[Debian-astro-maintainers] Bug#1105355: giza: FTBFS with make --shuffle=reverse: make[4]: *** [Makefile:1030: check_output_files] Error 1 shuffle=reverse

[Ole Streicher]

Control: tags -1 +patch

Hi Trupti,

on a first look your patch look right. However, could you make a pull 
request upstream, https://github.com/danieljprice/giza? That could make 
sure that we use the same change as upstream will, and we get another 
eye on it.

Cheers

Ole

Am 05.02.26 um 13:12 schrieb Trupti:
> 
> Hello,
> 
> I reproduced the FTBFS with GNU make shuffle mode (MAKEFLAGS=-- 
> shuffle=reverse).
> The failure happened because the PNG verification step was executed 
> before the test programs ran.
> The PNG files are created only when the tests are executed, but under 
> shuffle mode the check could run first due to missing ordering.
> 
> I fixed this by removing the custom check rule and moving the PNG 
> verification to check-local, with an explicit dependency on check-TESTS, 
> so the tests always run before the PNG check.
> 
> 
> --- giza-1.5.0.orig/test/F90/Makefile.am
> +++ giza-1.5.0/test/F90/Makefile.am
> @@ -10,13 +10,10 @@ TESTS = $(ftests)
>   check_PROGRAMS = $(ftests)
>   AM_DEFAULT_SOURCE_EXT = .f90
> 
> -# Ensure that the tests are built as part of the 'check' target
> -check: $(ftests) check_output_files
> -
>   test_output_files = giza-test.png giza-test-2D.png
> 
>   # Custom target to check for the existence of expected output files 
> from the Fortran tests
> -check_output_files: $(ftests)
> +check-local: check-TESTS
>          @echo "Checking for required png files..."
>          @for i in $(test_output_files); do \
>              if [ ! -s $$i ]; then \
> @@ -26,4 +23,6 @@ check_output_files: $(ftests)
>                  echo "$$i exists"; \
>              fi; \
>          done
> -       @echo "All required png files exist."
> \ No newline at end of file
> +       @echo "All required png files exist."
> +
> +
> 
> With this change, the package builds successfully under shuffle mode on 
> ppc64el
> 
> I have attached a patch that fixes the issue described above. I would 
> appreciate a review.
> 
> 
> 
> Thanks,
> Trupti