GOCART develop crashing with out of bounds error after updates

[bena-nasa]

After the PR merged into develop today, when I tried using the gocart develop branch, it crashed with this out of bounds error when using the "debug" build:

forrtl: severe (408): fort: (3): Subscript #3 of the array DPFLI has value 0 which is less than the lower bound of 1
Image              PC                Routine            Line        Source
GEOSgcm.x          000000000F493DCF  Unknown               Unknown  Unknown
GEOSgcm.x          000000000146F764  gocart2g_process_        6302  GOCART2G_Process.F90
GEOSgcm.x          0000000001158F67  su2g_gridcompmod_        1078  SU2G_GridCompMod.F90
libesmf.so         00002AAAAB40A9E4  _ZN5ESMCI6FTable1     Unknown  Unknown

The odd thing is that if you examine the code, it seems like it should be protected from getting there

6299 if(LH .lt. 1) continue
6300
6301 do k = LH, km
6302 qls(k) = dpfli(i,j,k)/pdog(i,j,k)*rhoa(i,j,k)
6303 end do

I tried changing the continue to cycle, thinking that maybe it was that but same behaviour. Will continue investigating.

[tclune]

Next step is to either explicitly print the lbounds of the arrays involved or to run under the debugger to see the same.

Am guessing that dpfli might be an "edge" variable which starts at 0? Not that this would explain things.

Also - please edit your initial message above to include compiler + version. Am suspecting a compiler bug.

[bena-nasa]

Hmm, maybe the cycle did do, I think I was rebuilding in the optimized build. Just rebuilt the debugged build and put the cycles in and it runs. So either it was the continue or we have a Heisenbug. I take a fresh look at this in the morning.

[tclune]

Hmm - I'd not even noticed that it said continue rather than cycle. The former should have been a syntax error, so maybe you ran without a new executable until that was fixed?

[weiyuan-jiang]

@tclune @bena-nasa I might introduce this bug. Let me take a took. It may be "exit" instead of cycle

[bena-nasa]

Yep, there were 4 continue statements that should have been cycle. Will make PR to fix

[weiyuan-jiang]

@bena-nasa @tclune I am sorry I messed up with C++'s "continue". The four "continue" statement should be 'cycle'.

[tclune]

No problem. It is just unfortunate that it is not a syntax error. Usually such other-language constructs are caught by the compiler.

[mathomp4]

No problem. It is just unfortunate that it is not a syntax error. Usually such other-language constructs are caught by the compiler.

Now that I understand the issue (I totally missed it yesterday looking at it with Ben because I guess I was in Python land...where continue is the right thing), it is annoying this isn't caught by the compiler. But, well, per the Standard:

1 Execution of a CONTINUE statement has no effect.

Since CONTINUE literally does nothing, I guess you can have them anywhere as long as it doesn't have a label at the front!