Support multiple west config files per config level #867

thorsten-klein · 2025-10-15T13:42:41Z

This is (another) alternative to PR #849. It is easier to discuss this solution while there is already some implementation existing.

Why?

In a big company there are often shared configs for build system and tooling (e.g., west and its configs).
Those generic configs must be applied first and contain some default or fallback values, if the user does not specify any value in any user config.
Depending on the company and team size, multiple different configs need to be stacked.
Remember, that users must always be able to overwrite values from those generic config files within their config.

Proposed Solution

Multiple config files can be specified in the according environment variable (WEST_CONFIG_SYSTEM, WEST_CONFIG_GLOBAL, WEST_CONFIG_LOCAL) for each config level (separated by os.pathsep).
The config files are applied in the same order as they are specified, whereby values from later files override earlier ones.

I personally like configuration through env variables, but I fear you don't.
Nevertheless I have prepared this proposal. Maybe it convinces you.

In a next step, this mechanism could also be combined with the approach from #849, so that multiple configs per config level are supported with a dropin config directory each. This would be the ultimative solution.

marc-hb · 2025-10-15T18:11:56Z

This is (another) alternative to PR #849.

I don't see how this is better than #849 dropins, sorry. It seems to have the same complexity (e.g.: how do you append to an existing value?) but with a less convenient and less standard user interface.

thorsten-klein · 2025-10-15T18:43:52Z

Perfectly fine for me. I just wanted to show it to you for completeness.
~~I will close this PR now 👍🏻~~

marc-hb · 2025-10-15T19:36:16Z

You could have waited for other opinions! They could still come but it's bit less likely in a closed PR.

codecov · 2025-10-16T15:06:15Z

Codecov Report

❌ Patch coverage is 81.81818% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.55%. Comparing base (dca6d0f) to head (3089760).
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/west/configuration.py	81.63%	18 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #867      +/-   ##
==========================================
+ Coverage   84.44%   84.55%   +0.11%     
==========================================
  Files          11       11              
  Lines        3394     3432      +38     
==========================================
+ Hits         2866     2902      +36     
- Misses        528      530       +2

Files with missing lines	Coverage Δ
src/west/app/config.py	`95.04% <100.00%> (+0.99%)`	⬆️
src/west/configuration.py	`83.24% <81.63%> (+0.74%)`	⬆️

... and 1 file with indirect coverage changes

pdgendt

Honestly, I like the simplicity here. A user that sets the environment variable, to change a config file location, already needs to know what they are doing.

Allowing to pass in multiple files is in line with the ConfigParser's functionality.

src/west/configuration.py

pdgendt

@thorsten-klein could you split the --list-paths into it's own PR? Both this and the dropins PR have this addition, we could maybe move ahead with this simpler addition and clean up this and the dropins PR.

src/west/app/config.py

carlescufi

I am fine with this, thanks @thorsten-klein.

tests/test_config.py

marc-hb

I found a couple issues while testing and reviewing but otherwise I agree this is definitely a minimal way to achieve the objective, nice!

marc-hb · 2025-10-23T22:17:09Z

src/west/configuration.py

+        if not self.paths:
+            raise WestNotFound('No config file exists that can be written')
+        if len(self.paths) > 1:
+            raise WestNotFound('Cannot write if multiple configs in use.')


Suggested change

raise WestNotFound('Cannot write if multiple configs in use.')

raise WestNotFound('Cannot write if multiple configs in use:' + os.pathsep.join(self.paths))

marc-hb · 2025-10-23T22:21:21Z

src/west/app/config.py

+For each configuration level (local, global, and system) also multiple config
+file locations can be specified. To do so, set according environment variable
+to contain all paths (separated by 'os.pathsep', which is ';' on Windows or
+':' otherwise)


Suggested change

':' otherwise)

':' otherwise): Latter configuration files have precedence in such lists.

marc-hb · 2025-10-23T22:31:06Z

src/west/configuration.py

+        self.paths = paths
+        read_files = self.cp.read(self.paths, encoding='utf-8')
+        if len(read_files) != len(self.paths):
+            raise FileNotFoundError(paths)


FileNotFoundError is not correct here. When a file in the list does not exist, things should still work and they apparently do (make sure you have a test for that, tests are all about edge cases).

On the other hand, while testing and trying to break this I found that a non-readable file throws this exception.

