v1.3.4 test suite fails

[umlaeute]

while updating the Debian package for libmysofa (well: attempting to update), I noticed that the test suite fails.

I can reproduce this with a local build of the v1.3.4 tag, which yields:

$ make test
[...]
68% tests passed, 13 tests failed out of 41

Total Test time (real) =  47.01 sec

The following tests FAILED:
	  5 - H20_44K_16bit_256tap_FIR_SOFA (Failed)
	  6 - MIT_KEMAR_large_pinna (Failed)
	  7 - MIT_KEMAR_normal_pinna (Failed)
	  8 - MIT_KEMAR_normal_pinna.old (Failed)
	  9 - dtf_nh2 (Failed)
	 10 - hrtf_c_nh898 (Failed)
	 32 - CIPIC_subject_003_hrir_final (Failed)
	 33 - FHK_HRIR_L2354 (Failed)
	 35 - Pulse (Failed)
	 36 - Tester (Failed)
	 37 - TU-Berlin_QU_KEMAR_anechoic_radius_0.5_1_2_3_m (Failed)
	 38 - TU-Berlin_QU_KEMAR_anechoic_radius_0.5m (Failed)
	 41 - example_dummy_sofa48_with_user_defined_variable (Failed)
Errors while running CTest

for more info see my LastTest.log

i haven't attempted to track this down yet.

[umlaeute]

hmm, it also fails with v1.3.2 (which I'm sure, used to work), so maybe this is just some missing dependencies.
(I guess the build system should detect missing deps and eventually skip some tests)

[umlaeute]

ah ok. the fogs are starting to lift...

the failures with v1.3.2 were related to not being able to find the mysofa2json binary in my out-of-tree builds, which was actually (partly) fixed with d8dc278

by passing the path to mysofa2json to the tests/notcrashed.sh script as well, I could make the test-suite pass with current main (PR upcoming)

HOWEVER the v1.3.4 tag (or current fix-main branch) still fails the mentioned tests.

[umlaeute]

@hoene since the only difference between current main (3f4cb66) and v1.3.4 (7a0c071) is the "optimized code of btree loop" (a1c0440), I conclude that this optimization broke the code

[umlaeute]

and just as an additional datapoint: the H20_44K_16bit_256tap_FIR_SOFA test extracts 15MB worth of JSON data from the .sofa file, whereas the reference file extracts to 26MB of data.

it seems that most differences are because of printing with lower precision... (seems more like an off-by-one error somewhere)