[Unison-hackers] unison and inotify principle

[Benjamin Pierce]

Right: this functionality needs to be finished -- my recollection,  
though, is that Alan and I had a pretty complete design sketched out  
and what was left was mostly writing the external watcher utilities  
(which should be small in each case, but will require a bit of  
familiarity with the OS's hooks for change notification).  Would be  
great to get this going!

     - B

On Feb 23, 2010, at 11:34 AM, Ryan Newton wrote:

> Hi all,
>
> I take it that no one is using this right now?  I notice I can send
> 'watch' to -repeat, but in the code there seems to be a dead end with
> the watcherfilename:
>
>  let watcherfilename = System.fspathFromString "" in
>  (* STOPPED HERE -- need to find the program using watcherosx  
> preference
>
> I do want to hack on this a bit once I actually get it building on OSX
> again ;-) ... (I'm having that same suite of problems with 64 bit
> builds).
>
> -Ryan
>
> On Sun, Jun 8, 2008 at 8:27 AM, Benjamin Pierce <bcpierce at cis.upenn.edu 
> > wrote:
>>> Concerning "portability", i think it will be really hard. There is  
>>> no
>>> POSIX way to poll this kind of event (even UNIXes doesn't have the
>>> same
>>> way to do this task). It will probably require to write a program
>>> per OS
>>> (BSD/Mac OS X/Linux/Win32).
>>
>> I didn't mean that exactly the same code would perform notification
>> on all architectures -- clearly the notification logic, low-level
>> APIs, etc. are going to have to work differently on different OSs.  I
>> meant that things have to be set up so that the Unison sources will
>> continue to *compile* on all architectures.
>>
>>     - Benjamin
>>
>>
>>
>>>
>> On Jun 8, 2008, at 5:37 AM, Sylvain Le Gall wrote:
>>> On 08-06-2008, Benjamin Pierce <bcpierce at cis.upenn.edu> wrote:
>>>> Hi Sylvain,
>>>>
>>>>> If the external program work well and is written using OCaml, is
>>>>> there
>>>>> anyway that the program gets included in unison itself after a
>>>>> certain
>>>>> amount of time (6 months working without problems)?
>>>>
>>>> Yes.  If it is written in OCaml in a portable way, then the  
>>>> simplest
>>>> may be simply to make it be part of Unison (to maintain the current
>>>> single-executable distribution model).  But I am also open to other
>>>> models where the notification program is written in another  
>>>> language
>>>> and not necessarily in a portable way.  At this point, we need to
>>>> experiment to see what's going to work best.
>>>>
>>>
>>> Full ack.
>>>
>>> The only thing i would like to underline, is that it should be  
>>> really
>>> great to have some kind of "driver" model where you can either use  
>>> an
>>> internal driver (integrated in unison) or an external one (using a
>>> program). The idea is to keep the driver external as long as  
>>> possible
>>> for testing and then integrate the one that can be integrated.
>>>
>>> Concerning "portability", i think it will be really hard. There is  
>>> no
>>> POSIX way to poll this kind of event (even UNIXes doesn't have the
>>> same
>>> way to do this task). It will probably require to write a program
>>> per OS
>>> (BSD/Mac OS X/Linux/Win32).
>>>
>>> I will probably try to write an inotify program for this. But i
>>> need to
>>> finish some other job before ;-)
>>>
>>> Regards,
>>> Sylvain Le Gall
>>>
>>> _______________________________________________
>>> Unison-hackers mailing list
>>> Unison-hackers at lists.seas.upenn.edu
>>> http://lists.seas.upenn.edu/mailman/listinfo/unison-hackers
>>
>> _______________________________________________
>> Unison-hackers mailing list
>> Unison-hackers at lists.seas.upenn.edu
>> http://lists.seas.upenn.edu/mailman/listinfo/unison-hackers
>>
> _______________________________________________
> Unison-hackers mailing list
> Unison-hackers at lists.seas.upenn.edu
> http://lists.seas.upenn.edu/mailman/listinfo/unison-hackers
>