[Notes] [Git][BuildStream/buildstream][raoul/802-refactor-artifactcache] 3 commits: artifactcache: add top level tmpdir for CASRemote

[Raoul Hidalgo Charman <gitlab mg gitlab com>]

Title: GitLab

Raoul Hidalgo Charman pushed to branch raoul/802-refactor-artifactcache at BuildStream / buildstream

Commits:

• 115b5593
  by Raoul Hidalgo Charman at 2019-01-14T11:06:49Z

  artifactcache: add top level tmpdir for CASRemote
  
  As CASRemote is used in other places such as SandboxRemote it makes sense for
  there to be a top level tmpdir it uses. This dir is passed to CASRemote when
  initialized.
  
  Part of #802
  

• 993f9957
  by Raoul Hidalgo Charman at 2019-01-14T11:11:19Z

  artifactcache: Move pull logic into CASRemote
  
  Seperates the pull logic into a remote/local API, so that artifact cache
  iterates over blob digests checks whether it has them, and then requests them
  if not. The request command allows batching of blobs where appropriate.
  
  Tests have been updated to ensure the correct tmpdir is set up in process
  wrappers, else invalid cross link errors happen in the CI. Additional asserts
  have been added to check that the temporary directories are cleared by the end
  of a pull.
  
  Part of #802
  

• 3a000f6c
  by Raoul Hidalgo Charman at 2019-01-14T11:11:34Z

  artifactcache: implement new push methods
  
  Similar to the pull methods, this implements a yield_directory_digests methods
  that iterates over blobs in the local CAS, with the upload_blob sending blobs
  to a remote and batching them where appropriate.
  
  Part of #802
  

13 changed files:

• buildstream/_artifactcache.py
• buildstream/_cas/__init__.py
• buildstream/_cas/cascache.py
• buildstream/_cas/casremote.py
• + buildstream/_cas/transfer.py
• buildstream/_context.py
• buildstream/data/userconfig.yaml
• buildstream/sandbox/_sandboxremote.py
• conftest.py
• tests/artifactcache/pull.py
• tests/artifactcache/push.py
• tests/integration/pullbuildtrees.py
• tests/testutils/runcli.py

Changes:

buildstream/_artifactcache.py

---

...	...	@@ -28,7 +28,8 @@ from ._message import Message, MessageType
28	28	from . import utils
29	29	from . import _yaml
30	30
31		-from ._cas import CASRemote, CASRemoteSpec
	31	+from ._cas import BlobNotFound, CASRemote, CASRemoteSpec
	32	+from ._cas.transfer import cas_directory_upload, cas_directory_download, cas_tree_download
32	33
33	34
34	35	CACHE_SIZE_FILE = "cache_size"
...	...	@@ -374,7 +375,7 @@ class ArtifactCache():
374	375	q = multiprocessing.Queue()
375	376	for remote_spec in remote_specs:
376	377
377		- error = CASRemote.check_remote(remote_spec, q)
	378	+ error = CASRemote.check_remote(remote_spec, self.context.tmpdir, q)
378	379
379	380	if error and on_failure:
380	381	on_failure(remote_spec.url, error)
...	...	@@ -385,7 +386,7 @@ class ArtifactCache():
385	386	if remote_spec.push:
386	387	self._has_push_remotes = True
387	388
388		- remotes[remote_spec.url] = CASRemote(remote_spec)
	389	+ remotes[remote_spec.url] = CASRemote(remote_spec, self.context.tmpdir)
389	390
390	391	for project in self.context.get_projects():
391	392	remote_specs = self.global_remote_specs
...	...	@@ -607,16 +608,41 @@ class ArtifactCache():
607	608
608	609	for remote in push_remotes:
609	610	remote.init()
	611	+ skipped_remote = True
610	612	display_key = element._get_brief_display_key()
611	613	element.status("Pushing artifact {} -> {}".format(display_key, remote.spec.url))
612	614
613		- if self.cas.push(refs, remote):
614		- element.info("Pushed artifact {} -> {}".format(display_key, remote.spec.url))
	615	+ try:
	616	+ for ref in refs:
	617	+ # Check whether ref is already on the server in which case
	618	+ # there is no need to push the ref
	619	+ root_digest = self.cas.resolve_ref(ref)
	620	+ response = remote.get_reference(ref)
	621	+ if (response is not None and
	622	+ response.hash == root_digest.hash and
	623	+ response.size_bytes == root_digest.size_bytes):
	624	+ element.info("Remote ({}) already has {} cached".format(
	625	+ remote.spec.url, element._get_brief_display_key()))
	626	+ continue
	627	+
	628	+ # upload blobs
	629	+ cas_directory_upload(self.cas, remote, root_digest)
	630	+ remote.update_reference(ref, root_digest)
	631	+
	632	+ skipped_remote = False
	633	+
	634	+ except CASError as e:
	635	+ if str(e.reason) == "StatusCode.RESOURCE_EXHAUSTED":
