraries loaded, then look at the sizes of the non-shared
libraries used on the two OS versions.

--
Steve Wampler -- swampler@noao.edu
Quantum materiae materietur marmota monax si marmota
                    monax materiam possit materiari?

Re: RPM RH9.0 conflict with unixODBC

From
Gaetano Mendola
Date:
Steve Wampler wrote:

> On Tue, 2003-11-25 at 12:39, Gaetano Mendola wrote:
>
>>Sander Steffann wrote:
>
> ...
>
>>>In short: both filesizes are correct, and I can't explain it :(
>>>Sander.
>>
>>However 6MB of difference for postgresql-server-7.4-0.3PGDG.i386.rpm
>>and     2.2 MB for postgresql-contrib-7.4-0.3PGDG.i386.rpm
>>are too much !
>
>
> If you do a 'file' command on the executables does one set show
> up stripped while the other doesn't.  The sizes of individual
> files are consistent with that difference (roughly).
>
> If they are the same, and 'ldd' shows the same set of shared
> libraries loaded, then look at the sizes of the non-shared
> libraries used on the two OS versions.

This mean that file for 9.0 must be stripped ?

RHAS-2.1:
tsearch2.so   152888 bytes

RH9.0:
tsearch2.so   951236 bytes



compiling this file from a postgresql-7.4.tar.gz
on a RH9.0 I obtain:
libtsearch2.so.0.0   158565 bytes

that is most like a RHAS2.1 size.

RH9.0 where I compiled the tar.gz:

ldd libtsearch2.so.0.0
         libm.so.6 => /lib/tls/libm.so.6 (0x4003d000)
         libc.so.6 => /lib/tls/libc.so.6 (0x42000000)
         /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x80000000)

RH9.0:

ldd tsearch2.so
         libm.so.6 => /lib/tls/libm.so.6 (0x40040000)
         libc.so.6 => /lib/tls/libc.so.6 (0x42000000)
         /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x80000000)

RHAS2.1:

ldd tsearch2.so
         libm.so.6 => /lib/i686/libm.so.6 (0x4002c000)
         libc.so.6 => /lib/i686/libc.so.6 (0x4004f000)
         /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x80000000)



RHAS-2.1:

 > gdb tsearch2.so
GNU gdb Red Hat Linux (5.3post-0.20021129.18rh)
Copyright 2003 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i386-redhat-linux-gnu"...(no debugging
symbols found)...


RH9.0:

 > gdb tsearch2.so
GNU gdb Red Hat Linux (5.3post-0.20021129.18rh)
Copyright 2003 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i386-redhat-linux-gnu"...

note that there is no warning about the debugging symbols :-(


is strange because the gdb on libtsearch2.so.0.0  ( 158565 bytes )
don't complain neather about the debugging symbols...

Am I crazy ?




Regards
Gateano Mendola











Re: RPM RH9.0 conflict with unixODBC

From
Lamar Owen
Date:
On Tuesday 25 November 2003 03:38 pm, Gaetano Mendola wrote:
> RHAS-2.1:
> tsearch2.so   152888 bytes

> RH9.0:
> tsearch2.so   951236 bytes

FWIW, on Fedora Core:
[lowen@localhost pgsql]$ ls -l tsearch2.so
-rwxr-xr-x    1 root     root       141348 Nov 21 21:53 tsearch2.so
[lowen@localhost pgsql]$

Something is certainly unusual here.  Sander, can you rebuild the RH9 set and
see why it is so large?  For some reason, I missed how many places were in
there, and missed the fact that there were multiple megabytes difference.
Debugging symbols or no, this is big.

And as Fedora Core uses gcc 3, that doesn't explain it.
--
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu


Re: RPM RH9.0 conflict with unixODBC

From
"Sander Steffann"
Date:
Hi,

> Something is certainly unusual here.  Sander, can you rebuild the RH9 set
and
> see why it is so large?  For some reason, I missed how many places were in
> there, and missed the fact that there were multiple megabytes difference.
> Debugging symbols or no, this is big.

The difference in size is 100% related to the stripped/unstripped binaries.
These are the RH9 files in /usr/bin as they are in the RPM:

-rwxr-xr-x    1 planeet  sander     208173 Nov 23 01:40 clusterdb
-rwxr-xr-x    1 planeet  sander     208816 Nov 23 01:40 createdb
-rwxr-xr-x    1 planeet  sander     258577 Nov 23 01:40 createlang
-rwxr-xr-x    1 planeet  sander     208161 Nov 23 01:40 createuser
-rwxr-xr-x    1 planeet  sander     205248 Nov 23 01:40 dropdb
-rwxr-xr-x    1 planeet  sander     257890 Nov 23 01:40 droplang
-rwxr-xr-x    1 planeet  sander     205322 Nov 23 01:40 dropuser
-rwxr-xr-x    1 planeet  sander     739671 Nov 23 01:40 pg_dump
-rwxr-xr-x    1 planeet  sander     176362 Nov 23 01:40 pg_dumpall
-rwxr-xr-x    1 planeet  sander      32745 Nov 23 01:41 pg_encoding
-rwxr-xr-x    1 planeet  sander      36674 Nov 23 01:40 pg_id
-rwxr-xr-x    1 planeet  sander     539901 Nov 23 01:40 pg_restore
-rwxr-xr-x    1 planeet  sander     908106 Nov 23 01:40 psql
-rwxr-xr-x    1 planeet  sander     143996 Nov 23 01:40 vacuumdb

