Forums

Please check number of committers vs number of commits here: https://www.ohloh.net/p/tikiwiki/analyses/latest

Number of commits are pretty steady: 2007-09 -> 532 2007-10 -> 773 2007-11 -> 676 2007-12 -> 642 2008-01 -> 620 2008-02 -> 494 2008-03 -> 922 2008-04 -> 834 2008-05 -> 608 2008-06 -> 572 2008-07 -> 596 2008-08 -> 850

Yet, number of committers falls from 13 to 20 range to 5 months at 0, and 3 months at less than 3. Committers: 2007-09 -> 13 2007-10 -> 15 2007-11 -> 0 2007-12 -> 0 2008-01 -> 0 2008-02 -> 0 2008-03 -> 0 2008-04 -> 1 2008-05 -> 2 2008-06 -> 3 2008-07 -> 13 2008-08 -> 20

I suspect there is an issue with the data because 0 committers can't make all these commits during that 5 month period :-)

Any idea?

Thanks!

Marc Laporte

12 months ago

Marc,

Just looked at latest analysis which shows 7 identified committers and the rest lumped into "Other".

Let me know if this isn't correct etc.

Thanks!

ssnow-blackduck

11 months ago

An image is worth a thousand words: http://tiki.org/tiki-downloadwikiattachment.php?attId=801

M ;-)

Marc Laporte

11 months ago

Marc,

I completely missed that graph. That certainly is odd. Best I can offer, I can re-fetch the trunk or CVS repository and see if it clears up the issue. I'll look at the web interface first for the trunk and CVS enlistments just to see if anything jumps out at me during that period (Nov 2007 through Mar 2008).

Let me know if you object to the re-fetch and I'll start it in a few hours.

Thanks!

ssnow-blackduck

11 months ago

Marc,

It occurs to me after reviewing the subversion web interface: Could it be that only commits to the branches/BRANCH-1-10 or branches/1.10 directory took place during that period? It isn't covered by any enlistment and may not show in the committers graph (because it filters commits made only to enlisted territory) but does show in the commits graph (which may show an overall number unfiltered?)

ssnow-blackduck

11 months ago

Marc,

I do see a few commits to trunk in that vicinity but possibly to areas covered by the ignore order? Hmmm.

ssnow-blackduck

11 months ago

Hi!

Thanks for this.

1- No objection to a re-fetch

2- branches/1.10: Why would ohcount filter one but not the other?

Ideally, we would want it to count the commits in all branches, but only the LoC in trunk. Here are some related feature requests: https://www.ohloh.net/forums/3491/topics/6043 https://www.ohloh.net/forums/8/topics/385

3- ignore order is for external code that we include. I don't see how we could have 5 months just depending on external code :-) But to help investigation, maybe rerun without any ignore order and we'll see what happens?

Maybe we have some remaining UTF-8 issues in CVS or SVN? https://tiki.org/Ohloh#UTF-8enlistmentissue

Thanks!

M ;-)

Marc Laporte

11 months ago

Marc,

I'm only speculating so far... Great way to put myself up a tree.

1. I hesitate if no change can come from it...

2. Could be a bug; could be a feature. We wouldn't analyze commits in any branch that wasn't enlisted to record the committers' names/id's so no record. Commit counts is easy either way - maybe we didn't use the right choice?

3. Wouldn't need to re-fetch to re-analyze after removing the ignore order? Then again, not sure? Depends if we grabbed all that stuff but didn't analyze it?

3 1/2. The UTF-8 issues will stop the job in mid-stream and cause it to fail. No failed jobs I found...

Hmmm.

Probably will save the current ignore order for trunk then wipe it out and see what happens...

ssnow-blackduck

11 months ago

Marc,

The removal of the ignore order seems to have done nothing to clear up the dead zone Nov, 2007 through Mar, 2008. I can reinstate the ignore order when you say the word. I'm a little surprised we haven't seen an update since Friday when I removed the ignore order. Will need to research this and see if something is off-kilter.

Thanks!

ssnow-blackduck

11 months ago

"I can reinstate the ignore order when you say the word."

Please do any changes any time to investigate this issue.

Thanks!