	636	+ element.warn("Failed to push element to {}: Resource exhuasted"
	637	+ .format(remote.spec.url))
	638	+ continue
	639	+ else:
	640	+ raise ArtifactError("Failed to push refs {}: {}".format(refs, e),
	641	+ temporary=True) from e
	642	+
	643	+ if skipped_remote is False:
615	644	pushed = True
616		- else:
617		- element.info("Remote ({}) already has {} cached".format(
618		- remote.spec.url, element._get_brief_display_key()
619		- ))
	645	+ element.info("Pushed artifact {} -> {}".format(display_key, remote.spec.url))
620	646
621	647	return pushed
622	648
...	...	@@ -644,19 +670,31 @@ class ArtifactCache():
644	670	display_key = element._get_brief_display_key()
645	671	element.status("Pulling artifact {} <- {}".format(display_key, remote.spec.url))
646	672
647		- if self.cas.pull(ref, remote, progress=progress, subdir=subdir, excluded_subdirs=excluded_subdirs):
648		- element.info("Pulled artifact {} <- {}".format(display_key, remote.spec.url))
649		- if subdir:
650		- # Attempt to extract subdir into artifact extract dir if it already exists
651		- # without containing the subdir. If the respective artifact extract dir does not
652		- # exist a complete extraction will complete.
653		- self.extract(element, key, subdir)
654		- # no need to pull from additional remotes
655		- return True
656		- else:
	673	+ root_digest = remote.get_reference(ref)
	674	+
	675	+ if not root_digest:
657	676	element.info("Remote ({}) does not have {} cached".format(
658		- remote.spec.url, element._get_brief_display_key()
659		- ))
	677	+ remote.spec.url, element._get_brief_display_key()))
	678	+ continue
	679	+
	680	+ try:
	681	+ cas_directory_download(self.cas, remote, root_digest, excluded_subdirs)
	682	+ except BlobNotFound:
	683	+ element.info("Remote ({}) is missing blobs for {}".format(
	684	+ remote.spec.url, element._get_brief_display_key()))
	685	+ continue
	686	+
	687	+ self.cas.set_ref(ref, root_digest)
	688	+
	689	+ if subdir:
	690	+ # Attempt to extract subdir into artifact extract dir if it already exists
	691	+ # without containing the subdir. If the respective artifact extract dir does not
	692	+ # exist a complete extraction will complete.
	693	+ self.extract(element, key, subdir)
	694	+
	695	+ element.info("Pulled artifact {} <- {}".format(display_key, remote.spec.url))
	696	+ # no need to pull from additional remotes
	697	+ return True
660	698
661	699	except CASError as e:
662	700	raise ArtifactError("Failed to pull artifact {}: {}".format(
...	...	@@ -671,15 +709,16 @@ class ArtifactCache():
671	709	#
672	710	# Args:
673	711	# project (Project): The current project
674		- # digest (Digest): The digest of the tree
	712	+ # tree_digest (Digest): The digest of the tree
675	713	#
676		- def pull_tree(self, project, digest):
	714	+ def pull_tree(self, project, tree_digest):
677	715	for remote in self._remotes[project]:
678		- digest = self.cas.pull_tree(remote, digest)
679		-
680		- if digest:
681		- # no need to pull from additional remotes
682		- return digest
	716	+ try:
	717	+ root_digest = cas_tree_download(self.cas, remote, tree_digest)
	718	+ except BlobNotFound:
	719	+ continue
	720	+ else:
	721	+ return root_digest
683	722
684	723	return None
685	724
...	...	@@ -708,7 +747,7 @@ class ArtifactCache():
708	747	return
709	748
710	749	for remote in push_remotes:
711		- self.cas.push_directory(remote, directory)
	750	+ cas_directory_upload(self.cas, remote, directory.ref)
712	751
713	752	# push_message():
714	753	#
...	...	@@ -917,6 +956,19 @@ class ArtifactCache():
917	956	stat = os.statvfs(volume)
918	957	return stat.f_bsize * stat.f_bavail, stat.f_bsize * stat.f_blocks
919	958
	959	+ def _fetch_directory(self, remote, root_digest, excluded_subdirs):
	960	+ for blob_digest in remote.yield_directory_digests(
	961	+ root_digest, excluded_subdirs=excluded_subdirs):
	962	+ if self.cas.check_blob(blob_digest):
	963	+ continue
	964	+ remote.request_blob(blob_digest)
	965	+ for blob_file in remote.get_blobs():
	966	+ self.cas.add_object(path=blob_file.name, link_directly=True)
	967	+
	968	+ # Request final CAS batch
	969	+ for blob_file in remote.get_blobs(complete_batch=True):
	970	+ self.cas.add_object(path=blob_file.name, link_directly=True)
	971	+
920	972
921	973	# _configured_remote_artifact_cache_specs():
922	974	#

