doc: git-push: clarify DESCRIPTION section & refspec definition #1964
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -19,31 +19,33 @@ SYNOPSIS | |
| DESCRIPTION | ||
| ----------- | ||
|
|
||
| Updates one or more branches, tags, or other references in a remote | ||
|
There was a problem hiding this comment. On the Git mailing list, Junio C Hamano wrote (reply to this): "Julia Evans via GitGitGadget" <[email protected]> writes:
> From: Julia Evans <[email protected]>
>
> Signed-off-by: Julia Evans <[email protected]>
>
> Be clearer about what we're describing ("which repository" instead of
> "what to push"), and start with a positive "try X, then Y, then Z"
> instead of a negative ("if X is not specified..").
Since I like this simple rule so much, if it is generally applicable
everywhere, I'd like to have it or a variant of it in one of our
developer facing documentation as a tip to write better
documentation.
> Signed-off-by: Julia Evans <[email protected]>
There is some funny ordering problem with the commit log body and
sign-off.
> ---
> Documentation/git-push.adoc | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/Documentation/git-push.adoc b/Documentation/git-push.adoc
> index e73b64f61fd0..5c934486c33d 100644
> --- a/Documentation/git-push.adoc
> +++ b/Documentation/git-push.adoc
> @@ -22,10 +22,10 @@ DESCRIPTION
> Updates one or more branches, tags, or other references in a remote
> repository from your local repository.
>
> -When the command line does not specify where to push with the
> -`<repository>` argument, `branch.*.remote` configuration for the
> -current branch is consulted to determine where to push. If the
> -configuration is missing, it defaults to 'origin'.
> +To decide which repository to push to, Git uses the `<repository>`
> +argument (for example `git push dev`), then if that's not specified the
> +`branch.*.remote` configuration for the current branch, and then defaults
> +to `origin`.
Very nicely done.
> When the command line does not specify what to push with `<refspec>...`
> arguments or `--all`, `--mirror`, `--tags` options, the command finds |
||
| repository from your local repository. | ||
|
|
||
| To decide which repository to push to, Git uses the `<repository>` | ||
| argument (for example `git push dev`), then if that's not specified the | ||
| `branch.*.remote` configuration for the current branch, and then defaults | ||
|
There was a problem hiding this comment. On the Git mailing list, Junio C Hamano wrote (reply to this): "Julia Evans via GitGitGadget" <[email protected]> writes:
> -When the command line does not specify what to push with `<refspec>...`
> -arguments or `--all`, `--mirror`, `--tags` options, the command finds
> -the default `<refspec>` by consulting `remote.*.push` configuration,
> -and if it is not found, honors `push.default` configuration to decide
> -what to push (See linkgit:git-config[1] for the meaning of `push.default`).
> +To decide which branches, tags, or other refs to push, Git uses
> +(in order of precedence):
> +
> +1. The <refspec> argument(s) (for example `main` in `git push origin main`)
> + or the `--all`, `--mirror`, or `--tags` options
> +2. The `remote.*.push` configuration for the current branch
> +3. The `push.default` configuration (See linkgit:git-config[1] for
> + the meaning of `push.default`).
The use of numbered list does make very good sense here, as we (at
least conceptually) examine these rules in the order.
> When neither the command-line nor the configuration specifies what to
> -push, the default behavior is used, which corresponds to the `simple`
> -value for `push.default`: the current branch is pushed to the
> -corresponding upstream branch, but as a safety measure, the push is
> -aborted if the upstream branch does not have the same name as the
> -local one.
> +push, the current branch is pushed to the branch with the same name
> +on the remote. The current branch must have a configured upstream with
> +the same name, so this will fail when pushing a new branch.
Is the last sentence correct?
$ cd /var/tmp/playpen
$ git clone https://github.com/git/git src
$ git clone --no-local --bare src dst
$ cd src
$ git checkout -b alter
$ git commit -m 'empty' --allow-empty
$ git -c push.default=simple push ../dst
Enumerating objects: 1, done.
Counting objects: 100% (1/1), done.
Writing objects: 100% (1/1), 185 bytes | 92.00 KiB/s, done.
Total 1 (delta 0), reused 0 (delta 0), pack-reused 0 (from 0)
To ../dst
* [new branch] alter -> alter
In "src" repository that is a fresh clone without any customization,
the current branch "alter" does not have any configured upstream.
Puzzled....
There was a problem hiding this comment. On the Git mailing list, "Julia Evans" wrote (reply to this): > Is the last sentence correct?
>
> $ cd /var/tmp/playpen
> $ git clone https://github.com/git/git src
> $ git clone --no-local --bare src dst
> $ cd src
> $ git checkout -b alter
> $ git commit -m 'empty' --allow-empty
> $ git -c push.default=simple push ../dst
> Enumerating objects: 1, done.
> Counting objects: 100% (1/1), done.
> Writing objects: 100% (1/1), 185 bytes | 92.00 KiB/s, done.
> Total 1 (delta 0), reused 0 (delta 0), pack-reused 0 (from 0)
> To ../dst
> * [new branch] alter -> alter
>
> In "src" repository that is a fresh clone without any customization,
> the current branch "alter" does not have any configured upstream.
>
> Puzzled....