Marc Laporte

11 months ago

Marc,

Thanks. I am out-of-town this week and will reinstate the ignore order today but for a proper investigation I may need to wait 'til I get home this weekend.

Thanks!

ssnow-blackduck

11 months ago

Marc,

Back home... (Not a real fan of buses...)

Almost all the commits I've viewed so far in the dead zone (Nov 1, 2007 thru Mar 31, 2008)(r9903 through r12288) were to 'branches/1.10/' with a very few to trunk. Clearly need to add this branch and also add the same? ignore order to this one and also a similar order tailored to the CVS repository as well, I think. Only preliminary data but you look and see if it makes sense to you.

Thanks!

ssnow-blackduck

11 months ago

Ok, I just added https://tikiwiki.svn.sourceforge.net/svnroot/tikiwiki/branches/1.10/ to try

Thanks!

Marc Laporte

11 months ago

Marc,

Cool!

The ignore order will probably need to be customized for 'branches/1.10/' since I see little to none of the directories referred to before.

Thanks!

ssnow-blackduck

11 months ago

Marc,

Apparently there was a UTF-8 issue with the 'branches/1.10/' enlistment which was tried 394 days ago with no joy. Let me know if I can help. The error log doesn't give me any help surrounding where the issue(s) are. I could try to get a full log of that repo and then scan through it by eye but would be hit-or-miss.

Let me know.

Thanks!

ssnow-blackduck

11 months ago

Ok, Jean-Marc (who fixed last time) will take a look. Thanks!

Marc Laporte

11 months ago

Jean-Marc cleaned up some potentially problematic entries. Can you try again?

In the past, I seem to remember that Abhay could give us hints of where (why) it crashed.

Thanks!

Marc Laporte

10 months ago

Marc,

As you've most certainly figured out, the 1.10 enlistment is progressing. I'll continue to monitor it and report back here when it finishes with the results.

Thanks!

ssnow-blackduck

10 months ago

Marc,

The 1.10 enlistment has finished. The remainder of the enlistments are 12 days old and should be updating soon.

Thanks!

ssnow-blackduck

10 months ago

Marc,

The update has run and finished OK about 7 hours ago. Analysis is at July 19, 2012 on code collected on July 18, 2012 and last commit is at July 19, 2012 about 7 hours ago.

Looking at the "Commits per Month" graph, the "dead zone" we talked about seems to be back in play. The commits shown during that period now seem to be on a par with the rest. Look things over and let me know if any additional attention is required.

Thanks!

ssnow-blackduck

10 months ago

So basically, we have not solved the problem and still have hundreds of commits made by 0 committers ;-)

Marc Laporte

10 months ago

Marc,

Okay, I see that you are right and the problem still exists. I will push the question up to management and see if someone can look at the raw data and from that see if they can guess at a cause.

(embarrassed face...)

Thanks!

ssnow-blackduck

10 months ago

Hi again!

I tried something and it looks like we may have a hint :-)

I removed the CVS enlistment and kept just SVN. We now have a negative number of LOCs and the number of contributors is at 0 from 2002-2008, while there are commits during that period.

See screenshots at: http://tiki.org/Ohloh#Numberofcommittersat0for5monthswhilethereareseveralhundredsofcommits

Thanks!

M ;-)

Marc Laporte

7 months ago

Marc,

Good detective work! Will pass on this information and we'll probably send out a bug-hunting party. I'm sure it's clear there should never be a negative number in LOC and it's also clear we need to discover why we're not picking up the commits before 2008. (Any possibility the commits before 2008 were from a conversion like cvs2svn? I have heard the conversion is sometimes iffy and there can be problems if the parameters aren't chosen just correctly.)

Thanks!

ssnow-blackduck

7 months ago

For the record, https://www.ohloh.net/p/smarty is also showing a negative value in LOC.

"Commits per Month" looks right, but "Contributors per Month" pre-2009 is all at 0.

Since a significant portion of older projects went from CVS to SVN, my hunch is that this will affect a significant number of projects, And thus, it will lead to quite a few new better, more representative stats / charts :-)

Thanks!

M ;-)

Marc Laporte

7 months ago

Post a Response