buildstream/_cas/__init__.py

---

...	...	@@ -18,4 +18,4 @@
18	18	# Tristan Van Berkom <tristan vanberkom codethink co uk>
19	19
20	20	from .cascache import CASCache
21		-from .casremote import CASRemote, CASRemoteSpec
	21	+from .casremote import CASRemote, CASRemoteSpec, BlobNotFound

buildstream/_cas/cascache.py

---

...	...	@@ -18,23 +18,16 @@
18	18	# Jürg Billeter <juerg billeter codethink co uk>
19	19
20	20	import hashlib
21		-import itertools
22	21	import os
23	22	import stat
24	23	import tempfile
25		-import uuid
26	24	import contextlib
27	25
28		-import grpc
29		-
30	26	from .._protos.build.bazel.remote.execution.v2 import remote_execution_pb2
31		-from .._protos.buildstream.v2 import buildstream_pb2
32	27
33	28	from .. import utils
34	29	from .._exceptions import CASCacheError
35	30
36		-from .casremote import BlobNotFound, _CASBatchRead, _CASBatchUpdate
37		-
38	31
39	32	# A CASCache manages a CAS repository as specified in the Remote Execution API.
40	33	#
...	...	@@ -183,73 +176,6 @@ class CASCache():
183	176
184	177	return modified, removed, added
185	178
186		- # pull():
187		- #
188		- # Pull a ref from a remote repository.
189		- #
190		- # Args:
191		- # ref (str): The ref to pull
192		- # remote (CASRemote): The remote repository to pull from
193		- # progress (callable): The progress callback, if any
194		- # subdir (str): The optional specific subdir to pull
195		- # excluded_subdirs (list): The optional list of subdirs to not pull
196		- #
197		- # Returns:
198		- # (bool): True if pull was successful, False if ref was not available
199		- #
200		- def pull(self, ref, remote, *, progress=None, subdir=None, excluded_subdirs=None):
201		- try:
202		- remote.init()
203		-
204		- request = buildstream_pb2.GetReferenceRequest(instance_name=remote.spec.instance_name)
205		- request.key = ref
206		- response = remote.ref_storage.GetReference(request)
207		-
208		- tree = remote_execution_pb2.Digest()
209		- tree.hash = response.digest.hash
210		- tree.size_bytes = response.digest.size_bytes
211		-
212		- # Check if the element artifact is present, if so just fetch the subdir.
213		- if subdir and os.path.exists(self.objpath(tree)):
214		- self._fetch_subdir(remote, tree, subdir)
215		- else:
216		- # Fetch artifact, excluded_subdirs determined in pullqueue
217		- self._fetch_directory(remote, tree, excluded_subdirs=excluded_subdirs)
218		-
219		- self.set_ref(ref, tree)
220		-
221		- return True
222		- except grpc.RpcError as e:
223		- if e.code() != grpc.StatusCode.NOT_FOUND:
224		- raise CASCacheError("Failed to pull ref {}: {}".format(ref, e)) from e
225		- else:
226		- return False
227		- except BlobNotFound as e:
228		- return False
229		-
230		- # pull_tree():
231		- #
232		- # Pull a single Tree rather than a ref.
233		- # Does not update local refs.
234		- #
235		- # Args:
236		- # remote (CASRemote): The remote to pull from
237		- # digest (Digest): The digest of the tree
238		- #
239		- def pull_tree(self, remote, digest):
240		- try:
241		- remote.init()
242		-
243		- digest = self._fetch_tree(remote, digest)
244		-
245		- return digest
246		-
247		- except grpc.RpcError as e:
248		- if e.code() != grpc.StatusCode.NOT_FOUND:
249		- raise
250		-
251		- return None
252		-
253	179	# link_ref():
254	180	#
255	181	# Add an alias for an existing ref.
...	...	@@ -263,73 +189,6 @@ class CASCache():
263	189
264	190	self.set_ref(newref, tree)
265	191
266		- # push():
267		- #
268		- # Push committed refs to remote repository.
269		- #
270		- # Args:
271		- # refs (list): The refs to push
272		- # remote (CASRemote): The remote to push to
273		- #
274		- # Returns:
275		- # (bool): True if any remote was updated, False if no pushes were required
276		- #
277		- # Raises:
278		- # (CASCacheError): if there was an error
279		- #
280		- def push(self, refs, remote):
281		- skipped_remote = True
282		- try:
283		- for ref in refs:
284		- tree = self.resolve_ref(ref)
285		-
286		- # Check whether ref is already on the server in which case
287		- # there is no need to push the ref
288		- try:
289		- request = buildstream_pb2.GetReferenceRequest(instance_name=remote.spec.instance_name)
290		- request.key = ref
291		- response = remote.ref_storage.GetReference(request)
292		-
293		- if response.digest.hash == tree.hash and response.digest.size_bytes == tree.size_bytes:
294		- # ref is already on the server with the same tree
295		- continue
296		-
297		- except grpc.RpcError as e:
298		- if e.code() != grpc.StatusCode.NOT_FOUND:
299		- # Intentionally re-raise RpcError for outer except block.
300		- raise
301		-
302		- self._send_directory(remote, tree)
303		-
304		- request = buildstream_pb2.UpdateReferenceRequest(instance_name=remote.spec.instance_name)
305		- request.keys.append(ref)
306		- request.digest.hash = tree.hash
307		- request.digest.size_bytes = tree.size_bytes
308		- remote.ref_storage.UpdateReference(request)
309		-
310		- skipped_remote = False
311		- except grpc.RpcError as e:
312		- if e.code() != grpc.StatusCode.RESOURCE_EXHAUSTED:
313		- raise CASCacheError("Failed to push ref {}: {}".format(refs, e), temporary=True) from e
314		-
315		- return not skipped_remote
316		-
317		- # push_directory():
318		- #
319		- # Push the given virtual directory to a remote.
320		- #
321		- # Args:
322		- # remote (CASRemote): The remote to push to
323		- # directory (Directory): A virtual directory object to push.
324		- #
325		- # Raises:
326		- # (CASCacheError): if there was an error
327		- #
328		- def push_directory(self, remote, directory):
329		- remote.init()
330		-
331		- self._send_directory(remote, directory.ref)
332		-
333	192	# objpath():
334	193	#
335	194	# Return the path of an object based on its digest.
...	...	@@ -591,6 +450,37 @@ class CASCache():
591	450	reachable = set()
592	451	self._reachable_refs_dir(reachable, tree, update_mtime=True)
593	452
	453	+ # Check to see if a blob is in the local CAS
	454	+ # return None if not
	455	+ def check_blob(self, digest):
	456	+ objpath = self.objpath(digest)
	457	+ if os.path.exists(objpath):
	458	+ # already in local repository
	459	+ return objpath
	460	+ else:
	461	+ return None
	462	+
	463	+ def yield_directory_digests(self, directory_digest):
	464	+ # parse directory, and recursively add blobs
	465	+ d = remote_execution_pb2.Digest()
	466	+ d.hash = directory_digest.hash
	467	+ d.size_bytes = directory_digest.size_bytes
	468	+ yield d
	469	+
	470	+ directory = remote_execution_pb2.Directory()
	471	+
	472	+ with open(self.objpath(directory_digest), 'rb') as f:
	473	+ directory.ParseFromString(f.read())
	474	+
	475	+ for filenode in directory.files:
	476	+ d = remote_execution_pb2.Digest()
	477	+ d.hash = filenode.digest.hash
	478	+ d.size_bytes = filenode.digest.size_bytes
	479	+ yield d
	480	+
	481	+ for dirnode in directory.directories:
	482	+ yield from self.yield_directory_digests(dirnode.digest)
	483	+
594	484	################################################
595	485	# Local Private Methods #
596	486	################################################
...	...	@@ -779,202 +669,3 @@ class CASCache():
779	669
780	670	for dirnode in directory.directories:
781	671	yield from self._required_blobs(dirnode.digest)
782		-
783		- # _ensure_blob():
784		- #