It looks like Git behaves differently depending on whether the remote
being pushed to is named "origin" or not: in this example
the push fails to "origin" but succeeds to a differently named remote.
$ git clone https://github.com/jvns/vue3-tiny-template src
$ cd src
$ git remote add origin2 https://github.com/jvns/vue3-tiny-template
$ git checkout -b alter
$ git -c push.default=simple push origin --dry-run
fatal: The current branch alter has no upstream branch.
$ git -c push.default=simple push origin2 --dry-run
To github.com:jvns/vue3-tiny-template
* [new branch] alter -> alter
I tried to find the responsible code by adding some debug print statements
(in this commit:
https://github.com/git/git/commit/541e5d7cf61f970a5653ab496e5c3111271654a1)
It looks like push.simple has some kind of "same remote" checking, and
if the branch has no tracking information, then origin is considered to be the
"same remote" (so pushing is not allowed), but origin2 is not the "same remote",
so it is allowedThere was a problem hiding this comment. On the Git mailing list, "D. Ben Knoble" wrote (reply to this): On Tue, Aug 26, 2025 at 4:40 PM Julia Evans via GitGitGadget
<[email protected]> wrote:
>
> From: Julia Evans <[email protected]>
>
> - Be more explicit about what we're describing ("which branches" instead
> of "what to push")
Reading this, I thought we would lose precision ("refspec" ->
"branches"), but: the patch (correctly) includes more than branches.
Perhaps
- Be more explicit about what we're describing ("which branches,
[etc.]" instead of "what to push")
?
> - Split out the ways to specify which branches into a numbered list,
> since there are 5 different ways to specify it and it's a lot to parse
> in paragraph form
> - The explanation of "push.default=simple" is confusing to some users,
> use an explanation more similar to the one in `man git-config`
> - Mention the most common case where push.default=simple is likely to
> fail, and how to handle it
>
> Signed-off-by: Julia Evans <[email protected]>
> ---
> Documentation/git-push.adoc | 23 +++++++++++++----------
> 1 file changed, 13 insertions(+), 10 deletions(-)
>
> diff --git a/Documentation/git-push.adoc b/Documentation/git-push.adoc
> index 5c934486c33d..0232195515c9 100644
> --- a/Documentation/git-push.adoc
> +++ b/Documentation/git-push.adoc
> @@ -27,18 +27,21 @@ argument (for example `git push dev`), then if that's not specified the
> `branch.*.remote` configuration for the current branch, and then defaults
> to `origin`.
>
> -When the command line does not specify what to push with `<refspec>...`
> -arguments or `--all`, `--mirror`, `--tags` options, the command finds
> -the default `<refspec>` by consulting `remote.*.push` configuration,
> -and if it is not found, honors `push.default` configuration to decide
> -what to push (See linkgit:git-config[1] for the meaning of `push.default`).
> +To decide which branches, tags, or other refs to push, Git uses
> +(in order of precedence):
> +
> +1. The <refspec> argument(s) (for example `main` in `git push origin main`)
> + or the `--all`, `--mirror`, or `--tags` options
> +2. The `remote.*.push` configuration for the current branch
> +3. The `push.default` configuration (See linkgit:git-config[1] for
> + the meaning of `push.default`).
>
> When neither the command-line nor the configuration specifies what to
> -push, the default behavior is used, which corresponds to the `simple`
> -value for `push.default`: the current branch is pushed to the
> -corresponding upstream branch, but as a safety measure, the push is
> -aborted if the upstream branch does not have the same name as the
> -local one.
> +push, the current branch is pushed to the branch with the same name
> +on the remote. The current branch must have a configured upstream with
> +the same name, so this will fail when pushing a new branch.
> +You can run `git push -u <repository> <current-branch>`
> +to configure the upstream.
I think we've lost the mention of `push.default` here, which we'd
probably like to keep?
Also, I (personally) dislike teach "git push -u <repo> <branch>"
because it leads to some confusion. Let me try to explain myself:
- The way Git treats the configured upstream is the place to pull
from, merge from, rebase onto, etc., configured by branch.X.merge and
branch.X.remote. There is a (possibly!) separate "push to" place
configured (somewhat confusingly) with push.default and
remote.pushDefault.
- The way GitHub and others encourage setting upstream to the place
you push means you have to work a bit harder to pull, merge, rebase
from the semantic upstream (the thing you forked from, not the place
you push to): "git pull <remote> <branch>", "git rebase origin/main",
etc.
- There is, of course, another way that upstream is used: for
"ahead-behind" information in Git's status output. It is convenient to
know where you stand, and Git unfortunately does not provide an easy
way to see the same information against @{push} instead of
@{upstream}. I use some version of "git show-branch HEAD HEAD@{push}"
and a custom "git-div" script [1] for this…
Anyway, point is, I think defaulting to @{upstream} as the place you
push has a nice benefit ("git status" shows me when I haven't pushed
recently), but also is specific to a workflow where you push and pull
from the same places. I find that _most_ of the time I'm actually
doing something triangular (even when I push to and pull from the same
repository, I rarely push to and pull from the same _branch_), and
configuring things a different way affords me many other conveniences.
For example, "git push" and "git pull" without extra flags are
arguments just DWIM.
