The comment was not formatted correctly, and because of that not picked up as
being deprecated.
updates b4ca1c7368
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
(cherry picked from commit e3fa7280ad)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Make the error more specific by stating that it's caused by a specific
environment variable and not an environment as a whole.
Also don't escape the variable to make it more readable.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
(cherry picked from commit 012b77952e)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
On Windows, ignore all variables that start with "=" when building an
environment variables map for stack.
For MS-DOS compatibility cmd.exe can set some special environment
variables that start with a "=" characters, which breaks the general
assumption that the first encountered "=" separates a variable name from
variable value and causes trouble when parsing.
These variables don't seem to be documented anywhere, but they are
described by some third-party sources and confirmed empirically on my
Windows installation.
Useful sources:
https://devblogs.microsoft.com/oldnewthing/20100506-00/?p=14133https://ss64.com/nt/syntax-variables.html
Known variables:
- `=ExitCode` stores the exit code returned by external command (in hex
format)
- `=ExitCodeAscii` - same as above, except the value is the ASCII
representation of the code (so exit code 65 (0x41) becomes 'A').
- `=::=::\` and friends - store drive specific working directory.
There is one env variable for each separate drive letter that was
accessed in the shell session and stores the working directory for that
specific drive.
The general format for these is:
`=<DRIVE_LETTER>:=<CWD>` (key=`=<DRIVE_LETTER>:`, value=`<CWD>`)
where <CWD> is a working directory for the drive that is assigned to
the letter <DRIVE_LETTER>
A couple of examples:
`=C:=C:\some\dir` (key: `=C:`, value: `C:\some\dir`)
`=D:=D:\some\other\dir` (key: `=C:`, value: `C:\some\dir`)
`=Z:=Z:\` (key: `=Z:`, value: `Z:\`)
`=::=::\` is the one that seems to be always set and I'm not exactly
sure what this one is for (what's drive `::`?). Others are set as
soon as you CD to a path on some drive. Considering that you start a
cmd.exe also has some working directory, there are 2 of these on start.
All these variables can be safely ignored because they can't be
deliberately set by the user, their meaning is only relevant to the
cmd.exe session and they're all are related to the MS-DOS/Batch feature
that are irrelevant for us.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
(cherry picked from commit a47058bbd5)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Tests mocking the output of GET images/json with fakeClient used an
array with one empty element as an empty response.
Change it to just an empty array.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
(cherry picked from commit a1953e19b2)
The error returned from "os/exec".Command when attempting to execute a
directory has been changed from syscall.EACCESS to syscall.EISDIR on
Go 1.20. 2b8f214094
Consequently, any runc runtime built against Go 1.20 will return an
error containing 'is a directory' and not 'permission denied'. Update
the string matching so the CLI exits with status code 126 on 'is a
directory' errors (EISDIR) in addition to 'permission denied' (EACCESS).
Signed-off-by: Cory Snider <csnider@mirantis.com>
(cherry picked from commit 9b5ceb52b0)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This prevents us needing to attempt to reconstruct the exact indentation
registry side, which is not canonical - so may differ.
Signed-off-by: Justin Chadwell <me@jedevc.com>
This behavior should not break any more use cases than before.
Previously, if the mismatch occured, we would actually push a manifest
that we then never referred to in the manifest list! If this was done in
a new repository, the command would fail with an obscure error from the
registry - the content wouldn't exist with the descriptor we expect it
to.
Signed-off-by: Justin Chadwell <me@jedevc.com>
The additionalHelp message is printed at the end of the --help output;
To get more help with docker, check out our guides at https://docs.docker.com/go/guides/
PS>
As this message may contain an URL, users may copy/paste the URL to open it
in their browser, but can easily end up copying their prompt (as there's
no whitespace after it), and as a result end up on a broken URL, for example:
https://docs.docker.com/go/guides/PS
This patch adds an extra newline at the end to provide some whitespace
around the message, making it less error-prone to copy the URL;
To get more help with docker, check out our guides at https://docs.docker.com/go/guides/
PS>
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This prevents the escape-characters being included when piping the
output, e.g. `docker --help > output.txt`, or `docker --help | something`.
