From 59929e19ae6c22e148154fbabd8863281aa21743 Mon Sep 17 00:00:00 2001 From: Joey Hess Date: Wed, 5 Apr 2017 11:41:59 -0400 Subject: comment --- ...8_a6c330f5ad2f64d86d8a24c23c49115a._comment.swp | Bin 0 -> 12288 bytes ...ent_8_a6c330f5ad2f64d86d8a24c23c49115a._comment | 21 +++++++++++++++++++++ 2 files changed, 21 insertions(+) create mode 100644 doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/.comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment.swp create mode 100644 doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment (limited to 'doc') diff --git a/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/.comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment.swp b/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/.comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment.swp new file mode 100644 index 000000000..24597d8f0 Binary files /dev/null and b/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/.comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment.swp differ diff --git a/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment b/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment new file mode 100644 index 000000000..5ea68b685 --- /dev/null +++ b/doc/bugs/add_fails_with_v6_repo_when_four_levels_deep/comment_8_a6c330f5ad2f64d86d8a24c23c49115a._comment @@ -0,0 +1,21 @@ +[[!comment format=mdwn + username="joey" + subject="""comment 8""" + date="2017-04-05T15:26:59Z" + content=""" +git-annex does use CoW in a few situations; it does so by running `cp +--reflink` and letting it use CoW features when available. However, +I don't see where `git annex add` would use that, and of course I don't see +why that feature would break for you even if it did run cp that way. + +git also uses CoW, at least it uses `mmap` with `MAP_PRIVATE`. I'm not +clear if/how that involves the filesystem layer. + +This remains puzzling, but knowing it's limited to btrfs on Arch Linux +with CoW is certianly a good start. + +It seems like a bug in btrfs would not be out of the question. +Trying some different kernel versions might be useful. + +It would perhaps be useful to get `strace -ff` logs. +"""]] -- cgit v1.2.3