Odd behaviour for random files

[Kevin Bullock]

(replies inline, top-posting is discouraged on this list)

On May 17, 2011, at 2:31 AM, Jan Olsen wrote:

> Unfortunately we run Mercurial over a cross country network (~30Mbit), so our commits takes a "tad" longer :).

Network speed makes no difference to commits. Commits are local. Unless you're running a networked -filesystem- over a cross-country 30Mbit network...

Are you referring to the speed of -pushing- rather than committing?

> We used to run Mercurial over a 1GB network and that time we never experienced any odd behaviour. Now on the 30Mbit we have experienced it twice.
> 
> A few days ago when I had to run the "hg update -r 1336" there was a problem with locked files in the repository, so I had to shut down five other computers, before it could finish the update. That was cumbersome but not a problem since it was during the weekend. During weekdays I'm afraid it wont be possible at all.


This would indicate to me that you're using a shared working directory on some kind of network filesystem. That's probably a bad idea—not because of Mercurial, but because of the inherent limitations of network filesystems, which DVCSes are designed to avoid.

The usual working pattern with Mercurial is for each developer to have their own local clone, commit locally, pull changes from each other (or push to a central repo), and merge each developer's changes early and often.

But perhaps a little more information on your current workflow could help us give you more specific advice?

pacem in terris / mir / shanti / salaam / heiwa
Kevin R. Bullock