Horton

Thread

  1. Re: [PORTS] Postgress installation in HP-UX 10.20.

    Thomas Lockhart <lockhart@alumni.caltech.edu> — 1998-01-06T01:55:02Z

    > Hi everybody:
> I have the following information for you.
> I installed Postgress 6.2.1 in the next platform:
> Apollo 9000 Serie 800 with PA-RISC running HP-UX 10.20
> - The version of PostgreSQL (v6.2.1, 6.1.1, beta 970703, etc.).
> v6.2.1
> - Your operating system (i.e. RedHat v4.0 Linux v2.0.26).
> HP-UX 10.20
> - Your hardware (SPARC, i486, etc.).
> Precision Arquitecture-RISC (32 o 64 Mz ???)
> - Install problems:
> I had problems running HP's make so I need to install 'gmake'. By the
> way, I couldn't find a program called gmake so I got 'make' instead.
> I did ftp from prep.ai.mit.edu. File:/pub/gnu/make-3.76.1.tar.gz
>
> I hadn't any problems to install gnu make.
>
> As to the Postgress installation I had the following problems:
> When I ran 'make all', I got a yacc error:
> Too many states compiling gram.y. (I got a clue from the yacc compiler:
> use Ns option)
> Solution: I edited the Makefile.global file. I modified the line with
> the YFLAGS variable (line 211), so I added the option Ns with a value of
> 5000(The default for HP was 1000)
> After this change The problem vanished but I found the next problem:The
> size of look ahead tables was not big enough. (Default 650) So I
> modified to 2500 with the Nl (En -el) option.
> At last the line was modified in the following way:
> Original line:
> YFLAGS= -d
> Modified line
> YFLAGS= -d -Ns5000 -Nl2500
>
> After this I got the next fatal error:
> cc -I../../include  -W l,-E -Ae -DNOFIXADE  -Dhpux -I.. -I.
> include   -c scan.c -o scan.o
> cc: "scan.c", line 145: error 1000: Unexpected symbol: "&".
>
> The problem was very simple to solve. One comment was erronous written.
> The '/' was missing. I just edited the file scan.c and everythig worked
> fine.

Oh, I assumed that this was a comment from scan.l, but I'm now guessing
that this was a comment inserted by HP's lex program. Yes?

> I ran the regress test and I could find some tests failed principally
> due to the floating point precision.

Which is OK.

Good information. Anyone interested in typing this up as a FAQ
(doc/FAQ_HP)?

                                                   - Tom