785		- # Fetch and add blob if it's not already local.
786		- #
787		- # Args:
788		- # remote (Remote): The remote to use.
789		- # digest (Digest): Digest object for the blob to fetch.
790		- #
791		- # Returns:
792		- # (str): The path of the object
793		- #
794		- def _ensure_blob(self, remote, digest):
795		- objpath = self.objpath(digest)
796		- if os.path.exists(objpath):
797		- # already in local repository
798		- return objpath
799		-
800		- with tempfile.NamedTemporaryFile(dir=self.tmpdir) as f:
801		- remote._fetch_blob(digest, f)
802		-
803		- added_digest = self.add_object(path=f.name, link_directly=True)
804		- assert added_digest.hash == digest.hash
805		-
806		- return objpath
807		-
808		- def _batch_download_complete(self, batch):
809		- for digest, data in batch.send():
810		- with tempfile.NamedTemporaryFile(dir=self.tmpdir) as f:
811		- f.write(data)
812		- f.flush()
813		-
814		- added_digest = self.add_object(path=f.name, link_directly=True)
815		- assert added_digest.hash == digest.hash
816		-
817		- # Helper function for _fetch_directory().
818		- def _fetch_directory_batch(self, remote, batch, fetch_queue, fetch_next_queue):
819		- self._batch_download_complete(batch)
820		-
821		- # All previously scheduled directories are now locally available,
822		- # move them to the processing queue.
823		- fetch_queue.extend(fetch_next_queue)
824		- fetch_next_queue.clear()
825		- return _CASBatchRead(remote)
826		-
827		- # Helper function for _fetch_directory().
828		- def _fetch_directory_node(self, remote, digest, batch, fetch_queue, fetch_next_queue, *, recursive=False):
829		- in_local_cache = os.path.exists(self.objpath(digest))
830		-
831		- if in_local_cache:
832		- # Skip download, already in local cache.
833		- pass
834		- elif (digest.size_bytes >= remote.max_batch_total_size_bytes or
835		- not remote.batch_read_supported):
836		- # Too large for batch request, download in independent request.
837		- self._ensure_blob(remote, digest)
838		- in_local_cache = True
839		- else:
840		- if not batch.add(digest):
841		- # Not enough space left in batch request.
842		- # Complete pending batch first.
843		- batch = self._fetch_directory_batch(remote, batch, fetch_queue, fetch_next_queue)
844		- batch.add(digest)
845		-
846		- if recursive:
847		- if in_local_cache:
848		- # Add directory to processing queue.
849		- fetch_queue.append(digest)
850		- else:
851		- # Directory will be available after completing pending batch.
852		- # Add directory to deferred processing queue.
853		- fetch_next_queue.append(digest)
854		-
855		- return batch
856		-
857		- # _fetch_directory():
858		- #
859		- # Fetches remote directory and adds it to content addressable store.
860		- #
861		- # Fetches files, symbolic links and recursively other directories in
862		- # the remote directory and adds them to the content addressable
863		- # store.
864		- #
865		- # Args:
866		- # remote (Remote): The remote to use.
867		- # dir_digest (Digest): Digest object for the directory to fetch.
868		- # excluded_subdirs (list): The optional list of subdirs to not fetch
869		- #
870		- def _fetch_directory(self, remote, dir_digest, *, excluded_subdirs=None):
871		- fetch_queue = [dir_digest]
872		- fetch_next_queue = []
873		- batch = _CASBatchRead(remote)
874		- if not excluded_subdirs:
875		- excluded_subdirs = []
876		-
877		- while len(fetch_queue) + len(fetch_next_queue) > 0:
878		- if not fetch_queue:
879		- batch = self._fetch_directory_batch(remote, batch, fetch_queue, fetch_next_queue)
880		-
881		- dir_digest = fetch_queue.pop(0)
882		-
883		- objpath = self._ensure_blob(remote, dir_digest)
884		-
885		- directory = remote_execution_pb2.Directory()
886		- with open(objpath, 'rb') as f:
887		- directory.ParseFromString(f.read())
888		-
889		- for dirnode in directory.directories:
890		- if dirnode.name not in excluded_subdirs:
891		- batch = self._fetch_directory_node(remote, dirnode.digest, batch,
892		- fetch_queue, fetch_next_queue, recursive=True)
893		-
894		- for filenode in directory.files:
895		- batch = self._fetch_directory_node(remote, filenode.digest, batch,
896		- fetch_queue, fetch_next_queue)
897		-
898		- # Fetch final batch
899		- self._fetch_directory_batch(remote, batch, fetch_queue, fetch_next_queue)
900		-
901		- def _fetch_subdir(self, remote, tree, subdir):
902		- subdirdigest = self._get_subdir(tree, subdir)
903		- self._fetch_directory(remote, subdirdigest)
904		-
905		- def _fetch_tree(self, remote, digest):
906		- # download but do not store the Tree object
907		- with tempfile.NamedTemporaryFile(dir=self.tmpdir) as out:
908		- remote._fetch_blob(digest, out)
909		-
910		- tree = remote_execution_pb2.Tree()
911		-
912		- with open(out.name, 'rb') as f:
913		- tree.ParseFromString(f.read())
914		-
915		- tree.children.extend([tree.root])
916		- for directory in tree.children:
917		- for filenode in directory.files:
918		- self._ensure_blob(remote, filenode.digest)
919		-
920		- # place directory blob only in final location when we've downloaded
921		- # all referenced blobs to avoid dangling references in the repository
922		- dirbuffer = directory.SerializeToString()
923		- dirdigest = self.add_object(buffer=dirbuffer)
924		- assert dirdigest.size_bytes == len(dirbuffer)
925		-
926		- return dirdigest
927		-
928		- def _send_directory(self, remote, digest, u_uid=uuid.uuid4()):
929		- required_blobs = self._required_blobs(digest)
930		-
931		- missing_blobs = dict()
932		- # Limit size of FindMissingBlobs request
933		- for required_blobs_group in _grouper(required_blobs, 512):
934		- request = remote_execution_pb2.FindMissingBlobsRequest(instance_name=remote.spec.instance_name)
935		-
936		- for required_digest in required_blobs_group:
937		- d = request.blob_digests.add()
938		- d.hash = required_digest.hash
939		- d.size_bytes = required_digest.size_bytes
940		-
941		- response = remote.cas.FindMissingBlobs(request)
942		- for missing_digest in response.missing_blob_digests:
943		- d = remote_execution_pb2.Digest()
944		- d.hash = missing_digest.hash
945		- d.size_bytes = missing_digest.size_bytes
946		- missing_blobs[d.hash] = d
947		-
948		- # Upload any blobs missing on the server
949		- self._send_blobs(remote, missing_blobs.values(), u_uid)
950		-
951		- def _send_blobs(self, remote, digests, u_uid=uuid.uuid4()):
952		- batch = _CASBatchUpdate(remote)
953		-
954		- for digest in digests:
955		- with open(self.objpath(digest), 'rb') as f:
956		- assert os.fstat(f.fileno()).st_size == digest.size_bytes
957		-
958		- if (digest.size_bytes >= remote.max_batch_total_size_bytes or
959		- not remote.batch_update_supported):
960		- # Too large for batch request, upload in independent request.
961		- remote._send_blob(digest, f, u_uid=u_uid)
962		- else:
963		- if not batch.add(digest, f):
964		- # Not enough space left in batch request.
965		- # Complete pending batch first.
966		- batch.send()
967		- batch = _CASBatchUpdate(remote)
968		- batch.add(digest, f)
969		-
970		- # Send final batch
971		- batch.send()
972		-
973		-
974		-def _grouper(iterable, n):
975		- while True:
976		- try:
977		- current = next(iterable)
978		- except StopIteration:
979		- return
980		- yield itertools.chain([current], itertools.islice(iterable, n - 1))

