Replace ftime() by clock_gettime()

[mreininghaus]

ftime() is considered obsolete in the POSIX standard and not available anymore on some platforms, e.g. OpenBSD. Therefore, I replaced it by clock_gettime(), however keeping ftime() as fallback.

[coveralls]

Coverage decreased (-0.02%) to 49.492% when pulling 42762cb on mreininghaus:ftime-clock_gettime into cc3cbd0 on vermaseren:master.

[coveralls]

Coverage decreased (-0.03%) to 49.475% when pulling b47169d on mreininghaus:ftime-clock_gettime into e900894 on vermaseren:master.

[tueda]

Thank you for your pull request! I totally agree with you that ftime() is obsolete and it should be replaced at some point. Some comments:

• Probably checking clock_gettime() in the configure is needed. Some system may require linking to rt and this is the reason why the CI failed. Once the check is done via configure, _POSIX_TIMERS is replaced with HAVE_CLOCK_GETTIME or something like that. For reference, my attempt for gettimeofday() is here.
• Because this is a set of rather small patches, would you mind squashing them into one commit and rebasing to the master such that fast-forward merge would work?

[mreininghaus]

Thank you for your comments. I'll implement your suggestions.

[tueda]

@mreininghaus Sorry that I didn't look into this for quite a while.

Do you still want to edit your commit history? Especially your 3 merge commits nonlinearly complicates the commit graph.

If you don't mind, I could git-cherry-pick your commits without the merge commits. The result will be like https://github.com/tueda/form/commits/mreininghaus-ftime-clock_gettime and this can be merged into the master branch.

[mreininghaus]

Feel free to do that. Sorry if I messed things up a bit.

[tueda]

OK, I will try to merge it in that way. Thanks for your contribution!