ros-time gives different timestamp than rosccp's ros::Time::now about roslisp

[gaya-]

Or should we consider wrapping the c-functions for accessing the sys-time?

If you just want the equivalent of C system time in SBCL (you mean the one that comes from the operating system in milliseconds from the epoch, is that correct?) try the (sb-ext:get-time-of-day) from the unix API.

My guess is that, originally, roslisp developers tried to keep it as Lisp-compiler-independent as possible, so they used general Common Lisp functions and most of them work on seconds neglecting milliseconds, so you have to get the milliseconds from the run-time and who knows if it creates an overhead etc.

[airballking]

Oh, that is a cool hint! I was not aware of that functionality.

Then the question is: Do we embrace such an sbcl-specific fix of (ros-time)?

[gaya-]

My guess might also be wrong, because, now that I look closer, right inside the same (ros-time) function there is a call to (mutex-owner *debug-stream-lock*) which is an SBCL-specific function. So, if you try the get-time-of-day and it turns out to work better than get-universal-time I wouldn't object to using it instead...

[moesenle]

Right. roslisp is already very SBCL specific. Porting it to other Lisp
compilers would be quite some effort. I would just go ahead and use
whatever SBCL function fixes the problem.

Btw. if you don't want to change the official release, you can also think
about monkey-patching roslisp.

On Thu, Feb 12, 2015 at 3:17 PM, Gayane Kazhoyan [email protected]
wrote:

My guess might also be wrong, because, now that I look closer, right
inside the same (ros-time) function there is a call to (mutex-owner
debug-stream-lock) which is an SBCL-specific function. So, if you try
the get-time-of-day and it turns out to work better than
get-universal-time I wouldn't object to using it instead...

—
Reply to this email directly or view it on GitHub
#22 (comment).

[gaya-]

@airballking Any update on this? I've noticed problems with cl-tf timestampes before, if sbcl time function solves the problem that would be nice. Have you tested this yet, Georg?

[airballking]

@gaya- I meant to but haven't got around to it. I'm prioritizing hardcore because of the upcoming review. Maybe I feel like checking this out over night.

But you are right, this could be the source of "all those error in cl-tf:lookup-transform". I had not even considered this!

[gaya-]

#24 should solve this.

[airballking]

Fixed through Gaya's pull request.