Thread: 723 failed to compile [FreeBSD-4.7-STABLE]
I have 722 installed and tried to build 723. 723 fails to compile. Attached to this message is the output of the build as captured by script(1). I already sent a PR to freebsd-ports but the answer was that this already happened and people got to compile it without changing anything (that they knew of, I guess). I am getting this error with FreeBSD-4.6.2p2 through 4.7RC0 and now 4.7-STABLE (updated last night) Someone mentioned that this error is related to the lack of awk in the system. I have: awk: /usr/bin/awk nawk: /usr/bin/nawk and just in case I added a link in /usr/local/bin (it's in the PATH): gawk -> /usr/bin/awk [fernan@pi] awk --version GNU Awk 3.0.6 Copyright (C) 1989, 1991-2000 Free Software Foundation. I am now turning to postgres people for help. I am writing pgsql-ports as said on the pgsql mailing lists page. [fernan@pi] uname -a FreeBSD pi.iib.unsam.edu.ar 4.7-STABLE FreeBSD 4.7-STABLE #1: Fri Oct 18 12:15:57 ART 2002 fernan@pi.iib.unsam.edu.ar:/usr/obj/usr/src/sys/PI i386 [fernan@pi] gcc --version 2.95.4 [fernan@pi] cc --version 2.95.4 Thanks in advance for any tip or suggestion, Fernan -- F e r n a n A g u e r o http://genoma.unsam.edu.ar/~fernan
Fernan Aguero <fernan@iib.unsam.edu.ar> writes:
> 723 fails to compile. Attached to this message is the output
> of the build as captured by script(1).
Er, you forgot the attachment...
regards, tom lane
Fernan Aguero <fernan@iib.unsam.edu.ar> writes:
> cc -O -pipe -Wall -Wmissing-prototypes -Wmissing-declarations -I../../../.=
> ./src/include -I/usr/local/include -c -o tuptoaster.o tuptoaster.c=0D
> tuptoaster.c: In function `toast_delete_datum':=0D
> tuptoaster.c:914: `F_OIDEQ' undeclared (first use in this function)=0D
> tuptoaster.c:914: (Each undeclared identifier is reported only once=0D
> tuptoaster.c:914: for each function it appears in.)=0D
> tuptoaster.c: In function `toast_fetch_datum':=0D
> tuptoaster.c:997: `F_OIDEQ' undeclared (first use in this function)=0D
> gmake[5]: *** [tuptoaster.o] Error 1=0D
This symptom usually means that the src/backend/utils/Gen_fmgrtab.sh script
failed to build a valid fmgroids.h file. This part of your trace looks
pretty suspicious:
/usr/local/bin/gmake -C utils fmgroids.h
gmake[4]: Entering directory `/usr/ports/databases/postgresql7/work/postgresql-7.2.3/src/backend/utils'
CPP='cc -E' AWK='nawk' /bin/sh Gen_fmgrtab.sh ../../../src/include/catalog/pg_proc.h
cc -E: not found
gmake[4]: Leaving directory
I think there's something wrong with your /bin/sh --- it looks to be
trying to interpret the -E switch as part of the cc filename. Anyone
seen anything like that before? Setting CPP to 'gcc -E' or 'cc -E' is
the usual value, so it's not like this is untested ...
regards, tom lane
Fernan Aguero <fernan@iib.unsam.edu.ar> writes:
> I guess that this might be the FreeBSD ports Makefile
> messing things up. In FreeBSD all ports are compiled from a
> Makefile (call it meta-Makefile) that sets some variables,
> drives the inclusion of FBSD specific patches, provides for
> a unified localization of non-system base binaries,
> libraries, etc. I guess that either the Makefile for the
> postgresql7 port (easy) or some of the more general bsd.*.mk
> files (that are included in each ports makefile, more
> difficult to fix, since this can break thousands of ports) is
> conflicting with postgresql's own Makefiles.
But how come we've not heard complaints from anyone else using FreeBSD?
