Error: Automatic array 'shifts' at (1) cannot have an initia

[mark.tolman]

I'm attempting to compile ADL 4.1.1 on Fedora 16. I'm getting errors when it gets to Wave_monitor_tc.f. Here's an example of the errors:

Code: Select all

Wave_monitor_tc.f:292.28:

         errshifts(maxlines), err_f(max_nspec_ccd), err_parm(64), &
                            1
Error: Automatic array 'errshifts' at (1) cannot have an initializer
Wave_monitor_tc.f:313.56:

       INTEGER (KIND=INT16) :: history_day(nwave_trends), &
                                                        1
Error: Automatic array 'history_day' at (1) cannot have an initializer
Wave_monitor_tc.f:297.44:

         history_scale(nwave_trends,maxctpx), &
                                            1
Error: Automatic array 'history_scale' at (1) cannot have an initializer

I have both the adl41IdpsAlg_p1.patch and SSEC_ADL4_1_1.patch patches applied. It is using gfortran 4.6.3. Is that too new a version? I was able to compile the same code using gfortran 4.4.6 on a CentOS server. Is there something else I'm missing?

Mark Tolman
MIT Lincoln Laboratory

[mark.tolman]

I found a posting that suggested adding the "-fno-whole-file" flag when using 4.6.x. Where does one do that in ADL?

I'm going to try editing PRO_Config - line 399.

[mark.tolman]

That didn't work. I'm still getting the error. Any thoughts?

[kbisanz]

Yes, PRO_Config around 399 would be the appropriate place. Or you could temporarly edit the Makefile directly just to see if it fixes your error. It sounds like it did not.

Raytheon only officially supports GCC 4.4. Unfortunately, with strict security policies, it's a nightmare for us to try out different compilers, so we do not have 4.6.x installed any place.

I believe the University of Wisconsin got ADL to build with 4.5 and those changes have been incorporate into the baseline.

I poked around google and found this page: http://gcc.1065356.n5.nabble.com/Bug-fo ... 36982.html

The poster indicated that removing -finit-local-zero allowed the code to compile. The flag -finit-local-zero is specified in $ADL_HOME/SDR/OMPS/total_column/src/Imakefile. If you're just trying to get OMPS to build, but don't plan on running it, it's probably ok to remove that. However, if you plan on running OMPS, I do NOT recommend removing -finit-local-zero. The reason is that some OMPS code has variables which are not initialized by the code before being used. The flag -finit-local-zero initializes them automatically. If they are not initialized automatically, they'll be set to some random value that happened to be in memory. This will change the way the code works. It's been a problem for IDPS. The fix was to use the compiler flag (-finit-local-zero or an AIX equivalent) to init things to 0.

[Kevin Howard]

hello,
I still have the same error.
Any other solutions?

thank you

[houchin]

You may want to consider installing the officially supported version of gcc/gfortran on your system. I have found that taking a short amount of time to use the officially supported versions saves a huge amount of other debugging time.

In the AeroADL distribution, I include the source for GCC 4.4.5, with GNU MP and GNU MPFR (two items not normally installed on typical CentOS and Scientific Linux systems) already bundled in. You can get it from subversion at:

https://svn.gravite.gov/aeroadl/trunk/s ... .5-src.tgz

There's also a build script for gcc, which includes unpacking and building at:

https://svn.gravite.gov/aeroadl/trunk/src/buildgcc.sh

More info on this is included in the AeroADL user documentation at:

https://svn.gravite.gov/aeroadl/trunk/src/doc

(You may need to translate those URLs depending on how you connect to the GRAVITE information portal. If you have a GRAVITE ICF account, AeroADL and all of the referenced files above are at:

/share/aero/aeroadl/latest

[scottm]

Good morning,

I was recently building ADL with the 4.7 compiler. The -finit-local-zero option continues to be a problem with OMPS. We routinely use the VIIRs, ATMS and CrIS SDR code with the 4.5 and 4.7 compiler. The products produced are tested to match the output of the IDPS. VIIRS EDR products are also produced and tested against IDPS.

[sara100]

thank you