7.1 RC1 RPM

Mike Cannon-Brookes wrote:

Any change of getting a 7.1 RC1 RPM? I'm using the beta4 RPMs at the moment
but don't seem to be any more recent ones.

I'm building a quickie RC1-1 RPM right now. There are some other things
I need to do on the RPMset before final release -- and I plan on working
a while this Saturday on that -- but my week is so loaded that I'm going
to put out a rebuild of 7.1beta6->7.1RC1 as is -- once I get it to
build.....

Stay tuned...
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Lamar Owen wrote:

Mike Cannon-Brookes wrote:

Any change of getting a 7.1 RC1 RPM? I'm using the beta4 RPMs at the moment
but don't seem to be any more recent ones.

I'm building a quickie RC1-1 RPM right now. There are some other things
I need to do on the RPMset before final release -- and I plan on working
a while this Saturday on that -- but my week is so loaded that I'm going
to put out a rebuild of 7.1beta6->7.1RC1 as is -- once I get it to
build.....

Stay tuned...

Ok, 7.1RC1 RPM uploaded to ftp://ftp.postgresql.org/pub/dev/test-rpms as
before.

BIG NOTE: These are built on Red Hat _7.0_ NOT 6.2 as previous ones
have been. The ODBC client build on 6.2 has been broken -- it built at
beta4, but now gives this error set at beta6/RC1:

ar crs libpsqlodbc.a info.o bind.o columninfo.o connection.o convert.o
drvconn.o environ.o execute.o lobj.o misc.o options.o pgtypes.o
psqlodbc.o qresult.o results.o socket.o parse.o statement.o gpps.o
tuple.o tuplelist.o dlg_specific.o
ranlib libpsqlodbc.a
gcc -O2 -Wall -Wmissing-prototypes -Wmissing-declarations -O2 -m486
-fno-strength-reduce -fpic -shared -Wl,-soname,libpsqlodbc.so.0
-Wl,-Bsymbolic info.o bind.o columninfo.o connection.o convert.o
drvconn.o environ.o execute.o lobj.o misc.o options.o pgtypes.o
psqlodbc.o qresult.o results.o socket.o parse.o statement.o gpps.o
tuple.o tuplelist.o dlg_specific.o -lm -Wl,-rpath,/usr/lib -o
libpsqlodbc.so.0.26
connection.o: In function `CC_connect':
connection.o(.text+0x914): undefined reference to
`check_client_encoding'
connection.o(.text+0x91a): undefined reference to
`check_client_encoding'
convert.o: In function `copy_statement_with_parameters':
convert.o(.text+0x110f): undefined reference to `multibyte_init'
convert.o(.text+0x1191): undefined reference to `multibyte_char_check'
convert.o(.text+0x11c3): undefined reference to `multibyte_strchr'
convert.o: In function `convert_special_chars':
convert.o(.text+0x2949): undefined reference to `multibyte_init'
convert.o(.text+0x29c4): undefined reference to `multibyte_char_check'
parse.o: In function `getNextToken':
parse.o(.text+0x151): undefined reference to `multibyte_char_check'
dlg_specific.o: In function `getDSNinfo':
dlg_specific.o(.text+0xd8c): undefined reference to
`check_client_encoding'
dlg_specific.o(.text+0xd98): undefined reference to
`check_client_encoding'
collect2: ld returned 1 exit status
make[3]: *** [libpsqlodbc.so.0.26] Error 1
make[3]: Leaving directory
`/root/rpm-building/redhat/BUILD/postgresql-7.1RC1/src/interfaces/odbc'
make[2]: *** [all] Error 2
make[2]: Leaving directory
`/root/rpm-building/redhat/BUILD/postgresql-7.1RC1/src/interfaces'
make[1]: *** [all] Error 2
make[1]: Leaving directory
`/root/rpm-building/redhat/BUILD/postgresql-7.1RC1/src'
make: *** [all] Error 2

There are no other errors in the ODBC build that I could find. Again --
this builds on Red Hat 7, but not RH 6.2.

Now, according to the cvsweb on postgresql.org, Bruce has just as of 14
hours ago done an update in this dir. What updates were done between
beta4 and beta6? (Is that not when the flap about patches occurred?).
Well, Red Hat 7 will build it -- and Red Hat 6.2 will now NOT build it
-- but it did build beta4 successfully.

Well, in any case, preliminary 7.1RC1 RPMS are up. There are some odd
issues with the packaging that I am working on. Be sure to read
README.rpm-dist -- attached to this message for your convenience.

THIS IS A PRERELEASE RPM. Please use for testing ONLY.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Lamar Owen <lamar.owen@wgcr.org> writes:

BIG NOTE: These are built on Red Hat _7.0_ NOT 6.2 as previous ones
have been. The ODBC client build on 6.2 has been broken -- it built at
beta4, but now gives this error set at beta6/RC1:

ar crs libpsqlodbc.a info.o bind.o columninfo.o connection.o convert.o
drvconn.o environ.o execute.o lobj.o misc.o options.o pgtypes.o
psqlodbc.o qresult.o results.o socket.o parse.o statement.o gpps.o
tuple.o tuplelist.o dlg_specific.o
ranlib libpsqlodbc.a
gcc -O2 -Wall -Wmissing-prototypes -Wmissing-declarations -O2 -m486
-fno-strength-reduce -fpic -shared -Wl,-soname,libpsqlodbc.so.0
-Wl,-Bsymbolic info.o bind.o columninfo.o connection.o convert.o
drvconn.o environ.o execute.o lobj.o misc.o options.o pgtypes.o
psqlodbc.o qresult.o results.o socket.o parse.o statement.o gpps.o
tuple.o tuplelist.o dlg_specific.o -lm -Wl,-rpath,/usr/lib -o
libpsqlodbc.so.0.26
connection.o: In function `CC_connect':
connection.o(.text+0x914): undefined reference to
`check_client_encoding'
connection.o(.text+0x91a): undefined reference to
`check_client_encoding'

It would appear you have a conflict about whether MULTIBYTE is defined
or not --- the code thinks so, but the makefile does not, since
multibyte.o is not seen in the link command.

