Forums : Technical Issue Help

I added my project today out of curiousity. In the svn-repository, there is 12-15 years of commits, but for some reason ohloh does only recognise the last month's commit. One reason may be that there was a directory change about at that time, but all directory changes are done with svn-move and, so svn know where things are done.

Any clue what I can do about it? Project is https://www.ohloh.net/p/opendtect

Kristofer Tingdahl over 11 years ago

Kristofer,

I see around 1907 commits and I'm not sure how many of these made changes in trunk. If ohloh is right, it should be 618 of them. I'll need to look closer at the repository to see.

Standby.

ssnow-blackduck over 11 years ago

Kristofer,

We may count commits differently. I wonder if we count only commits that touch recognizable source code? Hmmm. Will take a look. Total commits in trunk IS approximately 1907 but how many we actually count is the issue, I guess. Let me compare repository log to our commit log and see what I can determine.

Thanks!

ssnow-blackduck over 11 years ago

Total commits are around 36000. Since trunk was moved, I guess that is what screws it up. The history goes like this:

1. It was sitting under /svn/trunk up to September 2013.
2. It was then moved to /svn/versiondep/trunk
3. In early april 2014, it was moved back to /svn/trunk

I can live with this, but it screwed up the stats a bit.

• K

Kristofer Tingdahl over 11 years ago

Kristofer,

Looks like r33963 is the earliest commit we show. It was a massive one. The previous commit was Moved branches, tags & trunk from versiondep. It seems we aren't exploring any farther back than that.

Let me know what you think.

Thanks!

ssnow-blackduck over 11 years ago

Kristofer,

I wonder if you enlisted the entire subversion and then used ignore requests to kill off branches and tags etc. if it would do any better. The problem is that it might take forever to process the initial fetch because the ignore order doesn't count on the first fetch but only on the updates after that. I can try it on an experimental project if you'd like and see how long it will take.

Second issue: subversion is not a stellar performer for us - it's one step above CVS but that's not saying much... Could be a long time.

Thanks!

ssnow-blackduck over 11 years ago

Hello,

I'd leave it now, as I only have to live with that my project is younger than it is; we seem to be extremely productive ;-)

I may run experiments later on, perhaps trunking the revisions between september and april, or to try a git-mirror.

Thanks for looking into this.

Cheers,

Kristofer

Kristofer Tingdahl over 11 years ago

Kristofer,

Ping me here if you think of anything you'd like to try.

Thanks!

ssnow-blackduck over 11 years ago

Hi,

I re-submitted the URL last weekend with a slightly different url, and it loaded roughly 30000 revisions fine. The analysis does however look weird, as we clearly have contributors and increasing code base between 2005 and 2014.

When I loaded the new URL, it went to step 3 of 3, and then step 3 took long time until revision 6000-something. Then it gave an error message, and then restarted a couple of hours later and went on quite quickly.

Kristofer Tingdahl over 11 years ago

Kristofer,

The error message you saw was akin to committing suicide which is our cheeky way to indicate that the process wasn't getting attention and it's programmed to quit so a new process will take over. As you can see it did finally finish OK with 26680 total commits. You've also had one successful update which added 9 commits. Analysis is now at May 27, 2014 on code collected on May 27, 2014 and last commit is at May 26, 2014 - about 13 hours ago.

I'd examine the commit logs that we show and see if you can track down the source of the commit-free years. It certainly does look peculiar. It's possible that the current enlistment is leaving out some part of the repository that was significant for development for that period.

Let me know what you find.

Thanks!

ssnow-blackduck over 11 years ago

Assuming that the progress until 2005 was until the scan killed itself, it did a good job. At my last attempt, the nr of lines now should be around 6-700k, so I think something went wrong when it restarted.

Kristofer Tingdahl over 11 years ago

Kristofer,

I have initiated a re-count of the project since the commit counts are somewhat suspicious (import says 31762 commits and the interrupted count phase only got 26680 commits.)

Normally a restart proceeds without any data loss but this could be the exception. If this one finishes and is still suspicious, I'll run a re-fetch which takes about 3 days to finish, it seems.

Thanks!

ssnow-blackduck over 11 years ago

Kristofer,

Re-count provided 31770 commits. Please look it over and let me know if a re-fetch is warranted.

Thanks!

ssnow-blackduck over 11 years ago

Dear Support,

I have just gone through a major cleanup of history to make it easier for openhub to read it in. Would it be possible to re-fetch and reset the stats you currently have?

Thanks,

Kristofer Tingddahl

Kristofer Tingdahl almost 11 years ago

Kristofer,

We have a problem... The enlistment was failing after a LONG fetch with a Bad Gateway error. I'm trying a different enlistment with http:// transport instead of the secure version. Maybe it will work?

Thanks!

ssnow-blackduck almost 11 years ago

Thank you. I assume you ar now using http://opendtect.googlecode.com/svn/trunk

If that fails, I have an alternative location, but I would avoid that one if I could as I do not wish to mantain it.

thanks for your effort!

Kristofer Tingdahl almost 11 years ago

Hi,

after the troubles with the goolgecode repos, I changed to my own one (svn.opendtect.org/od/trunk), and it seemed to work well, until it failed after about a week of fetching. Any clue what went wrong?

I can work with either my own repos, or make another try at http://opendtect.googlecode.com/svn/trunk

But before I take action, it would be good to understand what went wrong?

Cheers,

Kristofer

Kristofer Tingdahl almost 11 years ago

Kristofer,

Fetch died after 2628/5034 commits were retrieved. Strictly an internal failure here at OpenHUB. I have rescheduled it and will keep an eye on it.

Thanks!

ssnow-blackduck almost 11 years ago

Kristofer,

Analyzed on Friday, March 06, 2015 @ 20:36:18PM UTC based on code collected on Thursday, March 05, 2015.

Thanks!

ssnow-blackduck almost 11 years ago