[Nut-upsdev] reposurgeon progress [.gitignore]

[Charles Lepple]

On Jan 9, 2012, at 7:55 AM, Charles Lepple wrote:
> you wrote:
>> Hmmm.  I instrumented reposurgeon to show all instances of .gitignore
>> generation, and I'm not spotting any that are obviously wrong to me.
> 
> Working backwards, then: are you seeing the symptom on your end? (That is, all of the .gitignore files have the same content - where the one under drivers/, for instance, should have a list of the driver executables that won't be present in other directories). The smallest interval I narrowed the regression down to in the history of reposurgeon is 2.0-pre4 to 2.0-pre5, but I did not start looking at the differences in the configuration yet.

I don't completely understand the way blobs are created and added to the fast-import stream, but compare these two chunks.

from 2.0-pre4:

[...]
M 100644 :1633 .cvsignore
M 100644 :1635 .gitignore
M 100644 :1634 clients/.cvsignore
M 100644 :1636 clients/.gitignore
[another commit]
M 100644 :1638 common/.cvsignore
M 100644 :1648 common/.gitignore
M 100644 :1639 conf/.cvsignore
M 100644 :1649 conf/.gitignore
M 100644 :1640 data/.cvsignore
M 100644 :1650 data/.gitignore
M 100644 :1641 data/html/.cvsignore
M 100644 :1651 data/html/.gitignore
M 100644 :1642 drivers/.cvsignore
M 100644 :1652 drivers/.gitignore

:1635 and :1636 contain the respective (different) svn:ignore properties for each of the directories.

from 2.0-pre7:

M 100644 :1498 .cvsignore
M 100644 :1500 .gitignore
M 100644 :1499 clients/.cvsignore
M 100644 :1500 clients/.gitignore
[another commit]
M 100644 :1502 common/.cvsignore
M 100644 :1500 common/.gitignore
M 100644 :1502 conf/.cvsignore
M 100644 :1500 conf/.gitignore
M 100644 :1502 data/.cvsignore
M 100644 :1500 data/.gitignore
M 100644 :1502 data/html/.cvsignore
M 100644 :1500 data/html/.gitignore
M 100644 :1503 drivers/.cvsignore

Note that the post-pre4 behavior matches what I see in the repository: namely, each .gitignore file contains the contents of an old .gitignore that only makes sense in the root directory. File blobs with marks other than :1500 and :1502 are created in the later fast-import stream (including the updated svn:ignore contents), but are not referenced by the subsequent commit blobs.

I'm stumped as to how the changes between 2.0-pre4 and -pre5 would lead to this issue, though. 

-- 
Charles Lepple
clepple at gmail