The identical technique is used in libpq's makefile, so I'm not sure
why you do not see a link failure in libpq as well.

regards, tom lane

Lamar Owen wrote:

Well, in any case, preliminary 7.1RC1 RPMS are up. There are some odd
issues with the packaging that I am working on. Be sure to read
README.rpm-dist -- attached to this message for your convenience.

Forgot to attach the file. :-(.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Tom Lane wrote:

It would appear you have a conflict about whether MULTIBYTE is defined
or not --- the code thinks so, but the makefile does not, since
multibyte.o is not seen in the link command.

The identical technique is used in libpq's makefile, so I'm not sure
why you do not see a link failure in libpq as well.

Hmmmm. Hiroshi committed an update to GNUmakefile to 'enable multibyte
support' for ODBC. But that was only 33 hours ago -- meaning it wasn't
updated in time for RC1. Lessee..... I'm rebuilding RC1 with Hiroshi's
GNUmakefile change as part of the RPMset patch -- and it succeeds.

But it succeeds on RH7 _without_ Hiroshi's patch. Odd. Lessee....
Examining my logs of the build on RH7 shows that the same error occurs
-- it just doesn't abort the build. Argh.

This means that the binary up on the ftp site right now for ODBC is
broken. I'll fix it tonight or tomorrow -- and the source RPM won't
rebuild on RH 6.2. I'll upload a -2 set tonight or tomorrow to fix
that, and a few other issues I found while dinking with it today.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

FYI: I successfully built a beta4 RPM for Mandrake, having to make only
the following changes:

1) add -fno-fast-math to the CFLAGS set from RPM_OPT_FLAGS. This fixes
the date/time rounding problems. Would seem to be a non-invasive fix to
put this in for every build, or to flag it as a Mandrake-only feature
(but I have not gotten a suggestion on how to do that).

2) explicitly remove Pg.bs from the list of files for the perl
installation. On my system (with perl-5.6.0) that is a zero-length file
which seems to disappear (or never gets copied) during the installation
process. Not sure why it does not propagate, but it does not seem to be
essential.

Comments?

- Thomas

Thomas Lockhart wrote:

FYI: I successfully built a beta4 RPM for Mandrake, having to make only
the following changes:

1) add -fno-fast-math to the CFLAGS set from RPM_OPT_FLAGS. This fixes
the date/time rounding problems. Would seem to be a non-invasive fix to
put this in for every build, or to flag it as a Mandrake-only feature
(but I have not gotten a suggestion on how to do that).

I've got the code Tom and I came up with to deal with the LinuxPPC
nonsense of a few releases ago... That or similar would be the ticket.
A sed pipeline re-setting CFLAGS by removing 'unsafe' options from the
flags would be useful.

I'm working on a 'what distribution is this' deal for the spec file.

2) explicitly remove Pg.bs from the list of files for the perl
installation. On my system (with perl-5.6.0) that is a zero-length file
which seems to disappear (or never gets copied) during the installation
process. Not sure why it does not propagate, but it does not seem to be
essential.

Interesting. I have found a couple of other non-props -- including the
README.rpm-dist, for some reason. I'm going to have to trace the build
in detail. And the html docs tree is going to the wrong place...

In any case, a unified or context diff against the 7.1beta4 spec would
be useful.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

Lamar Owen wrote:

Tom Lane wrote:

It would appear you have a conflict about whether MULTIBYTE is defined
or not --- the code thinks so, but the makefile does not, since
multibyte.o is not seen in the link command.

The identical technique is used in libpq's makefile, so I'm not sure
why you do not see a link failure in libpq as well.

Hmmmm. Hiroshi committed an update to GNUmakefile to 'enable multibyte
support' for ODBC. But that was only 33 hours ago -- meaning it wasn't
updated in time for RC1. Lessee..... I'm rebuilding RC1 with Hiroshi's
GNUmakefile change as part of the RPMset patch -- and it succeeds.

Oops I apologize for my mistake. I must have changed the
GNUmakefile when I committed multibyte support for ODBC.

regards,
Hiroshi Inoue

In any case, a unified or context diff against the 7.1beta4 spec would
be useful.

OK, here is a context diff of the spec file (note only two lines
changed). This addresses the "-fno-fast-math" problem on Mandrake (at
least 7.2 and earlier), and the "disappearing Pg.bs file" problem on
same.

- Thomas

Hiroshi Inoue wrote:

Lamar Owen wrote:

Hmmmm. Hiroshi committed an update to GNUmakefile to 'enable multibyte
support' for ODBC. But that was only 33 hours ago -- meaning it wasn't
updated in time for RC1. Lessee..... I'm rebuilding RC1 with Hiroshi's
GNUmakefile change as part of the RPMset patch -- and it succeeds.

Oops I apologize for my mistake. I must have changed the
GNUmakefile when I committed multibyte support for ODBC.

I'm afraid I may have been misunderstood. The change you made _fixed_
the problem -- it didn't cause it. You made no error.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11