Please describe the problem.
Using git annex add while in git-annex branch would make git-annex add to annex not to git.
There might be cases where direct manipulation within git-annex branch is desired. (in my case was to workaround missing yt: prefix). Then I wanted to use generic tools (sed + datalad run) but was surprised that we got content added to annex while within git-annex branch. I do not see when/why that could potentially be useful (but I might be wrong?!)
Sure could be argued to be "operator error" but it is more of question of assumptions and automations -- should all tools around git-annex guard for that?
I think annex add should avoid annexing, and just do git add while under its dedicated git-annex branch -- after all it would only be git-annex which would know how special this (or any other) branch for it.
What steps will reproduce the problem?
❯ git clone -b git-annex https://github.com/OpenNeuroDatasets/ds000003; builtin cd ds000003; echo 123 > 123; git annex add 123; git commit -m 123 123; git show
Cloning into 'ds000003'...
remote: Enumerating objects: 1121, done.
remote: Counting objects: 100% (29/29), done.
remote: Compressing objects: 100% (19/19), done.
remote: Total 1121 (delta 18), reused 10 (delta 10), pack-reused 1092 (from 1)
Receiving objects: 100% (1121/1121), 92.35 KiB | 331.00 KiB/s, done.
Resolving deltas: 100% (223/223), done.
Remote origin not usable by git-annex; setting annex-ignore
https://github.com/OpenNeuroDatasets/ds000003/config download failed: Not Found
add 123
ok
(recording state in git...)
[git-annex cbbce8c] 123
1 file changed, 1 insertion(+)
create mode 120000 123
commit cbbce8c1d594f9e675eb6784111ee2e7926bc6ec (HEAD -> git-annex)
Author: Yaroslav Halchenko <debian@onerussian.com>
Date: Fri Mar 28 07:21:14 2025 -0400
123
diff --git a/123 b/123
new file mode 120000
index 0000000..0f13084
--- /dev/null
+++ b/123
@@ -0,0 +1 @@
+.git/annex/objects/G6/qW/SHA256E-s4--181210f8f9c779c26da1d9b2075bde0127302ee0e3fca38c9a83f5b1dd8e5d3b/SHA256E-s4--181210f8f9c779c26da1d9b2075bde0127302ee0e3fca38c9a83f5b1dd8e5d3b
What version of git-annex are you using? On what operating system?
❯ git annex version
git-annex version: 10.20250115
...
Adding a large file to git just because the git-annex branch is currently checked out seems like it would be a large footbomb. That is generally harder to recover from than adding a file to the annex and then realizing it needs to be added to git instead.
Since git generally allows switching branches with new files staged. It would be entirely reasonable to check out the git-annex branch after adding a new annexed file but before committing it. And checking out the git-annex branch,
git-annex addof a large file without committing it, then switching back to the main branch and committing there is also possible if someone wants to do that for some reason.Since manual commits to the git-annex branch need extra steps anyway (eg removing .git/annex/index or committing using it instead of the usual index file), I don't see much point in refining it.