I also found that this WEST_CONFIG_LOCAL also throws this misleading exception: export WEST_CONFIG_LOCAL=~/.westconfig:$EMPTY_BECAUSE_TYPO

An empty path is apparently converted to "."? That seems wrong too.

Printing a stack trace is confusing here because it hides configparser which is at the root cause.

Finally, you should replace paths with os.pathsep.join(path) because this does not look great:

FileNotFoundError: [PosixPath('fu'), PosixPath('.'), PosixPath('bar')]

So I think you probably want to throw a WestNotFound (no stacktrace?) with a more generic message.

This is not cosmetic: users will misconfigure WEST_CONFIG_LOCAL at some point or another and west should give decent feedback when that happens.

The list is normalized above, so that only paths remain in the list that really exist.
If the count of read files is not matching, something went wrong. I will adapt to raise some WestNotFound instead.
I will also extend the test to use a non-readable file and catch this error.

WEST_CONFIG_LOCAL=~/.westconfig:$EMPTY_VAR_BECAUSE_TYPO. An empty path is apparently converted to "."? That seems wrong too.

The list is normalized above, so that only paths remain in the list that really exist.

I don't know where the "normalization" code was but it apparently didn't mind one of the paths to be the current directory "." (which always exists!)

I didn't test any newer version of this PR yet.

marc-hb · 2025-10-23T22:32:01Z

src/west/configuration.py

+            ret += self._system.paths
        if self._local and location in [ConfigFile.LOCAL, ConfigFile.ALL]:
-            ret.append(self._local.path)
+            ret += self._local.paths


Is this related to this new feature? If not, separate commit please to reduce mental load. Reviewers don't want to guess.

Yes, because now there is a list of paths stored (not a single path).
Therefore I cannot append, but I have to use extend or +=.

For clarity it might be better to use extend here, not blocking though.

marc-hb · 2025-10-23T22:32:35Z

src/west/configuration.py

-
+        paths = self.get_paths(configfile)
+        if len(paths) > 1:
+            raise WestNotFound('Cannot write if multiple configs in use.')


Simple and smart! Nice.

We have this same check at two places. Shall I keep both or only keep one? If yes, which one shall I keep?

I prefer to have the check at the "lowest level", so keep the check in _write at least.

This can be used as a fast return path, or we let it fail later on, both are fine IMO.

We need at both levels since otherwise _create is called with the path1;path2 which of course results in some errors.
We must catch it before we call this function.

marc-hb · 2025-10-23T22:43:59Z

EDIT: moved to new issue #871. I think #871 should be decided before this #867 PR gets merged.

I just thought about another test case: relative paths. As it stands, this PR allows relative paths in WEST_CONFIG_LOCAL and others. This means that the west config depends on the current directory!!

Everyone please correct me but this looks like a really bad idea to me. I can easily imagine users wasting hours or even days because of subtle compilation differences depending on where they compiled the same thing. There are also serious security implications. So I don't think this should be allowed.

This "bug" predates this PR. So it should probably be fixed before it.

marc-hb

(I would still like #871 to be discussed first and fixed separately)

marc-hb · 2025-10-28T02:36:01Z

src/west/configuration.py

+        # Normalize to a list of existing paths
+        paths = path
+        if isinstance(paths, Path):
+            paths = [Path(p) for p in str(paths).split(os.pathsep)]


This is confusing because it works only because a single Path is not always a real Path, sometimes it's a string with a list of paths shoehorned in a single, abused Path object! Which is why you convert here this fake Path into a string, split it and then convert back to a list of actual Path.

Rather than doing this, why not preserve correct typing and move the split to up in the callers? Right now the string -> Path conversion is performed at this point:

self._system_path = Path(_location(ConfigFile.SYSTEM)) self._global_path = Path(_location(ConfigFile.GLOBAL)) ...

After this PR, self._global_path can become something like this: PosixPath('/a/b/c:/d/e/f'). This is not a real Path and very misleading.

Instead you can probably do something like:

self._system_paths = [Path(p) for p in _locations(ConfigFile.SYSTEM)).split(os.pathsep)] ...

(this could be a function like paths_to_str())

This would keep strings as strings, every Path as an actual, single Path and use a proper Path list elsewhere.

You can still perform the sanity checks in the common from_path()

Changing the name of these to a plural self._system_paths and switching to self._system_paths[0] when writing helps makes sure no place ever forgets to check that the length is 1.