buildstream/_cas/casremote.py

---

1	1	from collections import namedtuple
2	2	import io
	3	+import itertools
3	4	import os
4	5	import multiprocessing
5	6	import signal
	7	+import tempfile
6	8	from urllib.parse import urlparse
7	9	import uuid
8	10
...	...	@@ -79,7 +81,7 @@ class BlobNotFound(CASRemoteError):
79	81	# Represents a single remote CAS cache.
80	82	#
81	83	class CASRemote():
82		- def __init__(self, spec):
	84	+ def __init__(self, spec, tmpdir):
83	85	self.spec = spec
84	86	self._initialized = False
85	87	self.channel = None
...	...	@@ -91,6 +93,16 @@ class CASRemote():
91	93	self.capabilities = None
92	94	self.max_batch_total_size_bytes = None
93	95
	96	+ # Need str because python 3.5 and lower doesn't deal with path like
	97	+ # objects here.
	98	+ self.tmpdir = str(tmpdir)
	99	+ os.makedirs(self.tmpdir, exist_ok=True)
	100	+
	101	+ self.__tmp_downloads = [] # files in the tmpdir waiting to be added to local caches
	102	+
	103	+ self.__batch_read = None
	104	+ self.__batch_update = None
	105	+
94	106	def init(self):
95	107	if not self._initialized:
96	108	url = urlparse(self.spec.url)
...	...	@@ -148,6 +160,7 @@ class CASRemote():
148	160	request = remote_execution_pb2.BatchReadBlobsRequest()
149	161	response = self.cas.BatchReadBlobs(request)
150	162	self.batch_read_supported = True
	163	+ self.__batch_read = _CASBatchRead(self)
151	164	except grpc.RpcError as e:
152	165	if e.code() != grpc.StatusCode.UNIMPLEMENTED:
153	166	raise
...	...	@@ -158,6 +171,7 @@ class CASRemote():
158	171	request = remote_execution_pb2.BatchUpdateBlobsRequest()
159	172	response = self.cas.BatchUpdateBlobs(request)
160	173	self.batch_update_supported = True
	174	+ self.__batch_update = _CASBatchUpdate(self)
161	175	except grpc.RpcError as e:
162	176	if (e.code() != grpc.StatusCode.UNIMPLEMENTED and
163	177	e.code() != grpc.StatusCode.PERMISSION_DENIED):
...	...	@@ -172,11 +186,11 @@ class CASRemote():
172	186	# in the main BuildStream process
173	187	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
174	188	@classmethod
175		- def check_remote(cls, remote_spec, q):
	189	+ def check_remote(cls, remote_spec, tmpdir, q):
176	190
177	191	def __check_remote():
178	192	try:
179		- remote = cls(remote_spec)
	193	+ remote = cls(remote_spec, tmpdir)
180	194	remote.init()
181	195
182	196	request = buildstream_pb2.StatusRequest()
...	...	@@ -254,6 +268,210 @@ class CASRemote():
254	268
255	269	return message_digest
256	270
	271	+ # get_reference():
	272	+ #
	273	+ # Args:
	274	+ # ref (str): The ref to request
	275	+ #
	276	+ # Returns:
	277	+ # (digest): digest of ref, None if not found
	278	+ #
	279	+ def get_reference(self, ref):
	280	+ try:
	281	+ self.init()
	282	+
	283	+ request = buildstream_pb2.GetReferenceRequest()
	284	+ request.key = ref
	285	+ return self.ref_storage.GetReference(request).digest
	286	+ except grpc.RpcError as e:
	287	+ if e.code() != grpc.StatusCode.NOT_FOUND:
	288	+ raise CASRemoteError("Failed to find ref {}: {}".format(ref, e)) from e
	289	+ else:
	290	+ return None
	291	+
	292	+ # update_reference():
	293	+ #
	294	+ # Args:
	295	+ # ref (str): Reference to update
	296	+ # digest (Digest): New digest to update ref with
	297	+ def update_reference(self, ref, digest):
	298	+ request = buildstream_pb2.UpdateReferenceRequest()
	299	+ request.keys.append(ref)
	300	+ request.digest.hash = digest.hash
	301	+ request.digest.size_bytes = digest.size_bytes
	302	+ self.ref_storage.UpdateReference(request)
	303	+
	304	+ def get_tree_blob(self, tree_digest):
	305	+ self.init()
	306	+ f = tempfile.NamedTemporaryFile(dir=self.tmpdir)
	307	+ self._fetch_blob(tree_digest, f)
	308	+
	309	+ tree = remote_execution_pb2.Tree()
	310	+ with open(f.name, 'rb') as tmp:
	311	+ tree.ParseFromString(tmp.read())
	312	+
	313	+ return tree
	314	+
	315	+ # yield_directory_digests():
	316	+ #
	317	+ # Recursively iterates over digests for files, symbolic links and other
	318	+ # directories starting from a root digest
	319	+ #
	320	+ # Args:
	321	+ # root_digest (digest): The root_digest to get a tree of
	322	+ # progress (callable): The progress callback, if any
	323	+ # subdir (str): The optional specific subdir to pull
	324	+ # excluded_subdirs (list): The optional list of subdirs to not pull
	325	+ #
	326	+ # Returns:
	327	+ # (iter digests): recursively iterates over digests contained in root directory
	328	+ #
	329	+ def yield_directory_digests(self, root_digest, *, progress=None,
	330	+ subdir=None, excluded_subdirs=None):
	331	+ self.init()
	332	+
	333	+ # Fetch artifact, excluded_subdirs determined in pullqueue
	334	+ if excluded_subdirs is None:
	335	+ excluded_subdirs = []
	336	+
	337	+ # get directory blob
	338	+ f = tempfile.NamedTemporaryFile(dir=self.tmpdir)
	339	+ self._fetch_blob(root_digest, f)
	340	+
	341	+ directory = remote_execution_pb2.Directory()
	342	+ with open(f.name, 'rb') as tmp:
	343	+ directory.ParseFromString(tmp.read())
	344	+
	345	+ yield root_digest
	346	+ for filenode in directory.files:
	347	+ yield filenode.digest
	348	+
	349	+ for dirnode in directory.directories:
	350	+ if dirnode.name not in excluded_subdirs:
	351	+ yield from self.yield_directory_digests(dirnode.digest)
	352	+
	353	+ # yield_tree_digests():
	354	+ #
	355	+ # Fetches a tree file from digests and then iterates over child digests
	356	+ #
	357	+ # Args:
	358	+ # tree_digest (digest): tree digest
	359	+ #
	360	+ # Returns:
	361	+ # (iter digests): iterates over digests in tree message
	362	+ def yield_tree_digests(self, tree):
	363	+ self.init()
	364	+
	365	+ tree.children.extend([tree.root])
	366	+ for directory in tree.children:
	367	+ for filenode in directory.files:
	368	+ yield filenode.digest
	369	+
	370	+ # add the directory to downloaded tmp files to be added
	371	+ f = tempfile.NamedTemporaryFile(dir=self.tmpdir)
