[Debian-med-packaging] Bug#724711: insighttoolkit4: Drops architecture support

Edmund Grimley Evans edmund.grimley.evans at gmail.com
Fri Jun 26 10:13:49 UTC 2015


> However, I make no distinction: software that "builds" but doesn't run
> correctly is no more valuable than software that doesn't build.

Are you sure that the software is not running correctly on other
architectures? It could just be that some tests are buggy. They often
are, in my experience.

Can you point out a particular test that you believe is good and
indicate how to run that test?

On arm64 the build takes a couple of days and generates a 600 MB log
that mentions "fail" in 336 places (grep -ic fail) but it's not
immediately obvious to me exactly what's failing.

I note that there are lots of what appear to be floating-point numbers
in the log. It is fairly difficult to write portable tests for
floating-point calculations so I would be rather surprised if the test
suite is completely portable.


You might want to suggest this "patch" for upstream:

find -type f -exec \
perl -i -pe 's/char/signed char/g if /complex\s*<\s*char\s*>/;' {} \;

It didn't noticeably improve the test results on arm64 but it did get
rid of 6002 warnings similar to this one:

cTraits.h:913:56: warning: comparison is always false due to limited
range of data type [-Wtype-limits]
   static bool IsNegative(Self val) { return val.real() < 0; }

Three comparison warnings remained:

ageMetricv4Test.cxx:528:29: warning: comparison is always true due to
limited range of data type [-Wtype-limits]
             useMovingFilter >= 0; useMovingFilter-- )

itkImageToImageMetricv4Test.cxx:531:32: warning: comparison is always
true due to limited range of data type [-Wtype-limits]
                 useFixedFilter >= 0; useFixedFilter-- )

TestsCXX.cxx:390:41: warning: comparison is always false due to
limited range of data type [-Wtype-limits]
   return (*reinterpret_cast<char*>(&uc) < 0)?0:1;

They are probably easy to fix.


When 4.7.2 has migrated to testing, will you be reenabling builds on
other architectures? People are more likely to look at the failures if
there are results to look at at
https://buildd.debian.org/status/package.php?p=insighttoolkit4&suite=sid



More information about the Debian-med-packaging mailing list