BTW it is this line that converts :$EMPTY_VAR: into the current directory.

marc-hb · 2025-10-28T02:36:15Z

src/west/configuration.py

-
-    def __init__(self, path: Path):
-        self.path = path
+    def from_path(path: Path | list[Path] | None) -> '_InternalCF | None':


This is never actually called with a list right now, is it?

I don't think any function or place in the code needs to accept alternatives like this, see below.

marc-hb · 2025-10-28T03:07:25Z

src/west/configuration.py

-
+        paths = self.get_paths(configfile)
+        if len(paths) > 1:
+            raise WestNotFound(f'Cannot write if multiple configs in use: {paths_to_str(paths)}')


Nit: I think the configfile == ConfigFile.ALL: check should be first because it's an internal bug.

BTW the elif on line 303 should be an if.

src/west/configuration.py

pdgendt · 2025-11-03T15:58:21Z

Otherwise I remove the commit from this branch again, so that it can be merged first (only first commit).
I will then immediately open a separate PR for the second commit.

Yes please because I would like more people to have a look at and discuss this.

@thorsten-klein is this still in the works? For me it can go either way, as we've reduced the security concern raised in #871 in #871 (comment)

This PR still only allows absolute paths, while @marc-hb is a proponent of having it relative to the workspace. I don't really have a strong opinion here.

marc-hb · 2025-11-03T22:29:08Z

src/west/configuration.py

+        if not paths:
+            return None
+        if not all(p.is_absolute() for p in paths):
+            raise WestNotFound(f"config file path(s) must be absolute: '{paths_to_str(paths)}'")


I don't think this is the right place to do this because _location() is used in a number of other places.

I'm working on fixing #871.

marc-hb · 2025-11-04T05:02:47Z

I just submitted #877 to fix the relative path issue #871.

The fix is totally different from the one here (EDIT: cause _location() is used in multiple places) and I think it must go first. Please help review, thanks!

src/west/util.py

src/west/configuration.py

tests/conftest.py

marc-hb

Thanks, I think we are getting close! I didn't review every line yet sorry but I think I already have enough comments to share for a first pass.

Note: all the tests must pass in every commit.

marc-hb · 2025-11-06T19:39:28Z

src/west/configuration.py

-    def __init__(self, path: Path):
-        self.path = path
+    def from_paths(paths: list[Path] | None) -> '_InternalCF | None':
+        paths = paths or []


In which case does this line make a difference?

marc-hb · 2025-11-06T19:42:08Z

src/west/configuration.py

+
+    def _write(self):
+        if not self.paths:
+            raise WestNotFound('No config file exists that can be written')


I think this looks like an internal error, not something that could even happen.

Suggested change

raise WestNotFound('No config file exists that can be written')

assert(self.paths)

marc-hb · 2025-11-06T19:47:17Z

src/west/configuration.py


    def _get(self, option, getter):
        section, key = _InternalCF.parse_key(option)
-


Don't mix code changes with whitespace

https://wiki.openstack.org/wiki/GitCommitMessages#Things_to_avoid_when_creating_commits

src/west/configuration.py

marc-hb · 2025-11-06T19:54:56Z

src/west/configuration.py

+            ret.extend(str_list_to_paths(paths))
        except WestNotFound:
            pass
-


Distracting change, remove. This is a complex enough feature.

marc-hb · 2025-11-06T21:00:57Z

src/west/configuration.py

+    if not config_file:
+        raise ValueError(f"no config found for configfile '{configfile}'")
+    if len(config_file) > 1:
+        raise ValueError(f'Cannot write if multiple configs in use: {list_to_str(config_file)}')


Rather than converting back, simply save the _location() output for later.

src/west/app/config.py

marc-hb · 2025-11-06T21:45:05Z

src/west/app/config.py


    def list_paths(self, args):
-        config_paths = self.config.get_paths(args.configfile or ALL)
+        config_paths = self.config.get_paths(args.configfile or ALL, existing_only=True)


This looks like a fix for the previous commit?

Please correct me:

before this PR, --list-paths shows only existing files.

after the previous commit, --list-paths shows some non-existing files.

thanks to this commit --list-paths goes back to showing only existing files.

If correct then just squash with the previous commit and avoid the intermediate state.

marc-hb · 2025-11-06T21:45:43Z

tests/conftest.py

Missing space in 'tmp_west_topdir'function commit subject.

