On Nov 9, 2008, at 11:11 PM, Bryan Blackburn wrote:
On Mon, Nov 10, 2008 at 01:33:25AM -0500, William Davis said:On Nov 10, 2008, at 12:19 AM, Joshua Root wrote:William Davis wrote:I'll try one more time: Ryan made a change that will cause portindexto be updated all at once by building it in a temp file and then overwritting the old file all at once. THEREFOREthere will be no problem with sections of portindex being locked whensomeone runs selfupdate. THEREFORE unless there are other constrains, it should now be possible to update portindex more often without end-users having problems.Server-side, the new PortIndex is only committed to svn after it is finished being regenerated, so an end user could never get a partial index from rsync. The issue with there being a partial index was only present when theindex in a local ports tree was rebuilt. The one advantage to end users of updating the PortIndex more often would be that new ports would beavailable sooner. - JoshNo doubt you are quite right. Still I wonder where those error mesgssaying something about port foo being updated I sometimes get insertedinto the middle of the file list as its copyed down come from......How often do you see these, and do you have an example? Perhaps you have caught it during the time it's syncing between subversion and what the rsyncserver offers.But! getting updates more often was indeed the point......I think portindex currently takes 7-8 minutes to run, and probably puts a load on the server, so that would be one limiting issue on just how often itis run.
The index makes no noticeable impact on server load. I can run it as often as you want. I havnt timed it on the server lately, but assuming its < 30m, then I can just run it in sync with the rsync server updates which are already every 30m.
-Bill
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ macports-dev mailing list [email protected] http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
