[MITgcm-devel] test on SunOS

Wed Mar 12 14:17:33 EDT 2008

Hi Martin,

I agree. However, might not have the time to do it right now,
so could think of using the same testreport trick temporarily
(if you want to pass a test on edvir1 this coming WE) 
until this getopt.h is fixed.
Cheers,
Jean-Michel

On Wed, Mar 12, 2008 at 09:54:44AM +0100, Martin Losch wrote:
> Hi Jean-Michel,
> this new flag seems to work for solaris (rays1).
> 
> 
> for edvir1 I can do the same, but I would still suggest that we  
> modify the ./configure script somehow to get around the getopt.h  
> option. The OS on that system is not as exotic as Solaris, and many  
> people are probably using that kind of platform. I guess anybody on  
> an IBM p690 or similar should have this problem (if he/she want's to  
> run testreport).
> 
> What do you think?
> 
> Martin
> 
> On 12 Mar 2008, at 03:36, Jean-Michel Campin wrote:
> 
> >Martin,
> >
> >I've just checked-in the little changes in testreport to allow you
> >to use the "backup" solution (i.e., build mpack with the old version
> >somewhere and specify "-mpackdir="somewhere" or -mpd somewhere).
> >It would be good if you could try it.
> >
> >Jean-Michel
> >
> >On Tue, Mar 11, 2008 at 08:33:24PM -0400, Jean-Michel Campin wrote:
> >>Hi Martin,
> >>
> >>we could fix the configure script for this getopt.h problem.
> >>
> >>Regarding stdint.h, it's more tricky:
> >>it is supposed to come with glibc package, so might be that
> >>you don't have a full gcc installation.
> >>And the changes made in this md5.h are precisely the ones
> >>that fixed our 64 bits problem that we have with the previous
> >>(old) version (I've just checked this).
> >>It means that we could "make up" a stdint.h file, but it
> >>will still be specific ;
> >>And do you know if the other C routines compile normally ?
> >>
> >>>can we please have an mpack that compile on all unix systems?
> >>I you have one, please go and check-in !
> >>
> >>Jean-Michel
> >>
> >>On Tue, Mar 11, 2008 at 09:27:44AM +0100, Martin Losch wrote:
> >>>when I comment out #include getopt.h
> >>>then unixpk.c and unixunpk.c compile but now on Solaris:
> >>>>gcc -DPACKAGE=\"mpack\" -DVERSION=\"1.6\" -DHAVE_DN_EXPAND=1 -
> >>>>DHAVE_STRCHR=1 -DHAVE_STRCASECMP=1 -DHAVE_GETOPT=1  -I. -I.     - 
> >>>>g -
> >>>>O2 -c codes.c
> >>>>In file included from codes.c:42:
> >>>>md5.h:37: stdint.h: No such file or directory
> >>>>*** Error code 1
> >>>>make: Fatal error: Command failed for target `codes.o'
> >>>and on uname -a
> >>>AIX edvir1 2 5 0010117A4C00
> >>>the mpack build completes without errors.
> >>>
> >>>can we please have an mpack that compile on all unix systems?
> >>>
> >>>Martin
> >>>
> >>>
> >>>On 10 Mar 2008, at 22:51, Jean-Michel Campin wrote:
> >>>
> >>>>Hi Martin,
> >>>>
> >>>>Ed's suggestion is good to try. If it works, I would be
> >>>>interested in knowing if an empty getopt.h in the mpack-1.6 dir
> >>>>is still fixing it (with the original unixpk.c & unixunpk.c).
> >>>>If nothing works, I m ready to add an option to testreport
> >>>>to specify the path of the mpack executable, and then you
> >>>>could use the old version (after the fix of Mon Feb 25,
> >>>>with cvs co -D "2008-02-25 23:00" you should get it)
> >>>>to build it somewhere once, and then to use this one.
> >>>>
> >>>>Cheers,
> >>>>Jean-Michel
> >>>>
> >>>>On Sun, Mar 09, 2008 at 11:45:30PM -0400, Ed Hill wrote:
> >>>>>On Sun, 9 Mar 2008 21:36:39 +0100 Martin Losch wrote:
> >>>>>
> >>>>>>>checking for getopt... yes
> >>>>>>>updating cache ./config.cache
> >>>>>>>creating ./config.status
> >>>>>>>creating Makefile
> >>>>>>>       gcc -DPACKAGE=\"mpack\" -DVERSION=\"1.6\" -
> >>>>>>>DHAVE_DN_EXPAND=1 -DHAVE_STRCHR=1 -DHAVE_STRCASECMP=1 -
> >>>>>>>DHAVE_GETOPT=1  -I. -I.      -g -O2 -c unixpk.c
> >>>>>>>unixpk.c:29:20: getopt.h: No such file or directory
> >>>>>>>make: The error code from the last command is 1.
> >>>>>>>
> >>>>>>>
> >>>>>>>Stop.
> >>>>>>
> >>>>>>
> >>>>>>I don't seem to have getopt.h on the system:
> >>>>>>>edvir1.awi-bremerhaven.de::mpack-1.6> locate getopt.h
> >>>>>>>locate: /opt/freeware/var/locatedb: No such file or directory
> >>>>>>
> >>>>>>
> >>>>>>I have no idea, what's that supposed to mean.
> >>>>>
> >>>>>
> >>>>>Hi Martin,
> >>>>>
> >>>>>The getopt() call is part of the POSIX standard and on  
> >>>>>Slowlaris and
> >>>>>other old Unices it is supposed to be supplied by the headers:
> >>>>>
> >>>>> #include <unistd.h>
> >>>>> #include <stdlib.h>
> >>>>> #include <stdio.h>
> >>>>>
> >>>>>so there is (again, supposedly) no need for the getopt.h  
> >>>>>header.  So,
> >>>>>please try commenting out the two lines matching:
> >>>>>
> >>>>> #include <getopt.h>
> >>>>>
> >>>>>in the files:
> >>>>>
> >>>>> unixpk.c
> >>>>> unixunpk.c
> >>>>>
> >>>>>and see if it helps.  If it does, then we can probably add a few
> >>>>>lines
> >>>>>to the "configure" script that will fix it for your machines.
> >>>>>
> >>>>>Ed
> >>>>>
> >>>>>-- 
> >>>>>Edward H. Hill III, PhD  |  ed at eh3.com  |  http://eh3.com/
> >>>>
> >>>>
> >>>>
> >>>>>_______________________________________________
> >>>>>MITgcm-devel mailing list
> >>>>>MITgcm-devel at mitgcm.org
> >>>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>>
> >>>>_______________________________________________
> >>>>MITgcm-devel mailing list
> >>>>MITgcm-devel at mitgcm.org
> >>>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>>
> >>>_______________________________________________
> >>>MITgcm-devel mailing list
> >>>MITgcm-devel at mitgcm.org
> >>>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >>_______________________________________________
> >>MITgcm-devel mailing list
> >>MITgcm-devel at mitgcm.org
> >>http://mitgcm.org/mailman/listinfo/mitgcm-devel
> >_______________________________________________
> >MITgcm-devel mailing list
> >MITgcm-devel at mitgcm.org
> >http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel