From: Junio C Hamano <hidden> Date: 2016-08-11 20:45:52
Shawn Pearce [off-list ref] writes:
Does anyone on the mailing list really have an objection to having
reflogs on by default?
When you talk about potential breakage for existing users, you
should not be asking people on THIS list. You instead should
talk with or at least think about people on linux-kernel, x.org
and wine people, and possibly others. git is maturing, and we
cannot expect that most of the users are paying attention to
what is happening on this list anymore.
I 100% agree that it makes sense to have reflog enabled for a
repository with an associated worktree. I would say that we do
not even need it to be conditional on the configuration variable
for such a repository.
My answer to your question is:
kernel.org:/pub/scm/
I would REALLY be worried to have reflog enabled at a public
distribution point where the only ways the owners interact with
it daily are 'git push' and 'git pull'. As you mentioned, there
is one extra potential receive-pack failure, and in general it
is one more thing that can go wrong, and hard to notice breakage
because it is on the other side of the connection.
Worse yet, there is no easy way to garbage collect. Even in an
end-user repository with a worktree, the only way to garbage
collect older reflog entries is to edit the reflog files to
remove the top part.
Maybe a check to say if $GIT_DIR is ".git" or ends with "/.git"
then enable it and otherwise honor the configuration variable,
without changing the default in the code (with your patch) nor
in the default configuration ("enable for new repositories" as I
suggested) might be a workable compromise.
This function does not really determine if the repo is bare. I have no
better name for it, though.
guess_if_bare_git_dir ?
I struggled to name that thing because it can't really tell, its just
guessing... but it is going to be right most of the time. Of course
I'm sure there's some Git user somewhere who will confuse it.
--
From: J. Bruce Fields <hidden> Date: 2016-08-11 19:18:47
On Thu, Dec 14, 2006 at 06:18:32PM -0500, Shawn Pearce wrote:
"J. Bruce Fields" [off-list ref] wrote:
quoted
Stupid question--I assume a mention in the reflog doesn't count as a
real reference to an object, so they won't save you in the case when you
pruned recently?
Not a stupid question. Your assumption is correct, its not a real
reference, so prune will remove things that the log mentions but
that refs don't currently mention.
OK, thanks. So we just need to make sure that's documented someplace.
Maybe a check to say if $GIT_DIR is ".git" or ends with "/.git"
then enable it and otherwise honor the configuration variable,
without changing the default in the code (with your patch) nor
in the default configuration ("enable for new repositories" as I
suggested) might be a workable compromise.
See my latest patch. Though that patch also sets the value in the
config file, much as core.filemode is also set in the config file,
based on the guess determined by init-db at the time it was executed.
--
This function does not really determine if the repo is bare. I have no
better name for it, though.
guess_if_bare_git_dir ?
I struggled to name that thing because it can't really tell, its just
guessing... but it is going to be right most of the time. Of course
I'm sure there's some Git user somewhere who will confuse it.
I think the name is fine, but probably a comment in front would
help unconfuse people.
/* Does it look like a repository without a working tree? */
Unfortunately there currently are public bare repositories that
have index under them because they were primed by rsync from
developers' working repositories. I do not think it is
unreasonable to persuade owners of them to drop index -- then we
could use absence of $GIT_DIR/index as a strong clue that the
repository is bare.
From: Shawn O. Pearce <hidden> Date: 2016-08-11 20:27:41
New and experienced Git users alike are finding out too late that
they forgot to enable reflogs in the current repository, and cannot
use the information stored within it to recover from an incorrectly
entered command such as `git reset --hard HEAD^^^` when they really
meant HEAD^^ (aka HEAD~2).
So enable reflogs by default in all future versions of Git, unless
the user specifically disables it with:
[core]
logAllRefUpdates = false
in their .git/config or ~/.gitconfig.
We only enable reflogs in repositories that have a working directory
associated with them, as shared/bare repositories do not have
an easy means to prune away old log entries, or may fail logging
entirely if the user's gecos information is not valid during a push.
This heuristic was suggested on the mailing list by Junio.
Documentation was also updated to indicate the new default behavior.
We probably should start to teach usuing the reflog to recover
from mistakes in some of the tutorial material, as new users are
likely to make a few along the way and will feel better knowing
they can recover from them quickly and easily, without fsck-objects'
lost+found features.
Signed-off-by: Shawn O. Pearce <redacted>
---
Documentation/config.txt | 7 +++++--
builtin-init-db.c | 4 ++++
cache.h | 1 +
environment.c | 9 +++++++++
4 files changed, 19 insertions(+), 2 deletions(-)
@@ -79,8 +79,11 @@ core.logAllRefUpdates:: file is automatically created for branch heads. This information can be used to determine what commit- was the tip of a branch "2 days ago". This value is- false by default (no automated creation of log files).+ was the tip of a branch "2 days ago".++ This value is true by default in a repository that has+ a working directory associated with it, and false by+ default in a bare repository. core.repositoryFormatVersion:: Internal variable identifying the repository format and layout
@@ -239,6 +239,10 @@ static void create_default_files(const char *git_dir, const char *template_path)git_config_set("core.filemode",filemode?"true":"false");}++/* Enable logAllRefUpdates if a working tree is attached */+git_config_set("core.logallrefupdates",+!is_bare_git_dir(git_dir)?"true":"false");}staticconstcharinit_db_usage[]=
On Thu, Dec 14, 2006 at 05:41:17PM -0500, Shawn O. Pearce wrote:
quoted
New and experienced Git users alike are finding out too late that
they forgot to enable reflogs in the current repository, and cannot
use the information stored within it to recover from an incorrectly
entered command such as `git reset --hard HEAD^^^` when they really
meant HEAD^^ (aka HEAD~2).
Stupid question--I assume a mention in the reflog doesn't count as a
real reference to an object, so they won't save you in the case when you
pruned recently?
Not a stupid question. Your assumption is correct, its not a real
reference, so prune will remove things that the log mentions but
that refs don't currently mention.
--
From: J. Bruce Fields <hidden> Date: 2016-08-11 20:46:38
On Thu, Dec 14, 2006 at 05:41:17PM -0500, Shawn O. Pearce wrote:
New and experienced Git users alike are finding out too late that
they forgot to enable reflogs in the current repository, and cannot
use the information stored within it to recover from an incorrectly
entered command such as `git reset --hard HEAD^^^` when they really
meant HEAD^^ (aka HEAD~2).
Stupid question--I assume a mention in the reflog doesn't count as a
real reference to an object, so they won't save you in the case when you
pruned recently?