These control-characters could cause issues if users copy/pasted the URL
from the output, resulting in them becoming part of the URL they tried
to visit, which would fail, e.g. when copying the output from:
To get more help with docker, check out our guides at https://docs.docker.com/go/guides/
Users ended up on URLs like;
https://docs.docker.com/go/guides/ESChttps://docs.docker.com/go/guides/%1B[0m
Before this patch, control characters ("bold") would be printed, even if
no TTY was attached;
docker --help > output.txt
cat output.txt | grep 'For more help' | od -c
0000000 033 [ 1 m F o r m o r e h e l
0000020 p o n h o w t o u s e
0000040 D o c k e r , h e a d t o
0000060 h t t p s : / / d o c s . d o c
0000100 k e r . c o m / g o / g u i d e
0000120 s / 033 [ 0 m \n
0000127
docker --help | grep 'For more help' | od -c
0000000 033 [ 1 m F o r m o r e h e l
0000020 p o n h o w t o u s e
0000040 D o c k e r , h e a d t o
0000060 h t t p s : / / d o c s . d o c
0000100 k e r . c o m / g o / g u i d e
0000120 s / 033 [ 0 m \n
0000127
With this patch, no control characters are included:
docker --help > output.txt
cat output.txt | grep 'For more help' | od -c
0000000 F o r m o r e h e l p o n
0000020 h o w t o u s e D o c k
0000040 e r , h e a d t o h t t p
0000060 s : / / d o c s . d o c k e r .
0000100 c o m / g o / g u i d e s / \n
0000117
docker --help | grep 'For more help' | od -c
0000000 F o r m o r e h e l p o n
0000020 h o w t o u s e D o c k
0000040 e r , h e a d t o h t t p
0000060 s : / / d o c s . d o c k e r .
0000100 c o m / g o / g u i d e s / \n
0000117
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Synchronize append on the `removed` slice with mutex because
containerRemoveFunc is called in parallel for each removed container by
`container rm` cli command.
Also reduced the shared access area by separating the scopes of test
cases.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
This comment was added in 7929888214
when this code was still in the Moby repository. That comment doesn't appear
to apply to the CLI's usage of this struct though, as nothing in the CLI
sets this field (or uses it), so this should be safe to remove.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Libtrust was only used for pushing schema 2, v1 images, which is no longer
supported; this TODO was likely left from when the CLI and daemon were
in the same repository.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This allows the cli to be initialized with a (custom) API client.
Currently to be used for unit tests, but could be used for other
scenarios.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Make sure that the container has multiple port-mappings to illustrate
that only the given port is matched.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
- use strings.Cut
- don't use nat.NewPort as we don't accept port ranges
- use an early return if there's no results
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
if a context is set (e.g. through DOCKER_CONTEXT or the CLI config file), but
wasn't found, then a "stub" context is added, including an error message that
the context doesn't exist.
DOCKER_CONTEXT=nosuchcontext docker context ls
NAME DESCRIPTION DOCKER ENDPOINT ERROR
default Current DOCKER_HOST based configuration unix:///var/run/docker.sock
nosuchcontext * context "nosuchcontext": context not found: …
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This updates `docker context ls` to:
- not abort listing contexts when failing one (or more) contexts
- instead, adding an ERROR column to inform the user there was
an issue loading the context.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This allows commands that don't require a client connection (such as `context use`)
to be functional, but still produces an error when trying to run a command that
needs to connect with the API;
mkdir -p ~/.docker/ && echo '{"currentContext":"nosuchcontext"}' > ~/.docker/config.json
docker version
Failed to initialize: unable to resolve docker endpoint: load context "nosuchcontext": context does not exist: open /root/.docker/contexts/meta/8bfef2a74c7d06add4bf4c73b0af97d9f79c76fe151ae0e18b9d7e57104c149b/meta.json: no such file or directory
docker context use default
default
Current context is now "default"
docker version
Client:
Version: 22.06.0-dev
API version: 1.42
...
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The "docker context show" command is intended to show the currently configured
context. While the context that's configured may not be valid (e.g., in case
an environment variable was set to configure the context, or if the context
was removed from the filesystem), we should still be able to _show_ the
context.
This patch removes the context validation, and instead only shows the context.
This can help in cases where the context is used to (e.g.) set the command-
prompt, but the user removed the context. With this change, the context name
can still be shown, but commands that _require_ the context will still fail.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This internalizes constructing the Client(), which allows us to provide
fallbacks when trying to determin the current API version.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Also move the resolveContextName() function together with the
method for easier cross-referencing.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
There's no strict need to perform this validation inside this function;
validating flags should happen earlier, to allow faster detecting of
configuration issues (we may want to have a central config "validate"
function though).
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
resolveContextName() is used to find which context to use, based on the
available configuration options. Once resolved, the context name is
used to load the actual context, which will fail if the context doesn't
exist, so there's no need to produce an error at this stage; only
check priority of the configuration options to pick the context
with the highest priority.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
CommonOptions was inherited from when the cli and daemon were in the same
repository, and some options would be shared between them. That's no longer
the case, and some options are even "incorrect" (for example, while the
daemon can be configured to run on multiple hosts, the CLI can only connect
with a single host / connection). This patch does not (yet) address that,
but merges the CommonOptions into the ClientOptions.
An alias is created for the old type, although it doesn't appear there's
any external consumers using the CommonOptions type (or its constructor).
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
- Make the package-level configMergeTests local to the test itself.
