Heads up: RC2 this evening

On Mon, Dec 20, 2004 at 02:04:46PM -0500, Tom Lane wrote:

Barring loud squawks, Marc will bundle up 8.0RC2 this evening sometime.
Anybody got last-minute stuff?

Is the following a plperl problem or does it need to be fixed in
DBD::PgSPI? I never saw any responses.

http://archives.postgresql.org/pgsql-bugs/2004-12/msg00097.php

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

Michael Fuhr <mike@fuhr.org> writes:

Is the following a plperl problem or does it need to be fixed in
DBD::PgSPI? I never saw any responses.

http://archives.postgresql.org/pgsql-bugs/2004-12/msg00097.php

AFAIK it's a PgSPI issue. plperl wraps its spi.c calls in a
subtransaction, but it looks like PgSPI isn't doing that.

regards, tom lane

Tom Lane said:

Barring loud squawks, Marc will bundle up 8.0RC2 this evening sometime.
Anybody got last-minute stuff?

I have not been able to build Cygwin with pltcl, and neither has anyone else
to the best of my knowledge.

I will investigate - probably a makefile issue - unless someone else has
solved the problem.

cheers

andrew

"Andrew Dunstan" <andrew@dunslane.net> writes:

Anybody got last-minute stuff?

I have not been able to build Cygwin with pltcl, and neither has anyone else
to the best of my knowledge.

Has that worked in prior releases?

regards, tom lane

Tom Lane wrote:

"Andrew Dunstan" <andrew@dunslane.net> writes:

Anybody got last-minute stuff?

I have not been able to build Cygwin with pltcl, and neither has anyone else
to the best of my knowledge.

Has that worked in prior releases?

I have no idea. It's hard to think of a reason in principle why it
shouldn't.

cheers

andrew

Andrew Dunstan <andrew@dunslane.net> writes:

Tom Lane wrote:

"Andrew Dunstan" <andrew@dunslane.net> writes:

I have not been able to build Cygwin with pltcl, and neither has anyone else
to the best of my knowledge.

Has that worked in prior releases?

I have no idea. It's hard to think of a reason in principle why it
shouldn't.

Nope. What are the symptoms exactly?

regards, tom lane

Andrew Dunstan <andrew@dunslane.net> writes:

Tom Lane wrote:

Nope. What are the symptoms exactly?

log attached. Looks like there is at least a missing "-ltcl" in the call
to dllwrap, but that's not all.

Agreed on missing -ltcl. It seems odd given that the Cygwin case in
Makefile.shlib does include $(SHLIB_LINK) and pltcl's Makefile does add
$(TCL_LIB_SPEC) to SHLIB_LINK. Is TCL_LIB_SPEC getting set reasonably
by configure? On my machine, Makefile.global ends up with

TCL_LIB_FILE = libtcl8.4.so
TCL_LIBS = -ldld -lm
TCL_LIB_SPEC = -L/opt/tcl8.4/lib -ltcl8.4
TCL_INCLUDE_SPEC = -I/opt/tcl8.4/include

No idea about the other errors ... one would think they'd be suppressed
by the -lpostgres, but evidently not.

regards, tom lane

Andrew Dunstan <andrew@dunslane.net> writes:

Tom Lane wrote:

Agreed on missing -ltcl. It seems odd given that the Cygwin case in
Makefile.shlib does include $(SHLIB_LINK) and pltcl's Makefile does add
$(TCL_LIB_SPEC) to SHLIB_LINK. Is TCL_LIB_SPEC getting set reasonably
by configure?

$ grep TCL_ Makefile.global
TCL_LIB_FILE = libtcl84.a
TCL_LIBS =
TCL_LIB_SPEC =
TCL_INCLUDE_SPEC = -I/nonexistent/include
TCL_SHARED_BUILD = 1
TCL_SHLIB_LD_LIBS =

Yuck.

here's what is in /usr/lib/tclConfig.sh - maybe there's a clue in there
- or maybe it's just a problem with the Cygwin-supplied package - I have
deliberately not tried to fix this by installing my own build of tcl.

The Cygwin-supplied package is evidently broken beyond belief.
configure is correctly copying these settings from tclConfig.sh, but the
data therein is wrong (or at least useless). If tclConfig.sh doesn't
provide usable information, that is not our problem to fix.

regards, tom lane

Tom Lane wrote:

here's what is in /usr/lib/tclConfig.sh - maybe there's a clue in there
- or maybe it's just a problem with the Cygwin-supplied package - I have
deliberately not tried to fix this by installing my own build of tcl.

The Cygwin-supplied package is evidently broken beyond belief.
configure is correctly copying these settings from tclConfig.sh, but the
data therein is wrong (or at least useless). If tclConfig.sh doesn't
provide usable information, that is not our problem to fix.

I am abandoning further effort altogether, because of this:

adunstan: ~/tcl8.4.9/win
$ ./configure --enable-shared
creating cache ./config.cache
checking for gcc... gcc
checking whether the C compiler (gcc ) works... yes
checking whether the C compiler (gcc ) is a cross-compiler... no
checking whether we are using GNU C... yes
checking whether gcc accepts -g... yes
checking for ar... ar
checking for ranlib... ranlib
checking for windres... windres
checking whether make sets ${MAKE}... yes
checking for Cygwin environment... yes
configure: error: Compiling under Cygwin is not currently supported.
A maintainer for the Cygwin port of Tcl/Tk is needed. See the README
file for information about building with Mingw.

cheers

andrew

Andrew Dunstan <andrew@dunslane.net> writes:

I am abandoning further effort altogether, because of this:

adunstan: ~/tcl8.4.9/win
$ ./configure --enable-shared
checking for Cygwin environment... yes
configure: error: Compiling under Cygwin is not currently supported.
A maintainer for the Cygwin port of Tcl/Tk is needed. See the README
file for information about building with Mingw.

Wow. That implies that the Tcl community knows about the problems and
considers them nontrivial to fix (else they'd have just fixed 'em).
So we'll just forget that for now...

regards, tom lane

Andrew Dunstan wrote:

I have not been able to build Cygwin with pltcl, and neither has
anyone else to the best of my knowledge.

This problem has existed for years. See the pgsql-cygwin(?) archives
for details.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/