This comment is still valid.

marc-hb · 2025-11-06T21:46:30Z

tests/test_config.py

This must be part of the main commit because all the tests must pass in every commit.

I prefer adding new tests in a separate commit, after the commits with functional changes.

Adding new tests in a later commit: no problem of course.
But not breaking existing tests and fixing them later, that's different.

For obvious git reasons, it must be possible to run and pass tests on every commit.

pdgendt · 2025-11-07T08:00:21Z

tests/test_config.py

I prefer adding new tests in a separate commit, after the commits with functional changes.

pdgendt · 2025-11-07T08:00:53Z

tests/conftest.py

This comment is still valid.

src/west/configuration.py

For easier conversion between types, a configuration._Converter class is created with following static methods: - str_to_paths(paths, sep): helper function to split a 'paths' string at given 'sep' into a list of 'Path' elements - str_list_to_paths(paths): convert each list element from 'str' to 'Path'

pdgendt · 2025-11-07T08:43:28Z

src/west/configuration.py



-def _gather_configs(cfg: ConfigFile, topdir: PathType | None) -> list[str]:
+def _gather_configs(cfg: ConfigFile, topdir: PathType | None) -> list[Path]:


Not sure if we need to change the return type here? You can simply replace append with extend in this function and be done with it.

multiple config files can be specified in the according environment variable for each config level (separated by 'os.pathsep', which is ';' on Windows or ':' otherwise). The config files are applied in the same order as they are specified, whereby values from later files override earlier ones. 'config --list-paths' prints all considered config files which are currently in use and exist. existing tests are adopted to work with the internal change of the return type of _location from 'str' to 'list[str]'

This helper function for tests is added. It temporarily creates a west topdir for the duration of the `with` block by creating a .west directory at given path. The diectory is removed again when the `with` block exits.

Verify that multiple config files can be specified in the according environment variable (WEST_CONFIG_SYSTEM, WEST_CONFIG_GLOBAL, WEST_CONFIG_LOCAL) for each config level.

marc-hb

I think you are almost there.

marc-hb · 2025-11-07T22:53:05Z

src/west/configuration.py

            self.cp.write(f)


+class _Converter:


I miss the value of wrapping two small static methods in this new class. Why not just functions? Or inside some existing class.

https://steve-yegge.blogspot.com/2006/03/execution-in-kingdom-of-nouns.html

Sorry, but I don't see that they belong to class _InternalCF which says

class _InternalCF: # For internal use only; convenience interface for reading and # writing INI-style [section] key = value configuration files, # but presenting a west-style section.key = value style API. # Unlike the higher level and public "Configuration" class, there is # one _InternalCF object for each location: LOCAL, GLOBAL,...

For understanding: Why are just functions better than functions within a (private) class?

marc-hb · 2025-11-07T22:53:38Z

src/west/configuration.py

+class _Converter:
+    @staticmethod
+    def str_to_paths(paths: str | None, sep: str = os.pathsep) -> list[Path]:
+        """Split a string into a list of Path objects using the given separator."""


I'm concerned pydocs automatically make these part of the API, is that correct? They're too basic and unrelated to the API to be part of it.

marc-hb · 2025-11-07T22:55:01Z

src/west/configuration.py


+class _Converter:
+    @staticmethod
+    def str_to_paths(paths: str | None, sep: str = os.pathsep) -> list[Path]:


Suggested change

def str_to_paths(paths: str | None, sep: str = os.pathsep) -> list[Path]:

def parse_paths(paths: str | None, sep: str = os.pathsep) -> list[Path]:

marc-hb · 2025-11-07T23:12:31Z

src/west/configuration.py

+        self.paths = paths
+        read_files = self.cp.read(self.paths, encoding='utf-8')
+        if len(read_files) != len(self.paths):
+            raise WestNotFound(f"Error while reading one of '{paths}'")


Suggested change

raise WestNotFound(f"Error while reading one of '{paths}'")

raise MalformedConfig(f"Error while reading one of '{paths}'")

All self.paths have been verified to .exists(), so "NotFound" is IMHO misleading.

marc-hb · 2025-11-07T23:26:39Z

src/west/configuration.py


-        with open(self.path, 'w', encoding='utf-8') as f:
-            self.cp.write(f)
+        self._write()


Sorry but I don't like the fact that you "tentatively" change the value in memory and fail only at the last moment before actually writing. This means the _InternatCF state in memory is "impossible" for some duration. It's OK now but then it's reasonably likely to become a problem in the future whenever someone refactoring happens for some other feature or re-org.