The confusion I alluded to earlier is that folks rarely learn this is
an option and (in my estimation) lose out on an essential aspect of
what makes Git distributed as opposed to centralized. It also leads to
some confusion over what the term upstream means. And the extra
"push.default explanation is sometimes inaccurate" doesn't help :)
I don't think this is a blocker for this patch, but I do have a bit of
a knee-jerk reaction to this particular piece of
(common-on-the-internet) advice :) Hope that makes sense.
[1]: https://github.com/benknoble/Dotfiles/blob/master/links/bin/git-div
PS I seek a better way to explain this whole idea, so my apologies for
the confused rant.
--
D. Ben Knoble |
||
| to `origin`. | ||
|
|
||
| To decide which branches, tags, or other refs to push, Git uses | ||
| (in order of precedence): | ||
|
|
||
| 1. The <refspec> argument(s) (for example `main` in `git push origin main`) | ||
| or the `--all`, `--mirror`, or `--tags` options | ||
| 2. The `remote.*.push` configuration for the current branch | ||
| 3. The `push.default` configuration (See linkgit:git-config[1] for | ||
| the meaning of `push.default`). | ||
|
|
||
| When neither the command-line nor the configuration specifies what to | ||
| push, the current branch is pushed to the branch with the same name | ||
| on the remote. The current branch must have a configured upstream with | ||
| the same name, so this will fail when pushing a new branch. | ||
| You can run `git push -u <repository> <current-branch>` | ||
| to configure the upstream. | ||
|
|
||
| You can make interesting things happen to a repository | ||
| every time you push into it, by setting up 'hooks' there. See | ||
| documentation for linkgit:git-receive-pack[1]. | ||
|
|
||
| OPTIONS[[OPTIONS]] | ||
| ------------------ | ||
|
|
@@ -55,96 +57,96 @@ OPTIONS[[OPTIONS]] | |
|
|
||
|
There was a problem hiding this comment. On the Git mailing list, Junio C Hamano wrote (reply to this): "Julia Evans via GitGitGadget" <[email protected]> writes:
> From: Julia Evans <[email protected]>
>
> - Originally it said that a refspec was `+<src>:<dst>`, but then later
> contradicted itself by saying that the `:<dst>` is optional.
> Mention that `:<dst>` is optional much earlier.
> - Put the complex sets of rules about different refspec forms
> in lists instead of in long paragraphs of prose
> - Add examples for the various types of refspecs
> (negative, deletion, pattern, etc)
> - Previously `*` and `^` were not mentioned, mention them
> - Explain what `+` does earlier
> - Remove "might be added in the future" (it's a given that software
> might change in the future)
>
> Signed-off-by: Julia Evans <[email protected]>
> ---
> Documentation/git-push.adoc | 164 ++++++++++++++++++------------------
> 1 file changed, 82 insertions(+), 82 deletions(-)
>
> diff --git a/Documentation/git-push.adoc b/Documentation/git-push.adoc
> index 0232195515c9..78d433c60c51 100644
> --- a/Documentation/git-push.adoc
> +++ b/Documentation/git-push.adoc
> @@ -57,77 +57,74 @@ OPTIONS[[OPTIONS]]
>
> <refspec>...::
> Specify what destination ref to update with what source object.
> - The format of a <refspec> parameter is an optional plus
> - `+`, followed by the source object <src>, followed
> - by a colon `:`, followed by the destination ref <dst>.
> -+
> -The <src> is often the name of the branch you would want to push, but
> -it can be any arbitrary "SHA-1 expression", such as `master~4` or
> -`HEAD` (see linkgit:gitrevisions[7]).
> -+
> -The <dst> tells which ref on the remote side is updated with this
> -push. Arbitrary expressions cannot be used here, an actual ref must
> -be named.
> -If `git push [<repository>]` without any `<refspec>` argument is set to
> -update some ref at the destination with `<src>` with
> -`remote.<repository>.push` configuration variable, `:<dst>` part can
> -be omitted--such a push will update a ref that `<src>` normally updates
> -without any `<refspec>` on the command line. Otherwise, missing
> -`:<dst>` means to update the same ref as the `<src>`.
> -+
> -If <dst> doesn't start with `refs/` (e.g. `refs/heads/master`) we will
> -try to infer where in `refs/*` on the destination <repository> it
> -belongs based on the type of <src> being pushed and whether <dst>
> -is ambiguous.
> +
> ---
> -* If <dst> unambiguously refers to a ref on the <repository> remote,
> - then push to that ref.
> -
> -* If <src> resolves to a ref starting with refs/heads/ or refs/tags/,
> - then prepend that to <dst>.
> -
> -* Other ambiguity resolutions might be added in the future, but for
> - now any other cases will error out with an error indicating what we
> - tried, and depending on the `advice.pushUnqualifiedRefname`
> - configuration (see linkgit:git-config[1]) suggest what refs/
> - namespace you may have wanted to push to.
> -
> ---
> -+
> -The object referenced by <src> is used to update the <dst> reference
> -on the remote side. Whether this is allowed depends on where in
> -`refs/*` the <dst> reference lives as described in detail below, in
> -those sections "update" means any modifications except deletes, which
> -as noted after the next few sections are treated differently.
> -+
> -The `refs/heads/*` namespace will only accept commit objects, and
> -updates only if they can be fast-forwarded.