And after stripping them:

-rwxr-xr-x    1 planeet  sander      22936 Nov 26 01:15 clusterdb
-rwxr-xr-x    1 planeet  sander      22944 Nov 26 01:15 createdb
-rwxr-xr-x    1 planeet  sander      28160 Nov 26 01:15 createlang
-rwxr-xr-x    1 planeet  sander      23000 Nov 26 01:15 createuser
-rwxr-xr-x    1 planeet  sander      20772 Nov 26 01:15 dropdb
-rwxr-xr-x    1 planeet  sander      27564 Nov 26 01:15 droplang
-rwxr-xr-x    1 planeet  sander      20836 Nov 26 01:15 dropuser
-rwxr-xr-x    1 planeet  sander     160776 Nov 26 01:15 pg_dump
-rwxr-xr-x    1 planeet  sander      28824 Nov 26 01:15 pg_dumpall
-rwxr-xr-x    1 planeet  sander       4464 Nov 26 01:15 pg_encoding
-rwxr-xr-x    1 planeet  sander       4592 Nov 26 01:15 pg_id
-rwxr-xr-x    1 planeet  sander      77120 Nov 26 01:15 pg_restore
-rwxr-xr-x    1 planeet  sander     152344 Nov 26 01:15 psql
-rwxr-xr-x    1 planeet  sander      13012 Nov 26 01:15 vacuumdb

This makes the difference between 4.1M (before) and 644K (after). I just
noticed that I disabled the debug-package that RH9 builds by default. That
very probably causes this difference. I will rebuild the RPMs with the
debug-package enabled to see what happens.

Sander.


Re: RPM RH9.0 conflict with unixODBC

From
"Sander Steffann"
Date:
Hi,

It turns out that preventing RH9 from building the debuginfo package also
prevented it from stripping the binaries. This was what caused the big
difference in filesize. I have rebuilt the RPMs for RH9 and put them on
http://opensource.nederland.net/.

I had to make a small modification to the specfile (again) because it seems
that macro's work differently for each RPM / RedHat version. There have been
no other changes to the sources or specfile, so the end-result is the same.

Sorry for the inconvenience I caused by disabling the debuginfo package!
Sander.


Re: RPM RH9.0 conflict with unixODBC

From
Gaetano Mendola
Date:
Sander Steffann wrote:

> Hi,
>
> It turns out that preventing RH9 from building the debuginfo package also
> prevented it from stripping the binaries. This was what caused the big
> difference in filesize. I have rebuilt the RPMs for RH9 and put them on
> http://opensource.nederland.net/.
>
> I had to make a small modification to the specfile (again) because it seems
> that macro's work differently for each RPM / RedHat version. There have been
> no other changes to the sources or specfile, so the end-result is the same.
>
> Sorry for the inconvenience I caused by disabling the debuginfo package!
> Sander.

Is this also related to the fact that gdb on libraries of RH9.0 don't
complain about the debugging info ?


Regards
Gaetano Mendola



Re: RPM RH9.0 conflict with unixODBC

From
Lamar Owen
Date:
On Tuesday 25 November 2003 07:56 pm, Sander Steffann wrote:
> It turns out that preventing RH9 from building the debuginfo package also
> prevented it from stripping the binaries.

Ah, ok.

> This was what caused the big
> difference in filesize. I have rebuilt the RPMs for RH9 and put them on
> http://opensource.nederland.net/.

> I had to make a small modification to the specfile (again) because it seems
> that macro's work differently for each RPM / RedHat version. There have
> been no other changes to the sources or specfile, so the end-result is the
> same.

Ooo... yet more macro differences.  Saturday I'll need to put some time into
understanding how to make a singular set...

> Sorry for the inconvenience I caused by disabling the debuginfo package!
> Sander.

Not a big problem.

I'll pull down the new packages tomorrow afternoon; I was going to be off, but
will have to come in for a couple of hours. So I'll transfer them across
then.

Many thanks!
--
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu


Re: RPM RH9.0 conflict with unixODBC

From
Lamar Owen
Date:
On Tuesday 25 November 2003 07:22 pm, Gaetano Mendola wrote:
> Sander Steffann wrote:
> > Sorry for the inconvenience I caused by disabling the debuginfo package!
> > Sander.

> Is this also related to the fact that gdb on libraries of RH9.0 don't
> complain about the debugging info ?

I would think so.  But with Sander confirming this by file sizes, we know
pretty much for certain.  Debugging info really jacks up the file size.  It
didn't use to be quite that big of a difference, back when I built the beta
packages (of which we didn't have any this cycle) with debug info turned on,
and unstripped.
--
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu