[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20030606: 20030605: Quick question on compiling gempak 5.6.j.

Stonie,

I introduced the .INIT in the make file rewrites a couple of 
releases ago....but its easy enough to back out of that...
in only the top level makefile. It was just supposed to
be easier than making the all: dependent on the target that
.INIT uses.

To test gmake, you probably want to use the -v and -n flags
to look at the environment being used by the program.
It sounds as if the NA_OS variable isn't in the environment 
so that the last line of Makeinc.common doesn't find
Makeinc.linux. If thats the case, and you are only building
on that one platform, I would probably change that last line to 
include $(NAWIPS)/config/Makeinc.linux
and see if that succeeds.

Make sure you have sourced $NAWIPS/Gemenviron before building
and see if the script is bailing early for some reason.

Steve Chiswell



>From: "Stonie R. Cooper" <address@hidden>
>Organization: Planetary Data, Incorporated
>Keywords: 200306061753.h56HrXLd022854

>Chiz,
>
>Still working through this (these) issues - and wonder if you have ever seen 
>this in your multiple OS dealings.
>
>First, the prerequisite of .INIT is not getting hit.  The first make failure 
>is because there is no $NAWIPS/lib/linux to stick the object files into.  I 
>can run "make .INIT", and it works - but that's not how .INIT, .DONE, and all 
>the implicit prerequisites are supposed to work.  This is gmake 3.8, gcc 3.3.
>
>The second thing - it doesn't appear that the Makeinc.common is getting called
>  
>correctly from the make . . . the Makeinc.linux is never invoked, thus why I 
>ended up with a bunch of big endian binaries.
>
>Again, this is on SuSE 8.2; I started with RedHat 9, and after the rpm 
>database folded in on itself after the first boot, I scrapped that POS with 
>the rest of RH's failings since 6.0.
>
>The purpose for all of this is because our-used-to-be-business-partner, SCO, 
>is committing corporate suicide, and we're bailing.
>
>Thanks for any pointers . . . 
>
>Stonie
>
>On Thursday 05 June 2003 17:18, Steve Chiswell wrote:
>> Stonie,
>>
>> The endian-ness is defined for the OPSYS type "Linux",
>> so your Makeinc.${NA_OS} must have "OPSYS = Linux".
>> That will allow the appropriate MCHPRM.Linux include file to
>> be found which sets MTMACH =  MTLNUX.
>>
>> The dm library code checks for MTLNUX and other little endian machine
>> types.
>>
>> Steve
>>
>> On Thu, 5 Jun 2003, Stonie R. Cooper wrote:
>> > Steve,
>> >
>> > Playing with a new Linux distro (SuSE 8.2), and the make did something
>> > weird with gempak 5.6.j - trying to debug; almost everything compiled
>> > without incident (except Garp), but everything was compiled big endian. 
>> > So, I have gui and interface - but no data, including maps, can be
>> > displayed.
>> >
>> > What is the magical macro that sets the compile to be little endian?  I
>> > also noticed the .INIT did not run, either - so had to make the install
>> > dirs by hand.  This is with gmake 3.80 and gcc 3.3.  Thanks!
>> >
>> > Stonie
>