> -+
> -The `refs/tags/*` namespace will accept any kind of object (as
> -commits, trees and blobs can be tagged), and any updates to them will
> -be rejected.
> -+
> -It's possible to push any type of object to any namespace outside of
> -`refs/{tags,heads}/*`. In the case of tags and commits, these will be
> -treated as if they were the commits inside `refs/heads/*` for the
> -purposes of whether the update is allowed.
> -+
> -I.e. a fast-forward of commits and tags outside `refs/{tags,heads}/*`
> -is allowed, even in cases where what's being fast-forwarded is not a
> -commit, but a tag object which happens to point to a new commit which
> -is a fast-forward of the commit the last tag (or commit) it's
> -replacing. Replacing a tag with an entirely different tag is also
> -allowed, if it points to the same commit, as well as pushing a peeled
> -tag, i.e. pushing the commit that existing tag object points to, or a
> -new tag object which an existing commit points to.
> -+
> -Tree and blob objects outside of `refs/{tags,heads}/*` will be treated
> -the same way as if they were inside `refs/tags/*`, any update of them
> -will be rejected.
> -+
> -All of the rules described above about what's not allowed as an update
> -can be overridden by adding an the optional leading `+` to a refspec
All of the above is fairly accurate, even though they are densely
written and at places somewhat redundant.
> +The format for a refspec is [+]<src>[:<dst>], for example `main`,
> +`main:other`, or `HEAD^:refs/heads/main`.
> ++
> +The `<src>` is often the name of the local branch to push, but it can be
> +any arbitrary "SHA-1 expression" (see linkgit:gitrevisions[7]).
> ++
> +The `<dst>` determines what to update on the remote side. It must be the
> +name of a branch, tag, or other ref, not an arbitrary expression.
Concise and simpler beginning is good.
> +`:<dst>` is optional.
It may be technically true, but I am not sure if it is a good idea
to say it here. Without "when missing, <dst> is inferred with these
rules", saying just "is optional" naturally invites a puzzlement:
when do we need to supply it and for what?
As you are going to say what happens when you omit it very soon, and
you already have said with [:<dst>] that it is optional, perhaps you
can scratch this sentence.
> +`+` is optional and does the same thing as `--force`.
Ditto; this one is less bad than the :<dst> thing, because at least
it tells us what it means. But we are going to talk about when an
update is not allowed (we haven't even hinted that some updates may
not be allowed yet) much later, "the same as `--force`" is probably
a bit premature at this point in the documentation.
> +You can write a refspec using the fully expanded form (for
> +example `main:refs/heads/main`) which specifies the exact source
This example is not fully expanded. refs/heads/main:refs/heads/main
would be, though.
> +and destination, or with a shorter form (for example `main` or
> +`main:other`). Here are the rules for how refspecs are expanded,
> +as well as various other special refspec forms:
> ++
I am not sure it is easier to read with numbered list. It is not
like these rules are applied in this order, or anything like that,
right?
> + 1. `<src>` without a `:<dst>` means to update the same ref as the
> + `<src>`, unless the `remote.<repository>.push` configuration specifies a
> + different <dst>. For example, if `main` is a branch, then the refspec
> + `main` expands to `main:refs/heads/main`.
> + 2. If <dst> unambiguously refers to a ref on the <repository> remote,
> + then expand it to that ref. For example, if `v1.0` is a tag on the
> + remote, then `HEAD:v1.0` expands to `HEAD:refs/tags/v1.0`.
> + 3. If <src> resolves to a ref starting with refs/heads/ or refs/tags/,
> + then prepend that to <dst>. For example, if `main` is a branch, then
> + `main:other` expands to `main:refs/heads/other`
> + 4. The special refspec `:` (or `+:` to allow non-fast-forward updates)
> + directs Git to push "matching" branches: for every branch that exists on
> + the local side, the remote side is updated if a branch of the same name
> + already exists on the remote side.
Good to see that the oddballs like this and an empty one, originally
described elsewhere, collected in the same list. An entry like this
that describes a special notation, not a concrete single refspec,
however is a bit hard to read when mixed with other more normal
rules. Perhaps move it down together with the "push void to remove"
at the top of the list?
> + 5. `tag <tag>` expands to `refs/tags/<tag>:refs/tags/<tag>`.
A tangent.
Is this a refspec you can write in .git/config, e.g.
[remote "origin"]
push = tag v1.0"
If not, it might be easier to explain if we tweaked the command line
synopsis to say that the command takes, after the destination
repository, zero or more refspec or "tag <tag>". I dunno.
> + 6. <src> may contain a * to indicate a simple pattern match.
> + This works like a glob that matches any ref matching the pattern.
> + There must be only one * in both the <src> and <dst>.
> + It will map refs to the destination by replacing the * with the
> + contents matched from the source. For example, `refs/heads/*:refs/heads/*`
> + will push all branches.
> + 7. A refspec starting with ^ is a negative refspec.
> + This specifies refs to exclude. A ref will be considered to
> + match if it matches at least one positive refspec, and does not
> + match any negative refspec. Negative refspecs can be pattern refspecs.
> + They must only contain a <src>.
> + Fully spelled out hex object names are also not supported.