I think you can avoid most of the duplication like this:

def set() / append() / etc.: self.single_file_check() # raises exception ... ... self._write() def _write(self): assert len(self.paths) == 1 # (in case some function forgot single_file_check) with open(self.paths[0], 'w', encoding='utf-8') as f: self.cp.write(f) def single_file_check(): assert paths len(self.paths) != 1 raise ValueError(f'Cannot write if multiple configs in use: {self.paths}')

marc-hb · 2025-11-07T23:39:00Z

tests/test_config.py

+    write_config(config_l1, 'sec', 'l', '1', 'l1', '1')
+    write_config(config_l2, 'sec', 'l', '2', 'l2', '2')
+
+    # use a non-readable config file (does not work on Windows)


Suggested change

# use a non-readable config file (does not work on Windows)

# config file without read permission (does not work on Windows)

marc-hb · 2025-11-07T23:39:52Z

tests/test_config.py

+        config_non_readable.touch()
+        config_non_readable.chmod(0o000)
+        os.environ["WEST_CONFIG_GLOBAL"] = f'{config_g1}{os.pathsep}{config_non_readable}'
+        _, stderr = cmd_raises('config --global some.section', WestNotFound)


"NotFound" is not right...

marc-hb · 2025-11-07T23:43:08Z

tests/test_config.py

+    config_g1 = config_dir / 'global 1'
+    config_g2 = config_dir / 'global 2'
+    config_l1 = config_dir / 'local 1'
+    config_l2 = config_dir / 'local 2'


Since you already imported Windows, try to break things using the "other" os.pathsep

if WINDOWS: other_sep = ':' else other_sep = ';' config_l2 = config_dir / 'local 2' + other_sep

marc-hb · 2025-11-07T23:45:49Z

tests/test_config.py

+    stdout = cmd('config --local sec.l1').rstrip()
+    assert stdout == '1'
+    stdout = cmd('config --local sec.l2').rstrip()
+    assert stdout == '2'


That's a lot of repetition. Use an array of key,value.

marc-hb · 2025-11-07T23:48:46Z

tests/test_config.py

+    msg = "'{file}' is relative but 'west topdir' is not defined"
+    cwd = pathlib.Path('any cwd')
+    cwd.mkdir()
+    with chdir(cwd):


I think you could have more coverage with less test code by:

having some relative files in the test mix, always

having some top-level test function that runs the ENTIRE test suite twice in two (or three) different directories. These tests do not involve git so I believe they are really quick to run.

marc-hb

On one hand you want to keep filtering non-existing paths early, at _InternalCF() construction time because it's very convenient.
On the other hand, now you need to remember non-existing files because you (rightfully) want to block modifications of WEST_CONFIG_x=existing_fileA:non_existing_fileB.

So I think that's why you added an new existing_only boolean parameter to the get_paths() method but that confused me a lot because it has a default value. Also, depending on that parameter get_paths() now perform again the .exists() filtering that was already performed when constructing the _InternalCF() objects... a bit confusing.

I think it would be much clearer to have two different functions:

get_path_definitions() that includes non-existing files
get_existing_paths() that returns only existing files from already filtered _InternalCF() objects.

I bet they won't have any code in common.

Build/compute things only once, name them really well and then re-use them wherever they are needed.

BTW do you have a test that makes sure changes to WEST_CONFIG_x=existing_fileA:non_existing_fileB are blocked?

marc-hb · 2025-11-08T00:50:24Z

src/west/configuration.py

+        if location in [ConfigFile.LOCAL, ConfigFile.ALL]:
+            ret.extend(self._local_paths)
+        if existing_only:
+            ret = [p for p in ret if p.exists()]


I don't see in which case this can make a difference. The self._*_paths have already been filtered in the _InternalCF.from_paths() (it's the reason why from_paths() exists). I understand why you want this but it does not seem to be happening. Do you have test coverage with missing files? Sorry I did not review the whole test code yet.