regards, tom lane
On Tue, 2002-10-22 at 08:34, Tom Lane wrote: > Fernan Aguero <fernan@iib.unsam.edu.ar> writes: > > I guess that this might be the FreeBSD ports Makefile > > messing things up. In FreeBSD all ports are compiled from a > > Makefile (call it meta-Makefile) that sets some variables, > > drives the inclusion of FBSD specific patches, provides for > > a unified localization of non-system base binaries, > > libraries, etc. I guess that either the Makefile for the > > postgresql7 port (easy) or some of the more general bsd.*.mk > > files (that are included in each ports makefile, more > > difficult to fix, since this can break thousands of ports) is > > conflicting with postgresql's own Makefiles. > > But how come we've not heard complaints from anyone else using FreeBSD? I just (yesterday) did a recompile of 7.2.3, and it worked fine from CURRENT FreeBSD Ports. FWIW. LER -- Larry Rosenman http://www.lerctr.org/~ler Phone: +1 972-414-9812 E-Mail: ler@lerctr.org US Mail: 1905 Steamboat Springs Drive, Garland, TX 75044-6749
Hi Tom, thanks for your reply. Indeed you hit the point rightly. See below. +----[ Esto dijo Tom Lane (tgl@sss.pgh.pa.us): | | Fernan Aguero <fernan@iib.unsam.edu.ar> writes: | > cc -O -pipe -Wall -Wmissing-prototypes -Wmissing-declarations -I../../../.= | > ./src/include -I/usr/local/include -c -o tuptoaster.o tuptoaster.c=0D | > tuptoaster.c: In function `toast_delete_datum':=0D | > tuptoaster.c:914: `F_OIDEQ' undeclared (first use in this function)=0D | > tuptoaster.c:914: (Each undeclared identifier is reported only once=0D | > tuptoaster.c:914: for each function it appears in.)=0D | > tuptoaster.c: In function `toast_fetch_datum':=0D | > tuptoaster.c:997: `F_OIDEQ' undeclared (first use in this function)=0D | > gmake[5]: *** [tuptoaster.o] Error 1=0D | | This symptom usually means that the src/backend/utils/Gen_fmgrtab.sh script | failed to build a valid fmgroids.h file. This part of your trace looks | pretty suspicious: | | /usr/local/bin/gmake -C utils fmgroids.h | gmake[4]: Entering directory `/usr/ports/databases/postgresql7/work/postgresql-7.2.3/src/backend/utils' | CPP='cc -E' AWK='nawk' /bin/sh Gen_fmgrtab.sh ../../../src/include/catalog/pg_proc.h | cc -E: not found | gmake[4]: Leaving directory The problem is here, no doubt. This was also pointed by someone at the freebsd-ports@freebsd.org mailing list. | | I think there's something wrong with your /bin/sh --- it looks to be | trying to interpret the -E switch as part of the cc filename. Anyone | seen anything like that before? Setting CPP to 'gcc -E' or 'cc -E' is | the usual value, so it's not like this is untested ... I guess that this might be the FreeBSD ports Makefile messing things up. In FreeBSD all ports are compiled from a Makefile (call it meta-Makefile) that sets some variables, drives the inclusion of FBSD specific patches, provides for a unified localization of non-system base binaries, libraries, etc. I guess that either the Makefile for the postgresql7 port (easy) or some of the more general bsd.*.mk files (that are included in each ports makefile, more difficult to fix, since this can break thousands of ports) is conflicting with postgresql's own Makefiles. Calling make like this worked, and pgsql compiled OK: env -i PATH=$PATH make (note that I don't call make within the extracted pgsql sources (i.e. on pgsql's Makefile) but on the ports meta-Makefile). So my guess, again, is that it was a pretty simple fix, and that the FreeBSD ports Makefiles (whichever) are the ones breaking it. Thanks again for your reply, Fernan | | regards, tom lane | +----] -- F e r n a n A g u e r o http://genoma.unsam.edu.ar/~fernan
+----[ Esto dijo Tom Lane (tgl@sss.pgh.pa.us): | | Fernan Aguero <fernan@iib.unsam.edu.ar> writes: | > I guess that this might be the FreeBSD ports Makefile | > messing things up. In FreeBSD all ports are compiled from a | > Makefile (call it meta-Makefile) that sets some variables, | > drives the inclusion of FBSD specific patches, provides for | > a unified localization of non-system base binaries, | > libraries, etc. I guess that either the Makefile for the | > postgresql7 port (easy) or some of the more general bsd.*.mk | > files (that are included in each ports makefile, more | > difficult to fix, since this can break thousands of ports) is | > conflicting with postgresql's own Makefiles. | | But how come we've not heard complaints from anyone else using FreeBSD? Don't know, but: i) you can always compile from sources without using the FreeBSD ports system. Perhaps this kind of users don't see this problems. ii) Not all users carry the exact same FreeBSD ports tree. Differences in the main bsd.*.mk files could have been introduced after updating the postgresql7 port to 723. But this is just a wild guess. The problem I reported has been experienced by others. See the following thread: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=263497+0+current/freebsd-ports Based on this problem reports (PRs): http://www.freebsd.org/cgi/query-pr.cgi?pr=44327 http://www.freebsd.org/cgi/query-pr.cgi?pr=44097 | regards, tom lane regards, Fernan (thanks again) | +----] -- F e r n a n A g u e r o http://genoma.unsam.edu.ar/~fernan
Fernan Aguero writes: > Calling make like this worked, and pgsql compiled OK: > env -i PATH=$PATH make > (note that I don't call make within the extracted pgsql > sources (i.e. on pgsql's Makefile) but on the ports > meta-Makefile). In that case it would be instructive to see what's actually in your environment. Try to unset each environment variable and see which one makes the problem go away. -- Peter Eisentraut peter_e@gmx.net