> + For example, `git push origin 'refs/heads/*' '^refs/heads/dev-*'`
> + will push all branches except for those starting with `dev-`
Good. Somehow we have added the description of these to "git fetch"
side, without updating "git push" side of the documentation.
> + 8. If `<src>` is empty, it deletes the <dst> ref from the remote
> + repository. For example, `git push origin :dev` will
> + delete the `dev` branch.
OK. This is specific to "push" (you do not delete your local branch
by fetching :refs/heads/to-be-removed from elsewhere), so even if we
wanted to unify the descriptions on both sides in the future, we'd
need to be a bit careful around here.
> + Deletions are always accepted without a leading `+` in the
> + refspec (or `--force`), except when forbidden by configuration or hooks.
This can be read in two ways, making two opposing answers to this
question possible: when forbidden, can you make a deletion accepted
by giving a `+` in front?
"except when forbidden, deletions are accepted with or without `+`"
might be less confusion-prone, but I dunno. I just wanted to make
sure that it is clear that forcing or prepending `+` would not
change anything when forbidden by configuration or hooks on the
remote end.
But because you haven't mentioned that not all updates are allowed,
this might be a bit out of place in this list. How about limiting
this bullet point to only say that this is the syntax to use to
delete a ref from the remote, and move the "deletions do not have to
be forced and operations forbidden at the remote cannot be forced
anyway" down, near the "Not all updates are allowed" below ...
> + See `receive.denyDeletes` in linkgit:git-config[1] and `pre-receive` and
> + `update` in linkgit:githooks[5].
... together with this?
> + 9. If the refspec can't be expanded unambiguously, error
> + out with an error indicating what was
> + tried, and depending on the `advice.pushUnqualifiedRefname`
> + configuration (see linkgit:git-config[1]) suggest what refs/
> + namespace you may have wanted to push to.
> +
> ++
> +Not all updates are allowed: it depends on what kind of destination
> +you're pushing to. In the following rules "update" means any
> +modifications except deletes, which as noted above are treated differently.
> ++
> +All of these rules
> +can be overridden by adding the optional leading `+` to a refspec
> (or using `--force` command line option). The only exception to this
> is that no amount of forcing will make the `refs/heads/*` namespace
> accept a non-commit object. Hooks and configuration can also overrideThere was a problem hiding this comment. On the Git mailing list, "Julia Evans" wrote (reply to this): >> +`:<dst>` is optional.
>
> It may be technically true, but I am not sure if it is a good idea
> to say it here. Without "when missing, <dst> is inferred with these
> rules", saying just "is optional" naturally invites a puzzlement:
> when do we need to supply it and for what?
>
> As you are going to say what happens when you omit it very soon, and
> you already have said with [:<dst>] that it is optional, perhaps you
> can scratch this sentence.
The reason I mentioned this early on is that when I read this for the first time
it was quite confusing. Since `main` (or another branch name on its own) was
the only push refspec I had experience using, it was disorienting to be told
that the format was `<src>:<dst>`. It made me wonder if maybe `main` was not a
refspec after all and I'd misunderstood the SYNOPSIS.
I know that I'm not alone in this: it's very common for Git users (even very
experienced ones!) to learn the form `git push origin main`, but to
otherwise not be familiar at all with the concept of a "refspec".
So I'd like to do something here to help those folks connect their
existing knowledge to the broader concept of a "refspec".
That said `:<dst> is optional` may not be necessary: I think the examples
here accomplish more or less the same thing.
>> +The format for a refspec is [+]<src>[:<dst>], for example `main`,
> > +`main:other`, or `HEAD^:refs/heads/main`.
>> +`+` is optional and does the same thing as `--force`.
>
> Ditto; this one is less bad than the :<dst> thing, because at least
> it tells us what it means. But we are going to talk about when an
> update is not allowed (we haven't even hinted that some updates may
> not be allowed yet) much later, "the same as `--force`" is probably
> a bit premature at this point in the documentation.
The reason I mentioned this so early is similar -- It's very common
for experienced users to be familiar with the `--force` option, and
so this is a quick way to explain to them what the `+` does.
As an aside, I noticed while reading this that it looks like
the description of `--force` right now is not quite accurate,
since the rule about pushing to tags is not "it must be
an ancestor".
>--force::
> Usually, the command refuses to update a remote ref that is
> not an ancestor of the local ref used to overwrite it.
I'm going to think about whether there's a way to format the
"rules for updates" so that they can be referred to from the `--force`
section. Or maybe just move them to the `--force` option, and refer
people there if they want to understand exactly when `--force`
is required.
It's certainly much more common for users to use `--force` to force
a push than to use `+`, from that perspective it would make more
sense to document the rules next to `--force` than next to the
description of `+`.
>> +You can write a refspec using the fully expanded form (for
>> +example `main:refs/heads/main`) which specifies the exact source
>
> This example is not fully expanded. refs/heads/main:refs/heads/main
> would be, though.
Makes sense, will change.
> I am not sure it is easier to read with numbered list. It is not
> like these rules are applied in this order, or anything like that,
> right?
I had a similar concern. I'm still trying to figure out how to manage
unordered lists, since lists with a `*` are formatted by AsciiDoc in a
weird way in the terminal: there's a tab character after the • character
which I don't understand the reason for.