EDIT: this is probably one of the places where the existing/non-existing confusion hit me.

marc-hb · 2025-11-08T00:57:47Z

src/west/configuration.py

+            if not configs:
+                raise WestNotFound(f'{configfile}: Cannot determine any config file')
+            if len(configs) > 1:
+                raise ValueError(f'Cannot set value if multiple configs in use: {configs}')


This exception is raised in multiple places: raise it from a single "check_single" very small function.

marc-hb · 2025-11-08T01:04:12Z

src/west/configuration.py

+    all_config_files = [_location(cfg, topdir=topdir) for cfg in considered_locations]
+    for config_files in all_config_files:
+        if len(config_files) > 1:
+            raise ValueError(f'Error: Multiple config paths in use: {config_files}')


Is this too restrictive? Let's say I have two GLOBAL files and a single LOCAL file. Now this looks like this blocks me from deleting things in my single LOCAL file, doesn't it?

Maybe this can be done like this:

Look for the value in every "location" (GLOBAL,...) requested. Remember where it was found. Do NOT delete anything yet!

For each location where it was actually found in step 1, make sure there is single file in the definition of that location. Raise an exception if not. Do not delete anything yet.

Finally and atomically delete it from all the location where it was found (and requested).

Clearly explain this (or whatever better algorithm you find) in a comment :-)

Write some good tests for it.

It restricts the configs only for considered_locations, which is [LOCAL] if --local was given. Nevertheless I have added a test for this to ensure that writing of a config is possible (even if the other "levels" use multiple configs)

marc-hb · 2025-11-08T01:11:08Z

src/west/configuration.py

+            '''
+            configs = self.get_paths(location)
+            if not configs:
+                raise WestNotFound(f'{configfile}: Cannot determine any config file')


You passed existing_only=False (the default), so I don't see how this could be raised. This looks like an assert, not an exception.

thorsten-klein marked this pull request as draft October 15, 2025 13:42

marc-hb mentioned this pull request Oct 15, 2025

Support 4th level west -c config.ini file(s) #429

Open

thorsten-klein closed this Oct 15, 2025

thorsten-klein reopened this Oct 15, 2025

thorsten-klein force-pushed the west-multiple-configs branch 4 times, most recently from e33851f to 8f92d18 Compare October 16, 2025 14:55

pdgendt reviewed Oct 16, 2025

View reviewed changes

src/west/configuration.py Outdated Show resolved Hide resolved

thorsten-klein force-pushed the west-multiple-configs branch from 8f92d18 to 3c9c122 Compare October 16, 2025 18:38

pdgendt reviewed Oct 20, 2025

View reviewed changes

src/west/app/config.py Outdated Show resolved Hide resolved

thorsten-klein mentioned this pull request Oct 21, 2025

support 'west config --list-paths' #870

Merged

thorsten-klein force-pushed the west-multiple-configs branch from 3c9c122 to 2fd4885 Compare October 22, 2025 16:18

thorsten-klein changed the title ~~[alternative] Support multiple west config files per config level~~ Support multiple west config files per config level Oct 23, 2025

thorsten-klein marked this pull request as ready for review October 23, 2025 14:12

carlescufi approved these changes Oct 23, 2025

View reviewed changes

marc-hb reviewed Oct 23, 2025

View reviewed changes

tests/test_config.py Outdated Show resolved Hide resolved

marc-hb reviewed Oct 23, 2025

View reviewed changes

tests/test_config.py Outdated Show resolved Hide resolved

marc-hb requested changes Oct 23, 2025

View reviewed changes

marc-hb mentioned this pull request Oct 23, 2025

WEST_CONFIG_* accept relative paths #871

Closed

thorsten-klein force-pushed the west-multiple-configs branch 5 times, most recently from 54cfdea to 032f4bb Compare October 24, 2025 12:01

marc-hb reviewed Oct 28, 2025

View reviewed changes

thorsten-klein force-pushed the west-multiple-configs branch from 620351a to 33dee41 Compare October 28, 2025 16:36

pdgendt reviewed Nov 3, 2025

View reviewed changes

src/west/configuration.py Outdated Show resolved Hide resolved

src/west/configuration.py Outdated Show resolved Hide resolved

thorsten-klein force-pushed the west-multiple-configs branch from 33dee41 to d220ba7 Compare November 3, 2025 16:54

marc-hb reviewed Nov 3, 2025

View reviewed changes

thorsten-klein force-pushed the west-multiple-configs branch from d220ba7 to 8178ece Compare November 4, 2025 06:10

