Hi Ken, Thank you for having a look at the canonicalize problem (on which I'm not yet convinced that it's a bug in Cygwin - it could also be a problem on Gnulib's side).
When I look at the config.cache file on Cygwin 2.9, and in particular the *_works variable whose value is 'no', I see a number of issues that could be worth fixing in Cygwin. You find the test case in the corresponding *.m4 file. * chown.m4 gl_cv_func_chown_ctime_works=no * fchmodat.m4 gl_cv_func_fchmodat_works=no * fcntl.m4 gl_cv_func_fcntl_f_dupfd_works=no * fma.m4 gl_cv_func_fma_works=no * fmaf.m4 gl_cv_func_fmaf_works=no * fmal.m4 gl_cv_func_fmal_works=no * frexpl.m4 gl_cv_func_frexpl_works=no * getdtablesize.m4 gl_cv_func_getdtablesize_works=no * ilogbl.m4 gl_cv_func_ilogbl_works=no * log2.m4 gl_cv_func_log2_works=no * log2f.m4 gl_cv_func_log2f_works=no * pwrite.m4 gl_cv_func_pwrite_works=no * strtod.m4 gl_cv_func_strtod_works=no * strfmon_l.m4 gl_cv_strfmon_l_works=no * uchar.m4 gl_cv_type_char16_t_works=no gl_cv_type_char32_t_works=no There are surely more issues, listed in *.m4 files or in the documentation, but it's harder to get a list of. Bruno