> A tangent.
> Is this a refspec you can write in .git/config, e.g.
>
> [remote "origin"]
> push = tag v1.0"
>
> If not, it might be easier to explain if we tweaked the command line
> synopsis to say that the command takes, after the destination
> repository, zero or more refspec or "tag <tag>". I dunno.
I checked and it's not a valid refspec:
$ git push
fatal: invalid refspec 'tag v1.0'
This list seems like a more natural place for that information than the
synopsis though: the synopsis is already quite hard to read and we
can't use it as the only place to communicate that information.
I can add a note to say that 'tag v1.0' is technically not a refspec.
Git already has enough weird exceptions like that that I don't think
it'll be too jarring.
> Perhaps move it down together with the "push void to remove"
> at the top of the list?
I'll also move it down.
> Good. Somehow we have added the description of these to "git fetch"
> side, without updating "git push" side of the documentation.
Yes, I copied this part from the `git fetch` documention after checking
that they worked with `git push` as well.
>> + Deletions are always accepted without a leading `+` in the
>> + refspec (or `--force`), except when forbidden by configuration or hooks.
>
> This can be read in two ways, making two opposing answers to this
> question possible: when forbidden, can you make a deletion accepted
> by giving a `+` in front?
> "except when forbidden, deletions are accepted with or without `+`"
> might be less confusion-prone, but I dunno. I just wanted to make
> sure that it is clear that forcing or prepending `+` would not
> change anything when forbidden by configuration or hooks on the
> remote end.
>
> But because you haven't mentioned that not all updates are allowed,
> this might be a bit out of place in this list. How about limiting
> this bullet point to only say that this is the syntax to use to
> delete a ref from the remote, and move the "deletions do not have to
> be forced and operations forbidden at the remote cannot be forced
> anyway" down, near the "Not all updates are allowed" below ...
Will do, that's much cleaner.
>> + See `receive.denyDeletes` in linkgit:git-config[1] and `pre-receive` and
>> + `update` in linkgit:githooks[5].
>
> ... together with this?
This makes sense.
- JuliaThere was a problem hiding this comment. On the Git mailing list, "D. Ben Knoble" wrote (reply to this): Thanks for the docs updates! A few nits below, but this looks nice to me :)
On Tue, Aug 26, 2025 at 4:40 PM Julia Evans via GitGitGadget
<[email protected]> wrote:
>
> From: Julia Evans <[email protected]>
>
> - Originally it said that a refspec was `+<src>:<dst>`, but then later
> contradicted itself by saying that the `:<dst>` is optional.
> Mention that `:<dst>` is optional much earlier.
> - Put the complex sets of rules about different refspec forms
> in lists instead of in long paragraphs of prose
> - Add examples for the various types of refspecs
> (negative, deletion, pattern, etc)
> - Previously `*` and `^` were not mentioned, mention them
> - Explain what `+` does earlier
> - Remove "might be added in the future" (it's a given that software
> might change in the future)
Excellent ideas!
> diff --git a/Documentation/git-push.adoc b/Documentation/git-push.adoc
> index 0232195515c9..78d433c60c51 100644
> --- a/Documentation/git-push.adoc
> +++ b/Documentation/git-push.adoc
> @@ -57,77 +57,74 @@ OPTIONS[[OPTIONS]]
> +The format for a refspec is [+]<src>[:<dst>], for example `main`,
> +`main:other`, or `HEAD^:refs/heads/main`.
> ++
> +The `<src>` is often the name of the local branch to push, but it can be
> +any arbitrary "SHA-1 expression" (see linkgit:gitrevisions[7]).
> ++
> +The `<dst>` determines what to update on the remote side. It must be the
> +name of a branch, tag, or other ref, not an arbitrary expression.
A welcome (to me) simplification from the original paragraph,
especially if we don't lose the original content but rearrange it
better :)
> +`:<dst>` is optional.
Here…
> ++
> +`+` is optional and does the same thing as `--force`.
> ++
…and here, I find it odd to start sentences with punctuation if we can avoid it.
> +You can write a refspec using the fully expanded form (for
> +example `main:refs/heads/main`) which specifies the exact source
> +and destination, or with a shorter form (for example `main` or
> +`main:other`). Here are the rules for how refspecs are expanded,
> +as well as various other special refspec forms:
> ++
> + 1. `<src>` without a `:<dst>` means to update the same ref as the
> + `<src>`, unless the `remote.<repository>.push` configuration specifies a
> + different <dst>. For example, if `main` is a branch, then the refspec
> + `main` expands to `main:refs/heads/main`.
> + 2. If <dst> unambiguously refers to a ref on the <repository> remote,
> + then expand it to that ref. For example, if `v1.0` is a tag on the
> + remote, then `HEAD:v1.0` expands to `HEAD:refs/tags/v1.0`.
> + 3. If <src> resolves to a ref starting with refs/heads/ or refs/tags/,
> + then prepend that to <dst>. For example, if `main` is a branch, then
> + `main:other` expands to `main:refs/heads/other`
> + 4. The special refspec `:` (or `+:` to allow non-fast-forward updates)
> + directs Git to push "matching" branches: for every branch that exists on
> + the local side, the remote side is updated if a branch of the same name
> + already exists on the remote side.
