Mike Gran wrote: > It makes me think that if Guile wants to use gnulib 'time' it also needs to > use 'gettimeofday'. is that correct?
Yes, and that is why there is a module dependency from gnulib module 'time' to gnulib module 'gettimeofday' (at least when REPLACE_TIME = 1, which is the case on mingw (see m4/time.m4)). But to be sure to not make assumptions about the current state of the module dependencies, you are welcome to add module 'gettimeofday' to the imports for Guile. (For the moment, it will not make a difference on mingw, but maybe in the future.) > Currently GNU Guile seems to use a plain AC_CHECK_FUNCS for > gettimeofday Once you have added module 'gettimeofday' to the imports, you can drop this AC_CHECK_FUNCS invocation and proceed as if $ac_cv_func_gettimeofday is true always — since the 'gettimeofday' module is documented [1] to always provide a gettimeofday() function. > but it also uses the gnulib 'time' which make a replacement > sys/time.h including rpl_timeval. > > On MinGW UCRT, this seems to be in conflict. MinGW's gettimeofday > is used directly How can MinGW's gettimeofday be used directly? You are supposed to #include <sys/time.h>, and to have -I options that make sure that the gnulib-generated sys/time.h gets precedence over the system's one [2]. Bruno [1] https://www.gnu.org/software/gnulib/manual/html_node/gettimeofday.html [2] https://www.gnu.org/software/gnulib/manual/html_node/_002dI-options.html
