Hi Akim, There are three issues with this patch:
1) Portability. As listed in https://www.gnu.org/software/gnulib/manual/html_node/clock_005fgettime.html Gnulib provides no replacement for clock_gettime. So, you'll probably need gettimeofday or module 'gettime' as fallback. For clock_gettime, you can use the Gnulib module 'clock-time' for the m4 macros. 2) Precision. When you use gettimeofday, the starting point is 1970, thus the number of microseconds stored in the 'wall' field will be around 1.6e15, thus when stored as 'float' the precision will be in the range of 3 to 10 minutes. A program which runs for less than a minute will have wall clock difference of 0 with high probability. => The 'wall' field should better be a 'double'. Or possibly solve both problems at once by using the Gnulib module 'gethrxtime', and 'xtime_t' instead of 'float' for the 'wall' field. Or 'gettime'. I don't know how 'gettime' compares against 'gethrxtime'. 3) Link dependency: List the link dependencies, like it's done e.g. in the 'clock-time', 'gettime', and 'gethrxtime' modules. Also use these dependencies when linking the test program (in module 'timevar-tests'). Bruno