	372	+ f.write(directory.SerializeToString())
	373	+ f.flush()
	374	+ self.__tmp_downloads.append(f)
	375	+
	376	+ # request_blob():
	377	+ #
	378	+ # Request blob, triggering download depending via bytestream or cas
	379	+ # BatchReadBlobs depending on size.
	380	+ #
	381	+ # Args:
	382	+ # digest (Digest): digest of the requested blob
	383	+ #
	384	+ def request_blob(self, digest):
	385	+ if (not self.batch_read_supported or
	386	+ digest.size_bytes > self.max_batch_total_size_bytes):
	387	+ f = tempfile.NamedTemporaryFile(dir=self.tmpdir)
	388	+ self._fetch_blob(digest, f)
	389	+ self.__tmp_downloads.append(f)
	390	+ elif self.__batch_read.add(digest) is False:
	391	+ self._download_batch()
	392	+ self.__batch_read.add(digest)
	393	+
	394	+ # get_blobs():
	395	+ #
	396	+ # Yield over downloaded blobs in the tmp file locations, causing the files
	397	+ # to be deleted once they go out of scope.
	398	+ #
	399	+ # Args:
	400	+ # complete_batch (bool): download any outstanding batch read request
	401	+ #
	402	+ # Returns:
	403	+ # iterator over NamedTemporaryFile
	404	+ def get_blobs(self, complete_batch=False):
	405	+ # Send read batch request and download
	406	+ if (complete_batch is True and
	407	+ self.batch_read_supported is True):
	408	+ self._download_batch()
	409	+
	410	+ while self.__tmp_downloads:
	411	+ yield self.__tmp_downloads.pop()
	412	+
	413	+ # upload_blob():
	414	+ #
	415	+ # Push blobs given an iterator over blob files
	416	+ #
	417	+ # Args:
	418	+ # digest (Digest): digest we want to upload
	419	+ # blob_file (str): Name of file location
	420	+ # u_uid (str): Used to identify to the bytestream service
	421	+ #
	422	+ def upload_blob(self, digest, blob_file, u_uid=uuid.uuid4()):
	423	+ with open(blob_file, 'rb') as f:
	424	+ assert os.fstat(f.fileno()).st_size == digest.size_bytes
	425	+
	426	+ if (digest.size_bytes >= self.max_batch_total_size_bytes or
	427	+ not self.batch_update_supported):
	428	+ # Too large for batch request, upload in independent request.
	429	+ self._send_blob(digest, f, u_uid=u_uid)
	430	+ else:
	431	+ if self.__batch_update.add(digest, f) is False:
	432	+ self.__batch_update.send()
	433	+ self.__batch_update = _CASBatchUpdate(self)
	434	+ self.__batch_update.add(digest, f)
	435	+
	436	+ # send_update_batch():
	437	+ #
	438	+ # Sends anything left in the update batch
	439	+ #
	440	+ def send_update_batch(self):
	441	+ # make sure everything is sent
	442	+ self.__batch_update.send()
	443	+ self.__batch_update = _CASBatchUpdate(self)
	444	+
	445	+ # find_missing_blobs()
	446	+ #
	447	+ # Does FindMissingBlobs request to remote
	448	+ #
	449	+ # Args:
	450	+ # required_blobs ([Digest]): list of blobs required
	451	+ #
	452	+ # Returns:
	453	+ # (Dict(Digest)): missing blobs
	454	+ def find_missing_blobs(self, required_blobs):
	455	+ self.init()
	456	+ missing_blobs = dict()
	457	+ # Limit size of FindMissingBlobs request
	458	+ for required_blobs_group in _grouper(required_blobs, 512):
	459	+ request = remote_execution_pb2.FindMissingBlobsRequest()
	460	+
	461	+ for required_digest in required_blobs_group:
	462	+ d = request.blob_digests.add()
	463	+ d.hash = required_digest.hash
	464	+ d.size_bytes = required_digest.size_bytes
	465	+
	466	+ response = self.cas.FindMissingBlobs(request)
	467	+ for missing_digest in response.missing_blob_digests:
	468	+ d = remote_execution_pb2.Digest()
	469	+ d.hash = missing_digest.hash
	470	+ d.size_bytes = missing_digest.size_bytes
	471	+ missing_blobs[d.hash] = d
	472	+
	473	+ return missing_blobs
	474	+
257	475	################################################
258	476	# Local Private Methods #
259	477	################################################
...	...	@@ -292,10 +510,31 @@ class CASRemote():
292	510	offset += chunk_size
293	511	finished = request.finish_write
294	512
295		- response = self.bytestream.Write(request_stream(resource_name, stream))
	513	+ try:
	514	+ response = self.bytestream.Write(request_stream(resource_name, stream))
	515	+ except grpc.RpcError as e:
	516	+ raise CASRemoteError("Failed to upload blob: {}".format(e), reason=e.code())
296	517
297	518	assert response.committed_size == digest.size_bytes
298	519
	520	+ def _download_batch(self):
	521	+ for _, data in self.__batch_read.send():
	522	+ f = tempfile.NamedTemporaryFile(dir=self.tmpdir)
	523	+ f.write(data)
	524	+ f.flush()
	525	+ self.__tmp_downloads.append(f)
	526	+
	527	+ self.__batch_read = _CASBatchRead(self)
	528	+
	529	+
	530	+def _grouper(iterable, n):
	531	+ while True:
	532	+ try:
	533	+ current = next(iterable)
	534	+ except StopIteration:
	535	+ return
	536	+ yield itertools.chain([current], itertools.islice(iterable, n - 1))
	537	+
299	538
300	539	# Represents a batch of blobs queued for fetching.
301	540	#
...	...	@@ -328,7 +567,11 @@ class _CASBatchRead():
328	567	if not self._request.digests:
329	568	return
330	569
331		- batch_response = self._remote.cas.BatchReadBlobs(self._request)
	570	+ try:
	571	+ batch_response = self._remote.cas.BatchReadBlobs(self._request)
	572	+ except grpc.RpcError as e:
	573	+ raise CASRemoteError("Failed to read blob batch: {}".format(e),
	574	+ reason=e.code()) from e
332	575
333	576	for response in batch_response.responses:
334	577	if response.status.code == code_pb2.NOT_FOUND:
...	...	@@ -376,7 +619,12 @@ class _CASBatchUpdate():
376	619	if not self._request.requests:
377	620	return
378	621
379		- batch_response = self._remote.cas.BatchUpdateBlobs(self._request)
	622	+ # Want to raise a CASRemoteError if
	623	+ try:
	624	+ batch_response = self._remote.cas.BatchUpdateBlobs(self._request)
	625	+ except grpc.RpcError as e:
	626	+ raise CASRemoteError("Failed to upload blob batch: {}".format(e),
	627	+ reason=e.code()) from e
380	628
381	629	for response in batch_response.responses:
382	630	if response.status.code != code_pb2.OK:

