SVN->GIT link broken?

[John Ralls]

On Mar 26, 2013, at 8:43 AM, John Ralls <jralls at ceridwen.us> wrote:

> 
> On Mar 26, 2013, at 8:00 AM, Derek Atkins <warlord at MIT.EDU> wrote:
> 
>> John Ralls <jralls at ceridwen.us> writes:
>> 
>>> On Mar 25, 2013, at 8:39 AM, Derek Atkins <warlord at MIT.EDU> wrote:
>>> 
>>>> 
>>>> Perhaps I should set up a cron job to flush it out every 6 hours, just
>>>> in case something "missed"?
>>> 
>>> Can't hurt. Would it be reasonable to have a "kicker" like the one for
>>> ceridwen, but in reverse, so that if it doesn't work right away a dev
>>> could push it? If one makes a change in trunk and wants to backport it
>>> it's kind of a pain if Github isn't updated right away.
>> 
>> Well, I don't want to just have anyone able to kick it; it's a potential
>> DoS.  Right now the way it works is that I have a cron job that runs
>> every minute to see if there's been a change, and if so it runs the
>> convert + push.  The change is detected because SVN touches a magic
>> file.
>> 
>> One thing I could try is to see if there was an error during the push
>> and, if so, reset itself to re-run by re-touching the file.  The
>> question is how easily can I detect that something went wrong?
> 
> We'll (I'll ;-) ) have to add some error handling to the push_repo function in git-svn-mirror (you're still using that, right?) to make it die if the push fails. Then your cron script can just test it and re-touch if it fails.

Alternatively, push_repo could handle the error on its own with an exponentially-timed retry period. The only concern I have about that is that there's no way to tell the difference between a temporary outage and Github doing something that requires human intervention -- but I suppose that's also true of re-touching your sentinel file.

BTW, why a cron job instead of an SVN commit hook?

Regards,
John Ralls