comment
This commit is contained in:
parent
3c5280b1cf
commit
d4fc506f27
1 changed files with 20 additions and 0 deletions
|
@ -0,0 +1,20 @@
|
||||||
|
[[!comment format=mdwn
|
||||||
|
username="joey"
|
||||||
|
subject="""comment 23"""
|
||||||
|
date="2021-07-29T16:30:46Z"
|
||||||
|
content="""
|
||||||
|
There are probably multiple causes. v8 upgrade does seem likely to be the
|
||||||
|
main one, but I see that adswa seems to have found a way to reproduce it from
|
||||||
|
scratch:
|
||||||
|
<https://github.com/datalad/datalad/issues/5750#issuecomment-888819347>
|
||||||
|
It's looking like perhaps inode instability in their case, due to use
|
||||||
|
of a networked filesystem in a cluster.
|
||||||
|
|
||||||
|
I have done some detailed auditing, and I'm pretty sure git-annex doesn't
|
||||||
|
just forget to record the inode in some case. Well, except for potentially
|
||||||
|
one case involving git-annex lock and annex.thin, which I fixed but was not
|
||||||
|
able to actually quite get to reproduce the problem.
|
||||||
|
|
||||||
|
(I did find a some potential races, that would cause other problems but not
|
||||||
|
this one, so the auditing was good for something..)
|
||||||
|
"""]]
|
Loading…
Reference in a new issue