mirrors/git-annex
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages 1 Wiki Activity Actions

restage: New git-annex command, handles restaging unlocked files

Browse source 
This is much easier and less failure-prone than having the user run
git update-index --refresh themselves.

Sponsored-by: Dartmouth College's DANDI project
This commit is contained in:
Joey Hess 2022-09-23 16:29:28 -04:00
parent f7146c153b
commit 2478e9e03a
No known key found for this signature in database
GPG key ID: DB12DB0FF05F8F38
10 changed files with 110 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

4
doc/bugs/reports_file___34__modified__34___whenever_it_is_not.mdwn
View file

@ -91,4 +91,6 @@ I think I get it after I `annex move` and then `annex get` that file back. Just
[[!meta author=yoh]]
[[!tag projects/dandi]]
> [[!meta title="annex.stalldetection prevents git-annex get from restaging unlocked files"]]
[[!meta title="annex.stalldetection prevents git-annex get from restaging unlocked files"]]
> [[fixed|done]] --[[Joey]]

33
doc/bugs/reports_file___34__modified__34___whenever_it_is_not/comment_16_4b8a360c38dfa4c8d22e3a96fbace0e5._comment Normal file
View file

@ -0,0 +1,33 @@
[[!comment format=mdwn
username="joey"
subject="""status update"""
date="2022-09-23T19:57:38Z"
content="""
I've implemented the log file. The stalled transferrer case is now handled.
This bug is fixed.
As to a few other cases I considered in comments upthread:
When a get/drop was interrupted before it could restage,
the next get/drop will cause the necessary restaging for the
interrupted process to happen. However, this doesn't help if there's
nothing left to get/drop. Should git-annex always run restagePointerFiles
on shutdown? That would make any git-annex command handle the restaging.
But it doesn't seem right for query commands to do potentially a lot of
work to handle this case. Anyway, I don't think this needs to be dealt
with in this bug report.
When multiple processes try to restage at the same time, one will
restage everything that all of them logged. The others will still display a
warning to the user that they couldn't restage. It would be hard to avoid
displaying that warning, since it does need to warn when it was
unable to restage because git has the index locked at the time. Anyway,
I think it's ok to display the message despite the files having been
restaged, because it's the same as a later git-annex process handling the
restaging. (It does seem like two transferrers belonging to the same parent
could collide in this way, and one display the warning, which isn't great..)
I also implemented a "git-annex restage" command that
is an easier way to restage in the cases where git-annex is not able
to do it itself.
"""]]