Re: Aw: News spin of repository conversion

[Daniel J S. <dan...@ie...>]

On 10/21/2017 02:35 AM, "Bastian Märkisch" wrote:
> 
>> Gesendet: Freitag, 20. Oktober 2017 um 23:01 Uhr
>> Von: "Eric S. Raymond" <es...@th...>
>> An: gnu...@li...
>> Betreff: News spin of repository conversion
>>
>> This one is 7a2323d15193540b226632f7b11ace3c524c63e3
>>
>> It was lifted using Daniel's enhanced algorithm for mining ChangeLog
>> files.
> 
> Unfortunately there still seems to be something slightly odd about the algorithm:
> Ethan commited only a few changes by me in the beginning, yet I get:
>    git log --committer=Ethan --author=Bastian --pretty="%H;%an;%cn;%cd;%s"  | wc -l
>    132.
> I can also not remember ever having comitted a change by Ethan:
>    git log --committer=Bastian --author=Ethan --pretty="%H;%an;%cn;%cd;%s"  | wc -l
>    154
> I also suspect that the 53 commits by EAM authored by HBB are "false positives".
> 
> Looking at some of these commits I see two cases where the algorithm fails:
> 1) A new ChangeLog entry without header is inserted. The algorithm seems to pick
>     the wrong header from the ChangeLog.
>     I thought that the backward search would identify these correctly.

Here's an example that Bastian is referring to.  In gitg:

Ethan A M <xxxxxx@xxxxxx>
07/24/2017 12:00:02 PM +0000
Committed by: Bastian M <xxxxxx@xxxxxx>
07/27/2017 11:13:43 AM +0200

And the expanded diff hunk in the ChangeLog (the 
first-change-search-back rule should hold):

sebald@ git-main $> git diff --unified=30 
f0736a1c9285dc882528e55ed749ba6b2ac59ba0 
e4e9a7594de9fc8a62f14d0dd4938f73aa781bb6
diff --git a/ChangeLog b/ChangeLog
index f46f85c..78f705d 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,35 +1,37 @@
  2017-07-27  Bastian Maerkisch  <bma...@we...>

         * src/win/wd2d.cpp:  Resize the swap chain buffers instead of
         recreating the swap chain when the window size changes.

+       * src/win/wgraph.c term/win.trm:  Default to Direct2D backend.
+
  2017-07-24  Ethan A Merritt  <merritt@u.washington.edu>

         * src/datafile.c (df_open):  Reject plot command if input and 
output
         both use the same data block.  Prevents memory corruption / 
segfault.

  2017-07-24  Bastian Maerkisch  <bma...@we...>


> 2) There have been additional corrections to / additions of previous ChangeLog entries
>     in the same commit.
>     Those are maybe hard to detect automatically.

Right.  Or at least too much programming detail for little gain.  I 
suspect there aren't too many of these cases.


> So it might be worth looking at the algorithm again. If the script is available, I
> can try to help tweaking it. Otherwise, we would have to come up with a list of
> manual corrections. In any case I think we have a lot of work to do to verify the
> attributions.

I'd say no manual correction for that.  The pool to search through is 
all entries, 6000+.  I was thinking the only manual adjustment might be 
the substitution of "empty log message" as one of the last steps. 
That's only 300 entries, much more reasonable to walk through.

Dan



> 
>    Bastian
> 
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> gnuplot-beta mailing list
> gnu...@li...
> Membership management via: https://lists.sourceforge.net/lists/listinfo/gnuplot-beta
> 


-- 

Dan Sebald
email: daniel(DOT)sebald(AT)ieee(DOT)org
URL: http://www(DOT)dansebald(DOT)com