|
From: | Aaron Bentley |
Subject: | Re: [Gnu-arch-users] Re: Sharing a library between multiple developers |
Date: | Thu, 08 Apr 2004 00:26:32 -0400 |
User-agent: | Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.6b) Gecko/20031205 Thunderbird/0.4 |
Andrew Suffield wrote:
On Wed, Apr 07, 2004 at 12:44:24PM -0700, Tom Lord wrote:> From: Matthieu Moy <address@hidden> > address@hidden (Julian T. J. Midgley) writes: > > In article <address@hidden>, > > Aaron Bentley <address@hidden> wrote: > >>Julian T. J. Midgley wrote: > >>> Are any problems likely to result from allowing many developers to> >>> share a single tla library?> A good question here is:> Is tla robust to concurent calls to 'tla library-add' ?I believe so. Bugs are bugs but, so far as I know, it's robust.FSVO "robust"; IIRC, there is at least a chance that one of the tla processes will return an error, which isn't really ideal and means you need external locking if you're going to automate it. It just won't mangle the revision library in the process.
You mean for the case where two developers try to add the same revision concurrently? It seems as though careful error handling could be applied here:
First, the revision is built in a temp directory, then renamed.If there's an error, and the status is EBUSY, ENOTEMPTY or EEXIST, then the requested revision has been added between checking for the library's presence and the rename, so we can just nuke the temp directory.
Concurrent library-remove and get could be problematic, though. :-) Aaron
[Prev in Thread] | Current Thread | [Next in Thread] |