- Rename fields to better describe intent
- Remove some redundant variables
- Reverse "expected" and "actual" fields for consistency
- Use assert.Check() to not fail early
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Various fixes:
- Don't capitalize error messages
- Rename variables that collided with imports or types
- Prefer assert.Check over assert.Assert to prevent tests covering multiple
cases from failing early
- Fix inconsistent order of expected <--> actual, which made it difficult to
check which output was the expected output.
- Fix formatting of some comments
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
When marshaling the type with `gopkg.in/yaml.v3`, unmarshaling would
recursively call the type's `MarshalYAML()` function, which ultimately
resulted in a crash:
runtime: goroutine stack exceeds 1000000000-byte limit
runtime: sp=0x140202e0430 stack=[0x140202e0000, 0x140402e0000]
fatal error: stack overflow
This applies a similar fix as was implemented in e7788d6f9a
for the `MarshalJSON()` implementation. An alternative would be to use
a type alias (to remove the `MarshalYAML()`), but keeping it simple.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The version was originally added in 570ee9cb54,
at the time the `expected` config did not have a `version:` field. A later
refactor in 0cf2e6353a updated the `expected`
config to have a `version:` included. However, the test was not updated,
which now resulted in the test using a compose file with a duplicate version
field:
version: '3.10'
version: "3.10"
services:
foo:
build:
This issue was masked by `yaml.Unmarshal()` from `gopkg.in/yaml.v2` which
silently ignores the duplicate, taking the value of the last occurrence. When
upgrading to `gopkg.in/yaml.v3`, the duplicate value resulted in an error:
yaml: unmarshal errors:
line 2: mapping key "version" already defined at line 1
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Looks like the linter uses an explicit -lang, which (for go1.19)
results in some additional formatting for octal values.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The existing `remove()` was unused, and using that as name makes it more
consistent with the metadata-store. Also renaming `removeAllEndpointData`
to just `removeEndpoint`, as it's part of the TLS-store, which should already
make it clear it's about (TLS)data.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
There's no reason to stop listing contexts if a context does not exist
while iterating over the directories,
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Go conventions are for interfaces to be defined on the receiver side,
and for producers to return concrete types. This patch changes the
constructor to return a concrete type.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The package defined various special errors; these errors existed for two reasons;
- being able to distinguish "not found" errors from other errors (as "not found"
errors can be ignored in various cases).
- to be able to update the context _name_ in the error message after the error
was created. This was needed in cases where the name was not available at the
location where the error was produced (e.g. only the "id" was present), and
the helpers to detect "not found" errors did not support wrapped errors (so
wrapping the error with a "name" could break logic); a `setContextName` interface
and corresponding `patchErrContextName()` utility was created for this (which
was a "creative", but not very standard approach).
This patch:
- Removes the special error-types, replacing them with errdefs definitions (which
is a more common approach in our code-base to detect error types / classes).
- Removes the internal utilities for error-handling, and deprecates the exported
utilities (to allow external consumers to adjust their code).
- Some errors have been enriched with detailed information (which may be useful
for debugging / problem solving).
- Note that in some cases, `patchErrContextName()` was called, but the code
producing the error would never return a `setContextName` error, so would
never update the error message.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This test was depending on the fact that contextDir's are a string,
and for the test is was using the context _name_ as a pseudo-ID.
This patch updates the test to be more explicit where ID's and where
names are used.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This allows callers to just pass the name, and handle the conversion to ID and
path internally. This also fixes a test which incorrectly used "names" as
pseudo-IDs.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Before this change, running `docker context rm --force` would fail if the context
did not exist. This behavior was different from other commands, which allowed
ignoring non-existing objects.
For example; when trying to remove a non-existing volume, the command would
fail without "force":
```bash
docker volume rm nosuchvolume
Error: No such volume: nosuchvolume
echo $?
1
```
But using the `-f` / `--force` option would make the command complete successfully
(the error itself is still printed for informational purposes);
```bash
docker volume rm -f nosuchvolume
nosuchvolume
echo $?
0
```
With this patch, `docker context rm` behaves the same:
```bash
docker context rm nosuchcontext
context "nosuchcontext" does not exist
echo $?
1
```
```bash
docker context rm -f nosuchcontext
nosuchcontext
echo $?
0
```
This patch also simplifies how we check if the context exists; previously we
would try to read the context's metadata; this could fail if a context was
corrupted, or if an empty directory was present. This patch now only checks
if the directory exists, without first validating the context's data.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Also removing redundant defer for env.PatchAll(), which is now automatically
handled in t.Cleanup()
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Laura Brehm
Sebastiaan van Stijn
Paweł Gronowski
Sebastiaan van Stijn
Nick Santos
Cory Snider
Akihiro Suda
Justin Chadwell
CrazyMax
David Karlsson
Patrick Daigle
Brian Goff
Sam Thibault
Drew Erny
mspiess