buildstream/_cas/transfer.py

---

	1	+#
	2	+# Copyright (C) 2017-2018 Codethink Limited
	3	+#
	4	+# This program is free software; you can redistribute it and/or
	5	+# modify it under the terms of the GNU Lesser General Public
	6	+# License as published by the Free Software Foundation; either
	7	+# version 2 of the License, or (at your option) any later version.
	8	+#
	9	+# This library is distributed in the hope that it will be useful,
	10	+# but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	12	+# Lesser General Public License for more details.
	13	+#
	14	+# You should have received a copy of the GNU Lesser General Public
	15	+# License along with this library. If not, see <http://www.gnu.org/licenses/>.
	16	+#
	17	+# Authors:
	18	+# Raoul Hidalgo Charman <raoul hidalgocharman codethink co uk>
	19	+
	20	+from ..utils import _message_digest
	21	+
	22	+
	23	+def cas_directory_download(caslocal, casremote, root_digest, excluded_subdirs):
	24	+ for blob_digest in casremote.yield_directory_digests(
	25	+ root_digest, excluded_subdirs=excluded_subdirs):
	26	+ if caslocal.check_blob(blob_digest):
	27	+ continue
	28	+ casremote.request_blob(blob_digest)
	29	+ for blob_file in casremote.get_blobs():
	30	+ caslocal.add_object(path=blob_file.name, link_directly=True)
	31	+
	32	+ # Request final CAS batch
	33	+ for blob_file in casremote.get_blobs(complete_batch=True):
	34	+ caslocal.add_object(path=blob_file.name, link_directly=True)
	35	+
	36	+
	37	+def cas_tree_download(caslocal, casremote, tree_digest):
	38	+ tree = casremote.get_tree_blob(tree_digest)
	39	+ for blob_digest in casremote.yield_tree_digests(tree):
	40	+ if caslocal.check_blob(blob_digest):
	41	+ continue
	42	+ casremote.request_blob(blob_digest)
	43	+ for blob_file in casremote.get_blobs():
	44	+ caslocal.add_object(path=blob_file.name, link_directly=True)
	45	+
	46	+ # Get the last batch
	47	+ for blob_file in casremote.get_blobs(complete_batch=True):
	48	+ caslocal.add_object(path=blob_file.name, link_directly=True)
	49	+
	50	+ # get root digest from tree and return that
	51	+ return _message_digest(tree.root.SerializeToString())
	52	+
	53	+
	54	+def cas_directory_upload(caslocal, casremote, root_digest):
	55	+ required_blobs = caslocal.yield_directory_digests(root_digest)
	56	+ missing_blobs = casremote.find_missing_blobs(required_blobs)
	57	+ for blob in missing_blobs.values():
	58	+ blob_file = caslocal.objpath(blob)
	59	+ casremote.upload_blob(blob, blob_file)
	60	+
	61	+ # send remaining blobs
	62	+ casremote.send_update_batch()

buildstream/_context.py

---

...	...	@@ -191,10 +191,11 @@ class Context():
191	191	_yaml.node_validate(defaults, [
192	192	'sourcedir', 'builddir', 'artifactdir', 'logdir',
193	193	'scheduler', 'artifacts', 'logging', 'projects',
194		- 'cache', 'prompt', 'workspacedir', 'remote-execution'
	194	+ 'cache', 'prompt', 'workspacedir', 'remote-execution', 'tmpdir',
195	195	])
196	196
197		- for directory in ['sourcedir', 'builddir', 'artifactdir', 'logdir', 'workspacedir']:
	197	+ for directory in ['sourcedir', 'builddir', 'artifactdir', 'logdir',
	198	+ 'tmpdir', 'workspacedir']:
198	199	# Allow the ~ tilde expansion and any environment variables in
199	200	# path specification in the config files.
200	201	#

buildstream/data/userconfig.yaml

---

...	...	@@ -19,6 +19,9 @@ builddir: ${XDG_CACHE_HOME}/buildstream/build
19	19	# Location to store local binary artifacts
20	20	artifactdir: ${XDG_CACHE_HOME}/buildstream/artifacts
21	21
	22	+# Location to store temporary files, e.g. used when downloading from a casremote
	23	+tmpdir: ${XDG_CACHE_HOME}/buildstream/tmp
	24	+
22	25	# Location to store build logs
23	26	logdir: ${XDG_CACHE_HOME}/buildstream/logs
24	27

buildstream/sandbox/_sandboxremote.py

---

...	...	@@ -39,6 +39,7 @@ from .._exceptions import SandboxError
39	39	from .. import _yaml
40	40	from .._protos.google.longrunning import operations_pb2, operations_pb2_grpc
41	41	from .._cas import CASRemote, CASRemoteSpec
	42	+from .._cas.transfer import cas_tree_download, cas_directory_upload
42	43
43	44
44	45	class RemoteExecutionSpec(namedtuple('RemoteExecutionSpec', 'exec_service storage_service action_service')):
...	...	@@ -279,10 +280,9 @@ class SandboxRemote(Sandbox):
279	280
280	281	context = self._get_context()
281	282	cascache = context.get_cascache()
282		- casremote = CASRemote(self.storage_remote_spec)
	283	+ casremote = CASRemote(self.storage_remote_spec, context.tmpdir)
283	284
284		- # Now do a pull to ensure we have the necessary parts.
285		- dir_digest = cascache.pull_tree(casremote, tree_digest)
	285	+ dir_digest = cas_tree_download(cascache, casremote, tree_digest)
286	286	if dir_digest is None or not dir_digest.hash or not dir_digest.size_bytes:
287	287	raise SandboxError("Output directory structure pulling from remote failed.")
288	288
...	...	@@ -306,8 +306,9 @@ class SandboxRemote(Sandbox):
306	306
307	307	def _run(self, command, flags, *, cwd, env):
308	308	# set up virtual dircetory
	309	+ context = self._get_context()
309	310	upload_vdir = self.get_virtual_directory()
310		- cascache = self._get_context().get_cascache()
	311	+ cascache = context.get_cascache()
311	312	if isinstance(upload_vdir, FileBasedDirectory):
312	313	# Make a new temporary directory to put source in
313	314	upload_vdir = CasBasedDirectory(cascache, ref=None)
...	...	@@ -340,11 +341,11 @@ class SandboxRemote(Sandbox):
340	341	action_result = self._check_action_cache(action_digest)
341	342
342	343	if not action_result:
343		- casremote = CASRemote(self.storage_remote_spec)
	344	+ casremote = CASRemote(self.storage_remote_spec, context.tmpdir)
344	345
345	346	# Now, push that key (without necessarily needing a ref) to the remote.
346	347	try:
347		- cascache.push_directory(casremote, upload_vdir)
	348	+ cas_directory_upload(cascache, casremote, upload_vdir.ref)
348	349	except grpc.RpcError as e:
349	350	raise SandboxError("Failed to push source directory to remote: {}".format(e)) from e
350	351

conftest.py

---

...	...	@@ -50,8 +50,10 @@ def integration_cache(request):
50	50
51	51	# Clean up the artifacts after each test run - we only want to
52	52	# cache sources
	53	+ # Also clean up the tmp dir, should be empty but isn't in CI tests
53	54	try:
54	55	shutil.rmtree(os.path.join(cache_dir, 'artifacts'))
	56	+ shutil.rmtree(os.path.join(cache_dir, 'tmp'))
55	57	except FileNotFoundError:
56	58	pass
57	59