This was referenced Nov 4, 2025

override remote urls thorsten-klein/west#2

Open

Allow modifying remote/project urls of imported manifests #869

Draft

thorsten-klein force-pushed the west-multiple-configs branch 2 times, most recently from 7d86ab8 to 0756f8b Compare November 6, 2025 14:27

pdgendt reviewed Nov 6, 2025

View reviewed changes

src/west/util.py Outdated Show resolved Hide resolved

src/west/configuration.py Show resolved Hide resolved

tests/conftest.py Outdated Show resolved Hide resolved

thorsten-klein force-pushed the west-multiple-configs branch from 0756f8b to bbc808f Compare November 6, 2025 17:30

marc-hb requested changes Nov 6, 2025

View reviewed changes

thorsten-klein force-pushed the west-multiple-configs branch 3 times, most recently from d2efd35 to dca2152 Compare November 7, 2025 07:59

pdgendt reviewed Nov 7, 2025

View reviewed changes

thorsten-klein force-pushed the west-multiple-configs branch 2 times, most recently from 9ee38dc to d4691bd Compare November 7, 2025 08:12

pdgendt reviewed Nov 7, 2025

View reviewed changes

thorsten-klein force-pushed the west-multiple-configs branch from d4691bd to 71ba34d Compare November 7, 2025 17:08

thorsten-klein added 3 commits November 7, 2025 18:09

add 'tmp_west_topdir' function for usage in tests

8e877d6

This helper function for tests is added. It temporarily creates a west topdir for the duration of the `with` block by creating a .west directory at given path. The diectory is removed again when the `with` block exits.

add tests for multiple west config files

3089760

Verify that multiple config files can be specified in the according environment variable (WEST_CONFIG_SYSTEM, WEST_CONFIG_GLOBAL, WEST_CONFIG_LOCAL) for each config level.

thorsten-klein force-pushed the west-multiple-configs branch from 71ba34d to 3089760 Compare November 7, 2025 17:09

marc-hb reviewed Nov 7, 2025

View reviewed changes

marc-hb reviewed Nov 8, 2025

View reviewed changes

	raise WestNotFound('Cannot write if multiple configs in use.')
	raise WestNotFound('Cannot write if multiple configs in use:' + os.pathsep.join(self.paths))

	':' otherwise)
	':' otherwise): Latter configuration files have precedence in such lists.

	raise WestNotFound('No config file exists that can be written')
	assert(self.paths)


		def _get(self, option, getter):
		section, key = _InternalCF.parse_key(option)



		def _gather_configs(cfg: ConfigFile, topdir: PathType \| None) -> list[str]:
		def _gather_configs(cfg: ConfigFile, topdir: PathType \| None) -> list[Path]:

	def str_to_paths(paths: str \| None, sep: str = os.pathsep) -> list[Path]:
	def parse_paths(paths: str \| None, sep: str = os.pathsep) -> list[Path]:

	raise WestNotFound(f"Error while reading one of '{paths}'")
	raise MalformedConfig(f"Error while reading one of '{paths}'")

	# use a non-readable config file (does not work on Windows)
	# config file without read permission (does not work on Windows)

Support multiple west config files per config level #867

Are you sure you want to change the base?

Support multiple west config files per config level #867

Conversation

thorsten-klein commented Oct 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why?

Proposed Solution

Uh oh!

marc-hb commented Oct 15, 2025

Uh oh!

thorsten-klein commented Oct 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

marc-hb commented Oct 15, 2025

Uh oh!

codecov bot commented Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

pdgendt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

pdgendt left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

carlescufi left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

marc-hb left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

marc-hb commented Oct 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

marc-hb left a comment

Choose a reason for hiding this comment

Uh oh!

marc-hb Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

pdgendt commented Nov 3, 2025

Uh oh!

thorsten-klein commented Oct 15, 2025 •

edited

Loading

thorsten-klein commented Oct 15, 2025 •

edited

Loading

codecov bot commented Oct 16, 2025 •

edited

Loading

marc-hb commented Oct 23, 2025 •

edited

Loading

marc-hb Oct 28, 2025 •

edited

Loading

marc-hb commented Nov 4, 2025 •

edited

Loading

marc-hb left a comment •

edited

Loading

marc-hb Nov 6, 2025 •

edited

Loading