After upgrading lcms2 from 2.12-3.fc36 to 2.13-1.fc36, t/tiff.t fails like this:

t/viewer-preferences.t ...... ok
#   Failed test 'alpha + flate'
#   at t/tiff.t line 164.
#          got: '# ImageMagick pixel enumeration: 1000,100,255,graya
# 0,0: (0,0,0,255)  #000000FF  graya(0,1)
[...]
# 999,99: (255,255,255,255)  #FFFFFFFF  graya(255,1)
# '
#   Failed test 'alpha + lzw'
#   at t/tiff.t line 300.
#          got: '# ImageMagick pixel enumeration: 1000,100,255,graya
# 0,0: (0,0,0,255)  #000000FF  graya(0,1)
[...]
# 999,99: (255,255,255,255)  #FFFFFFFF  graya(255,1)
# '
t/tiff.t .................... 
Dubious, test returned 2 (wstat 512, 0x200)
Failed 2/19 subtests 
	(less 1 skipped subtest: 16 okay)
Test Summary Report
-------------------
t/tiff.t                  (Wstat: 512 Tests: 19 Failed: 2)
  Failed tests:  9, 14
  Non-zero exit status: 2

There seems to be changes in lmcs-2.13.0 regarding alpha in TIFF:

2.13 Featured release
-----------------------
Added support for premultiplied alpha
tifficc can now handle alpha channels, both unassociated and premultiplied
Better documentation
CGATS parser can now deal with very long strings
Added Projects for Visual Studio 2020
Travis CI discontinued, GitHub actions used instead
Added a very preliminar meson build script (thanks to xclaesse)
Added ARM64 target to visual studio 2019 (thanks to gaborkertesz-linaro)
Added thread safe code to get time
Added automatic linear space detection 
Added cmsGetStageContextID function
Added cmsDetectRGBProfileGamma function
configure now accepts --without-fastfloat to turn plugin off
autogen.sh has now a --distclean toggle to get rid of all autotools generated files
Checked to work on STM32 Cortex-A, Cortex-M families
Bug & typos fixing (thanks to many reporters and contributors)