Showing example pounders/formquad failure

[jmlarson1]

I was doing some test and encountered an example problem that failed on develop. I see that it also fails on main.

When nf = 489 we enter a model-improving step and get an error:

Traceback (most recent call last):
  File "/home/jlarson/research/poptus/IBCDFO/pounders/py/tests/test_showing_failure.py", line 45, in <module>
    [X, F, flag, xkin] = pdrs.pounders(Ffun, X0, n, npmax, nfmax, gtol, delta, nfs, 1, F0, xind, Low, Upp, printf, spsolver, hfun, combinemodels)
  File "/home/jlarson/research/poptus/IBCDFO/ibcdfo_pypkg/src/ibcdfo/pounders/pounders.py", line 289, in pounders
    Hres = Hres + Hresdel
ValueError: operands could not be broadcast together with shapes (4,4,1) (0,0) 

I don't know why we haven't encountered this before. But you can see that it happens in the GitHub Actions:
https://github.com/POptUS/IBCDFO/actions/runs/9488351840/job/26147086251?pr=149#step:5:2979

[coveralls]

Pull Request Test Coverage Report for Build 9488161893

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 96.855%

---

Totals
Change from base Build 7744656838:	0.0%
Covered Lines:	1386
Relevant Lines:	1431

---

💛 - Coveralls

[coveralls]

Pull Request Test Coverage Report for Build 9488250299

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 96.855%

---

Totals
Change from base Build 7744656838:	0.0%
Covered Lines:	1386
Relevant Lines:	1431

---

💛 - Coveralls

[coveralls]

Pull Request Test Coverage Report for Build 9488342748

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+3.1%) to 100.0%

---

Totals
Change from base Build 7744656838:	3.1%
Covered Lines:	0
Relevant Lines:	0

---

💛 - Coveralls

[jmlarson1]

@wildsm If I take this python example directly to matlab, the crash/error is not present. The iterates (or function evals) differ enough.

Instead, if I save all of the inputs to the last call to formquad.py in pounders.py before, and then load this in matlab and re-create the calls to formquad.m, I'm able to recreate the issue (in R2023b at least).

Running in matlab produces the following:

>> cd ~/research/poptus/IBCDFO/pounders/m/
>> test_formquad
Error using  + 
Arrays have incompatible sizes for this operation.

Error in test_formquad (line 14)
    Hres = Hres + Hresdel;

So, like in python, the matlab can produce an empty Hresdel.

[wildsm]

@wildsm If I take this python example directly to matlab, the crash/error is not present. The iterates (or function evals) differ enough.

Instead, if I save all of the inputs to the last call to formquad.py in pounders.py before, and then load this in matlab and re-create the calls to formquad.m, I'm able to recreate the issue (in R2023b at least).

Running in matlab produces the following:

>> cd ~/research/poptus/IBCDFO/pounders/m/
>> test_formquad
Error using  + 
Arrays have incompatible sizes for this operation.

Error in test_formquad (line 14)
    Hres = Hres + Hresdel;

So, like in python, the matlab can produce an empty Hresdel.

I can recreate the above in R2023a, thank you!

I will look into a solution.

[jmlarson1]

@wildsm I know that the first test_formquad failure didn't provide enough context to debug.

I have now saved the state of every variable at the top of the last loop of pounders.py. You can now load that into matlab and see the model/Res/D values in matlab, which I have done in the example larger_test_formquad.m. If you run in matlab, you should see:

>> cd ~/research/poptus/IBCDFO/pounders/m/
>> larger_test_formquad
Error using  + 
Arrays have incompatible sizes for this operation.

Error in larger_test_formquad (line 230)
            Hres = Hres + Hresdel;