| Re: [stellation-res] New workspace code status, with news of a very obscure bug. |
On Thu, 2002-08-29 at 20:11, Florin Iucha wrote: > On Thu, Aug 29, 2002 at 01:50:59PM -0400, Dave Shields wrote: > > I just fixed a very obscure bug in the workspace code. The code was > > failing several tests in an intermittent fashion. After several hours > > of debugging, I learned that the resolution of File.lastModified() is > > only one second in Linux, and the new workspace code was so much > > faster that some tests were updating files in the same second as the > > checkout, so the change was not detected! Adding a 'sleep 1' at end > > of Checkout fixed the problem. > > > > I revisited, and largely rewrote, all the code related to signatures > > (two versions of an artifact are deemed to be different if and only if > > they have different signatures). The signature is now the MD5 checksum > > I am not sure about the second if. I know it is hard to come with two > different files that have the same MD5 but still... We've been doing signature based differences for a while. If you work out the odds, having a full MD5 hashcode incorrectly tell you that the files are the same is actually less likely than a random gamma ray hitting your processor causing data to get altered in a way that will cause invalid behavior. It's just below the threshhold of reasonable concern. -Mark -- Mark Craig Chu-Carroll, IBM T.J. Watson Research Center *** The Stellation project: Advanced SCM for Collaboration *** http://www.eclipse.org/stellation *** Work Email: mcc@xxxxxxxxxxxxxx ------- Personal Email: markcc@xxxxxxxxxxx
Attachment:
signature.asc
Description: This is a digitally signed message part