Failed "make check" on x86-64

[Torbjorn Granlund]

Alexander Kruppa <alexander.kruppa at mytum.de> writes:

  Alexander Kruppa wrote:
  > Hi,
  >
  > I tried to compile GMP 4.2 on a AMD Opteron 150, but "make check" fails.
  >

  The problem persisted with a stock gcc 3.4.6.

  The problem turned out to be that the test programs "t-sub" etc. are
  linked dynamically. Thus, at execution they are linked against a
  previously installed GMP dynamic library instead of the newly compiled
  GMP 4.2. This caused the state of the random number generator to be
  initalised incorrectly. A possible fix is to add the "-static" flag to
  the compilaton of the tests programs. A workaround is installing the new
  library first, and only then doing "make check".

This is interesting for two reasons.

First, I don't understand why the shell script "t-sub" doesn't
get the shared lib search path setup correctly.  Have you checked
the environment inherited into it, if you have LD_LIBRARY_PATH or
some variant of it set?

Please try to understand why this is happening.

Secondly, I am interested in why t-sub is incompatible with the
installed libgmp.  It sounds like you have already investigated
that.  Perhaps it is a benign incompatibility, i.e., the new
t-sub uses features not available in your installed libgmp?

--
Torbjörn