cplcore/cpl_init.c | 9 0 9 - 0 !
1 file changed, 9 deletions(-)

 remove wcs versioninfo

Since wcslib does not provide a version string at runtime, CPL
included the WCSLIB version number at compile time. This is misleading
when the wcslib is updated afterwards. To omit confusion, this patch
removes the WCSLIB version number from the CPL version string.

cplcore/cpl_init.c | 18 0 18 - 0 !
1 file changed, 18 deletions(-)

 remove cfitsio version check

ESO is paranoid against the binary incompability of cfitsio and checks
that the version at runtime equals the compile time version. Since the
libcfitsio3 is binary compatible up to now, this check is
obsolete. For Debian, runtime dependency checks should be put into the
install dependency anyway instead of checking it at runtime.

cplcore/cpl_test.c | 2 1 1 - 0 !
cpldrs/cpl_wcs.c | 2 1 1 - 0 !
cpldrs/tests/cpl_wcs-test.c | 2 1 1 - 0 !
3 files changed, 3 insertions( ), 3 deletions(-)

 use standard paths for libraries

Check cfitsio, wcs, and fftw in the standard paths if they are not
mentioned as configure packages. For wcslib, also the #include
wcslib.h directives were changed so that the software can be compiled
without additional -I flags of the compiler if wcslib is installed in
the standard directory.

Makefile.am | 3 1 2 - 0 !
1 file changed, 1 insertion( ), 2 deletions(-)

 remove private library libltdl and use the installed one.

cpldrs/tests/cpl_fit-test.c | 2 1 1 - 0 !
1 file changed, 1 insertion( ), 1 deletion(-)

 loose tests

On several machines -- even 64 bit -- the fit test may fail because of
a significant loss of precision in
cpl_fit_imagelist_polynomial_window(). Since the fit accuracy is not
defined anyway, we will use the loose requirement here for all
architectures.

configure.ac | 1 0 1 - 0 !
1 file changed, 1 deletion(-)

 remove cpu check at compile time

On Linux machines, CPL_CHECK_CPU checks the CPU 2nd level cache during
compile time for some obscure code optimizations. In a distribution,
the produced code should be independent of the build environment, so
we use the default here which works well for non-Linux platforms. The
performance loss is probably negleglible.

cplcore/tests/Makefile.am | 2 1 1 - 0 !
cpldfs/tests/Makefile.am | 2 1 1 - 0 !
cpldrs/tests/Makefile.am | 2 1 1 - 0 !
cplui/tests/Makefile.am | 2 1 1 - 0 !
libcext/tests/Makefile.am | 2 1 1 - 0 !
5 files changed, 5 insertions( ), 5 deletions(-)

 force serial-tests to get better verbose output

cplcore/cpl_test.c | 9 0 9 - 0 !
1 file changed, 9 deletions(-)

 remove timestamp macros from source for repoducible builds

cplcore/cpl_imagelist_basic.c | 4 4 0 - 0 !
1 file changed, 4 insertions( )

 work around mips compiler optimization problem

Switch off optimization for cpl_imagelist_basic.c on MIPS There is a
loop optimization problem around cpl_imagelist_basic_body.h:111

cpldrs/cpl_ppm.c | 2 1 1 - 0 !
1 file changed, 1 insertion( ), 1 deletion(-)

 fix angle tolerance in i386

Reason is (IMO) that in `cpl_ppm_match_points()`, outliers are detected with
(`cpl_ppm.c`. line 1757):
```

cplcore/cpl_mask.c | 2 1 1 - 0 !
1 file changed, 1 insertion( ), 1 deletion(-)

 don't run a loop which may cause unaligned access (sigbus)

This happens on armhf and sparc64 (and sparc, which is not a Debian
arch anymore)