CMake and Autotools builds broken when V2 API is disabled

[edhartnett]

Thanks @gsjaardema for bringing this to light. Greg submitted a fix in a PR, but it introduces a new header file which is not needed. config.h should have all the information needed to determine if V2 is built.

Here's the error:

nc_test4/CMakeFiles/nc_test4_tst_vars2.dir/build.make:62: recipe for target 'nc_test4/CMakeFiles/nc_test4_tst_vars2.dir/tst_vars2.c.o' failed
make[2]: *** [nc_test4/CMakeFiles/nc_test4_tst_vars2.dir/tst_vars2.c.o] Error 1
CMakeFiles/Makefile2:4131: recipe for target 'nc_test4/CMakeFiles/nc_test4_tst_vars2.dir/all' failed
make[1]: *** [nc_test4/CMakeFiles/nc_test4_tst_vars2.dir/all] Error 2
Makefile:162: recipe for target 'all' failed
make: *** [all] Error 2
Build step 'Execute shell' marked build as failure
Finished: FAILURE

The autotools build fails like this:

tst_vars2.c: In function ‘main’:
tst_vars2.c:568:14: warning: implicit declaration of function ‘nccreate’; did you mean ‘nc_create’? [-Wimplicit-function-declaration]
       ncid = nccreate(FILE_NAME, NC_NETCDF4);
              ^~~~~~~~
              nc_create
tst_vars2.c:571:18: warning: implicit declaration of function ‘ncdimdef’; did you mean ‘nc_redef’? [-Wimplicit-function-declaration]
       dimid[0] = ncdimdef(ncid, LAT, LAT_LEN);
                  ^~~~~~~~
                  nc_redef
tst_vars2.c:571:33: error: ‘LAT’ undeclared (first use in this function)
       dimid[0] = ncdimdef(ncid, LAT, LAT_LEN);
                                 ^~~
tst_vars2.c:571:33: note: each undeclared identifier is reported only once for each function it appears in
tst_vars2.c:571:38: error: ‘LAT_LEN’ undeclared (first use in this function); did you mean ‘DIM1_LEN’?
       dimid[0] = ncdimdef(ncid, LAT, LAT_LEN);
                                      ^~~~~~~
                                      DIM1_LEN
tst_vars2.c:572:33: error: ‘LON’ undeclared (first use in this function)
       dimid[1] = ncdimdef(ncid, LON, LON_LEN);
                                 ^~~
tst_vars2.c:572:38: error: ‘LON_LEN’ undeclared (first use in this function); did you mean ‘LAT_LEN’?
       dimid[1] = ncdimdef(ncid, LON, LON_LEN);
                                      ^~~~~~~
                                      LAT_LEN
tst_vars2.c:575:15: warning: implicit declaration of function ‘ncvardef’; did you mean ‘nc_redef’? [-Wimplicit-function-declaration]
       varid = ncvardef(ncid, CXX_VAR_NAME, NC_FLOAT, VAR_DIMS2, dimid);
               ^~~~~~~~
               nc_redef
tst_vars2.c:578:7: warning: implicit declaration of function ‘ncclose’; did you mean ‘nc_close’? [-Wimplicit-function-declaration]
       ncclose(ncid);
       ^~~~~~~
       nc_close
tst_vars2.c:581:14: warning: implicit declaration of function ‘ncopen’; did you mean ‘nc_open’? [-Wimplicit-function-declaration]
       ncid = ncopen(FILE_NAME, 0);
              ^~~~~~
              nc_open
tst_vars2.c:582:7: warning: implicit declaration of function ‘ncinquire’; did you mean ‘nc_inq_var’? [-Wimplicit-function-declaration]
       ncinquire (ncid, &ndims, &nvars, &natts, &recdim);
       ^~~~~~~~~
       nc_inq_var

The config.h in both cases does include this:

/* do not build the netCDF version 2 API */
#define NO_NETCDF_2 1

The macro name is awkward, but it's a historical oddity, so we tolerate it, like the Roman-style robes we make graduates wear.

Greg also very thoughtfully submitted a fix (PR #656), but it unnecessarily introduces the netcdf_meta.h file, which is not needed and should not be used in tests.

I have fixed this on my current branch by using NO_NETCDF_2 where necessary in tst_vars2.c. I have also added some disable-v2 builds to my CI system to prevent it breaking again. ;-)

I will put my branch up for merge sometime next week.

[edhartnett]

This is now fixed in master. Thanks @gsjaardema.