_stream: use full element name in `artifact log`/`list-contents` by c1assik · Pull Request #2174 · apache/buildstream

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik · 2026-08-15T09:07:59Z

artifact_log() and artifact_list_contents() were keying their result dicts by obj.name, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. bst artifact list-contents element.bst nested.bst:element.bst would only show one entry, since both targets resolved to the same key.

Use obj._get_full_name() instead, which includes the junction prefix, falling back to the artifact ref for ArtifactElement targets (i.e. when an artifact is referred to by its ref rather than its element name) as before.

Also drops a dead obj.name = {ref: "No artifact cached"} assignment in artifact_list_contents() that was immediately followed by continue and never used.

Add tests covering artifact log and artifact list-contents with a cross-junction element alongside a same-named top-level element.

nathanwilliams-ct · 2026-08-17T10:59:18Z

 continue

- artifact_logs[obj.name] = obj._get_logs()
+ name = ref if isinstance(obj, ArtifactElement) else obj._get_full_name()


Perhaps it would be better if ArtifactElement overrode the _get_full_name method or the __full_name variable, instead of having this awkward condition here?
Suggested change
name=refifisinstance(obj, ArtifactElement) elseobj._get_full_name()
name=obj._get_full_name()
In ArtifactElement
def_get_full_name(self) returnself.get_artifact_name()

Added a method _get_full_name() to ArtifactElement.
return self.get_artifact_name(), unconditionally fails with AssertionError: assert key is not None in get_artifact_name() so if the cache key isn't set yet, fall back to super()._get_full_name() (the base Plugin/Element name, which is always safe since it's precomputed earlier in Plugin.init). Once construction finishes and the key is set, it returns the full artifact ref as intended

I think it makes more sense to set the name in ArtifactElement (you can set it before chaining up to the super class __init__() so that it gets picked up). I feel it would be more elegant than trying to override _get_full_name().

nathanwilliams-ct · 2026-08-18T08:02:35Z

LGTM

`artifact_log()` and `artifact_list_contents()` were keying their result dicts by `obj.name`, which is just the bare, project-relative element name. For elements loaded across a junction, this drops the owning junction prefix, so e.g. `bst artifact list-contents element.bst nested.bst:element.bst` would only show one entry, since both targets resolved to the same key. Use `obj._get_full_name()` instead, which includes the junction prefix, falling back to the artifact ref for `ArtifactElement` targets (i.e. when an artifact is referred to by its ref rather than its element name) as before. Also drops a dead `obj.name = {ref: "No artifact cached"}` assignment in `artifact_list_contents()` that was immediately followed by `continue` and never used. Add tests covering `artifact log` and `artifact list-contents` with a cross-junction element alongside a same-named top-level element.

c1assik marked this pull request as ready for review August 15, 2026 09:08

c1assik requested review from BenjaminSchubert, abderrahim, cs-shadow, gtristan and juergbi as code owners August 15, 2026 09:08

nathanwilliams-ct reviewed Aug 17, 2026
View reviewed changes