I'm not 100% sure this belongs as an item in an ordered list here,
since it implies (structurally) something about the order of possible
expansions tried. But the introduction to the list does say "rules
[and] special refspec forms"… Hm. Maybe it's worth splitting the
special ones out? idk.
I see Junio mentioned something similar.
> + 5. `tag <tag>` expands to `refs/tags/<tag>:refs/tags/<tag>`.
> + 6. <src> may contain a * to indicate a simple pattern match.
> + This works like a glob that matches any ref matching the pattern.
> + There must be only one * in both the <src> and <dst>.
> + It will map refs to the destination by replacing the * with the
Should src/dst/* have backticks here? I'm not sure.
> + contents matched from the source. For example, `refs/heads/*:refs/heads/*`
> + will push all branches.
> + 7. A refspec starting with ^ is a negative refspec.
Ditto the "^"
> + This specifies refs to exclude. A ref will be considered to
> + match if it matches at least one positive refspec, and does not
> + match any negative refspec. Negative refspecs can be pattern refspecs.
> + They must only contain a <src>.
> + Fully spelled out hex object names are also not supported.
> + For example, `git push origin 'refs/heads/*' '^refs/heads/dev-*'`
> + will push all branches except for those starting with `dev-`
I learned something new today! This isn't in the manual I have for
2.48.1 or 2.51.x locally. Thanks! [Junio mentions it's on the fetch
side, which I see now]
> + 8. If `<src>` is empty, it deletes the <dst> ref from the remote
Backticks for dst ;)
> + repository. For example, `git push origin :dev` will
> + delete the `dev` branch.
> + Deletions are always accepted without a leading `+` in the
> + refspec (or `--force`), except when forbidden by configuration or hooks.
Maybe "except when forbidden on the remote by…" ? (This came from the
original and does not need tweaked in this series, though.)
> + See `receive.denyDeletes` in linkgit:git-config[1] and `pre-receive` and
> + `update` in linkgit:githooks[5].
> + 9. If the refspec can't be expanded unambiguously, error
> + out with an error indicating what was
> + tried, and depending on the `advice.pushUnqualifiedRefname`
> + configuration (see linkgit:git-config[1]) suggest what refs/
> + namespace you may have wanted to push to.
Wrapping looks strange to me here.
> +
> ++
> +Not all updates are allowed: it depends on what kind of destination
> +you're pushing to. In the following rules "update" means any
> +modifications except deletes, which as noted above are treated differently.
> ++
> +All of these rules
> +can be overridden by adding the optional leading `+` to a refspec
Ditto.
> (or using `--force` command line option). The only exception to this
> is that no amount of forcing will make the `refs/heads/*` namespace
> accept a non-commit object. Hooks and configuration can also override
> @@ -135,18 +132,21 @@ or amend these rules, see e.g. `receive.denyNonFastForwards` in
> linkgit:git-config[1] and `pre-receive` and `update` in
> linkgit:githooks[5].
> +
> -Pushing an empty <src> allows you to delete the <dst> ref from the
> -remote repository. Deletions are always accepted without a leading `+`
> -in the refspec (or `--force`), except when forbidden by configuration
> -or hooks. See `receive.denyDeletes` in linkgit:git-config[1] and
> -`pre-receive` and `update` in linkgit:githooks[5].
> -+
> -The special refspec `:` (or `+:` to allow non-fast-forward updates)
> -directs Git to push "matching" branches: for every branch that exists on
> -the local side, the remote side is updated if a branch of the same name
> -already exists on the remote side.
> -+
> -`tag <tag>` means the same as `refs/tags/<tag>:refs/tags/<tag>`.
> +1. If the destination is a **branch** (`refs/heads/*`): the source must
> + be a commit object, and only fast-forward updates are allowed.
> +2. If the destination is a **tag** (`refs/tags/*`): the source can
> + be any object (as commits, trees and blobs can be tagged), and any
> + updates to them will be rejected.
> +3. For destinations outside of `refs/{tags,heads}/*`:
> + * If the source is a tree or blob object, any updates will be rejected
> + * If the source is a tag or commit object, any fast-forward update
> + is allowed, even in cases where what's being fast-forwarded is not a
> + commit, but a tag object which happens to point to a new commit which
> + is a fast-forward of the commit the last tag (or commit) it's
> + replacing. Replacing a tag with an entirely different tag is also
> + allowed, if it points to the same commit, as well as pushing a peeled
> + tag, i.e. pushing the commit that existing tag object points to, or a
> + new tag object which an existing commit points to.
I didn't close-read this bit, but it seems reasonable.
