Richard Braun rbraun at
Wed Apr 11 15:08:27 UTC 2012

On Wed, Apr 11, 2012 at 09:17:01AM -0400, Rich Felker wrote:
> This is something of an aside, but I'd be really interested in knowing
> what it's needed for. A number of folks go around asserting that
> CLOCK_MONOTONIC is not really monotonic on Linux, but according to the
> documentation it is monotonic and the only thing non-"raw" about it is
> that it can run slightly faster or slower when ntpd smoothly adjusts
> for clock error. In particular, it ignores discontinuous clock
> changes. If there's a bug and it sometimes is discontinuous, my
> feeling is that the CLOCK_MONOTONIC constant should just be remapped
> to CLOCK_MONOTONIC_RAW rather than leaving the standard clock broken
> and forcing applications to use a nonstandard one...

I was considering this clock because of the weird behaviour I observed
with CLOCK_MONOTONIC. But this is probably caused by a bug fixed in the
kernel (commit 8aa3149405e33cec4f866cfe7f92c2b40d259613). Your
description of the difference between the two clocks is right, and it's
not what my problem is about, so I can't answer you as to why the _RAW
clock is needed for, sorry.

Richard Braun

More information about the uClibc mailing list