--
D. Ben Knoble |
||
| <refspec>...:: | ||
| Specify what destination ref to update with what source object. | ||
| + | ||
| The format for a refspec is [+]<src>[:<dst>], for example `main`, | ||
| `main:other`, or `HEAD^:refs/heads/main`. | ||
| + | ||
| The `<src>` is often the name of the local branch to push, but it can be | ||
| any arbitrary "SHA-1 expression" (see linkgit:gitrevisions[7]). | ||
| + | ||
| The `<dst>` determines what to update on the remote side. It must be the | ||
| name of a branch, tag, or other ref, not an arbitrary expression. | ||
| `:<dst>` is optional. | ||
| + | ||
| `+` is optional and does the same thing as `--force`. | ||
| + | ||
| You can write a refspec using the fully expanded form (for | ||
| example `main:refs/heads/main`) which specifies the exact source | ||
| and destination, or with a shorter form (for example `main` or | ||
| `main:other`). Here are the rules for how refspecs are expanded, | ||
| as well as various other special refspec forms: | ||
| + | ||
| 1. `<src>` without a `:<dst>` means to update the same ref as the | ||
| `<src>`, unless the `remote.<repository>.push` configuration specifies a | ||
| different <dst>. For example, if `main` is a branch, then the refspec | ||
| `main` expands to `main:refs/heads/main`. | ||
| 2. If <dst> unambiguously refers to a ref on the <repository> remote, | ||
| then expand it to that ref. For example, if `v1.0` is a tag on the | ||
| remote, then `HEAD:v1.0` expands to `HEAD:refs/tags/v1.0`. | ||
| 3. If <src> resolves to a ref starting with refs/heads/ or refs/tags/, | ||
| then prepend that to <dst>. For example, if `main` is a branch, then | ||
| `main:other` expands to `main:refs/heads/other` | ||
| 4. The special refspec `:` (or `+:` to allow non-fast-forward updates) | ||
| directs Git to push "matching" branches: for every branch that exists on | ||
| the local side, the remote side is updated if a branch of the same name | ||
| already exists on the remote side. | ||
| 5. `tag <tag>` expands to `refs/tags/<tag>:refs/tags/<tag>`. | ||
| 6. <src> may contain a * to indicate a simple pattern match. | ||
| This works like a glob that matches any ref matching the pattern. | ||
| There must be only one * in both the <src> and <dst>. | ||
| It will map refs to the destination by replacing the * with the | ||
| contents matched from the source. For example, `refs/heads/*:refs/heads/*` | ||
| will push all branches. | ||
| 7. A refspec starting with ^ is a negative refspec. | ||
| This specifies refs to exclude. A ref will be considered to | ||
| match if it matches at least one positive refspec, and does not | ||
| match any negative refspec. Negative refspecs can be pattern refspecs. | ||
| They must only contain a <src>. | ||
| Fully spelled out hex object names are also not supported. | ||
| For example, `git push origin 'refs/heads/*' '^refs/heads/dev-*'` | ||
| will push all branches except for those starting with `dev-` | ||
| 8. If `<src>` is empty, it deletes the <dst> ref from the remote | ||
| repository. For example, `git push origin :dev` will | ||
| delete the `dev` branch. | ||
| Deletions are always accepted without a leading `+` in the | ||
| refspec (or `--force`), except when forbidden by configuration or hooks. | ||
| See `receive.denyDeletes` in linkgit:git-config[1] and `pre-receive` and | ||
| `update` in linkgit:githooks[5]. | ||
| 9. If the refspec can't be expanded unambiguously, error | ||
| out with an error indicating what was | ||
| tried, and depending on the `advice.pushUnqualifiedRefname` | ||
| configuration (see linkgit:git-config[1]) suggest what refs/ | ||
| namespace you may have wanted to push to. | ||
|
|
||
| + | ||
| Not all updates are allowed: it depends on what kind of destination | ||
| you're pushing to. In the following rules "update" means any | ||
| modifications except deletes, which as noted above are treated differently. | ||
| + | ||
| All of these rules | ||
| can be overridden by adding the optional leading `+` to a refspec | ||
| (or using `--force` command line option). The only exception to this | ||
| is that no amount of forcing will make the `refs/heads/*` namespace | ||
| accept a non-commit object. Hooks and configuration can also override | ||
| or amend these rules, see e.g. `receive.denyNonFastForwards` in | ||
| linkgit:git-config[1] and `pre-receive` and `update` in | ||
| linkgit:githooks[5]. | ||
| + | ||
| 1. If the destination is a **branch** (`refs/heads/*`): the source must | ||
| be a commit object, and only fast-forward updates are allowed. | ||
| 2. If the destination is a **tag** (`refs/tags/*`): the source can | ||
| be any object (as commits, trees and blobs can be tagged), and any | ||
| updates to them will be rejected. | ||
| 3. For destinations outside of `refs/{tags,heads}/*`: | ||
| * If the source is a tree or blob object, any updates will be rejected | ||
| * If the source is a tag or commit object, any fast-forward update | ||
| is allowed, even in cases where what's being fast-forwarded is not a | ||
| commit, but a tag object which happens to point to a new commit which | ||
| is a fast-forward of the commit the last tag (or commit) it's | ||
| replacing. Replacing a tag with an entirely different tag is also | ||
| allowed, if it points to the same commit, as well as pushing a peeled | ||
| tag, i.e. pushing the commit that existing tag object points to, or a | ||
| new tag object which an existing commit points to. | ||
|
|
||
| --all:: | ||
| --branches:: | ||
|
|
||
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the Git mailing list, "D. Ben Knoble" wrote (reply to this):
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the Git mailing list, Junio C Hamano wrote (reply to this):
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the Git mailing list, "Julia Evans" wrote (reply to this):
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the Git mailing list, "D. Ben Knoble" wrote (reply to this):
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the Git mailing list, "Kristoffer Haugsbakk" wrote (reply to this):