tests/artifactcache/pull.py

---

...	...	@@ -110,7 +110,7 @@ def test_pull(cli, tmpdir, datafiles):
110	110	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
111	111	process = multiprocessing.Process(target=_queue_wrapper,
112	112	args=(_test_pull, queue, user_config_file, project_dir,
113		- artifact_dir, 'target.bst', element_key))
	113	+ artifact_dir, tmpdir, 'target.bst', element_key))
114	114
115	115	try:
116	116	# Keep SIGINT blocked in the child process
...	...	@@ -126,14 +126,18 @@ def test_pull(cli, tmpdir, datafiles):
126	126	assert not error
127	127	assert cas.contains(element, element_key)
128	128
	129	+ # Check that the tmp dir is cleared out
	130	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
129	131
130		-def _test_pull(user_config_file, project_dir, artifact_dir,
	132	+
	133	+def _test_pull(user_config_file, project_dir, artifact_dir, tmpdir,
131	134	element_name, element_key, queue):
132	135	# Fake minimal context
133	136	context = Context()
134	137	context.load(config=user_config_file)
135	138	context.artifactdir = artifact_dir
136	139	context.set_message_handler(message_handler)
	140	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
137	141
138	142	# Load the project manually
139	143	project = Project(project_dir, context)
...	...	@@ -218,7 +222,7 @@ def test_pull_tree(cli, tmpdir, datafiles):
218	222	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
219	223	process = multiprocessing.Process(target=_queue_wrapper,
220	224	args=(_test_push_tree, queue, user_config_file, project_dir,
221		- artifact_dir, artifact_digest))
	225	+ artifact_dir, tmpdir, artifact_digest))
222	226
223	227	try:
224	228	# Keep SIGINT blocked in the child process
...	...	@@ -239,6 +243,9 @@ def test_pull_tree(cli, tmpdir, datafiles):
239	243	# Assert that we are not cached locally anymore
240	244	assert cli.get_element_state(project_dir, 'target.bst') != 'cached'
241	245
	246	+ # Check that the tmp dir is cleared out
	247	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
	248	+
242	249	tree_digest = remote_execution_pb2.Digest(hash=tree_hash,
243	250	size_bytes=tree_size)
244	251
...	...	@@ -246,7 +253,7 @@ def test_pull_tree(cli, tmpdir, datafiles):
246	253	# Use subprocess to avoid creation of gRPC threads in main BuildStream process
247	254	process = multiprocessing.Process(target=_queue_wrapper,
248	255	args=(_test_pull_tree, queue, user_config_file, project_dir,
249		- artifact_dir, tree_digest))
	256	+ artifact_dir, tmpdir, tree_digest))
250	257
251	258	try:
252	259	# Keep SIGINT blocked in the child process
...	...	@@ -267,13 +274,18 @@ def test_pull_tree(cli, tmpdir, datafiles):
267	274	# Ensure the entire Tree stucture has been pulled
268	275	assert os.path.exists(cas.objpath(directory_digest))
269	276
	277	+ # Check that the tmp dir is cleared out
	278	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
	279	+
270	280
271		-def _test_push_tree(user_config_file, project_dir, artifact_dir, artifact_digest, queue):
	281	+def _test_push_tree(user_config_file, project_dir, artifact_dir, tmpdir,
	282	+ artifact_digest, queue):
272	283	# Fake minimal context
273	284	context = Context()
274	285	context.load(config=user_config_file)
275	286	context.artifactdir = artifact_dir
276	287	context.set_message_handler(message_handler)
	288	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
277	289
278	290	# Load the project manually
279	291	project = Project(project_dir, context)
...	...	@@ -304,12 +316,14 @@ def _test_push_tree(user_config_file, project_dir, artifact_dir, artifact_digest
304	316	queue.put("No remote configured")
305	317
306	318
307		-def _test_pull_tree(user_config_file, project_dir, artifact_dir, artifact_digest, queue):
	319	+def _test_pull_tree(user_config_file, project_dir, artifact_dir, tmpdir,
	320	+ artifact_digest, queue):
308	321	# Fake minimal context
309	322	context = Context()
310	323	context.load(config=user_config_file)
311	324	context.artifactdir = artifact_dir
312	325	context.set_message_handler(message_handler)
	326	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
313	327
314	328	# Load the project manually
315	329	project = Project(project_dir, context)

tests/artifactcache/push.py

---

...	...	@@ -89,7 +89,7 @@ def test_push(cli, tmpdir, datafiles):
89	89	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
90	90	process = multiprocessing.Process(target=_queue_wrapper,
91	91	args=(_test_push, queue, user_config_file, project_dir,
92		- artifact_dir, 'target.bst', element_key))
	92	+ artifact_dir, tmpdir, 'target.bst', element_key))
93	93
94	94	try:
95	95	# Keep SIGINT blocked in the child process
...	...	@@ -105,14 +105,18 @@ def test_push(cli, tmpdir, datafiles):
105	105	assert not error
106	106	assert share.has_artifact('test', 'target.bst', element_key)
107	107
	108	+ # Check tmpdir for downloads is cleared
	109	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
108	110
109		-def _test_push(user_config_file, project_dir, artifact_dir,
	111	+
	112	+def _test_push(user_config_file, project_dir, artifact_dir, tmpdir,
110	113	element_name, element_key, queue):
111	114	# Fake minimal context
112	115	context = Context()
113	116	context.load(config=user_config_file)
114	117	context.artifactdir = artifact_dir
115	118	context.set_message_handler(message_handler)
	119	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
116	120
117	121	# Load the project manually
118	122	project = Project(project_dir, context)
...	...	@@ -196,9 +200,10 @@ def test_push_directory(cli, tmpdir, datafiles):
196	200	queue = multiprocessing.Queue()
197	201	# Use subprocess to avoid creation of gRPC threads in main BuildStream process
198	202	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
199		- process = multiprocessing.Process(target=_queue_wrapper,
200		- args=(_test_push_directory, queue, user_config_file,
201		- project_dir, artifact_dir, artifact_digest))
	203	+ process = multiprocessing.Process(
	204	+ target=_queue_wrapper,
	205	+ args=(_test_push_directory, queue, user_config_file, project_dir,
	206	+ artifact_dir, tmpdir, artifact_digest))
202	207
203	208	try:
204	209	# Keep SIGINT blocked in the child process
...	...	@@ -215,13 +220,17 @@ def test_push_directory(cli, tmpdir, datafiles):
215	220	assert artifact_digest.hash == directory_hash
216	221	assert share.has_object(artifact_digest)
217	222
	223	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
218	224
219		-def _test_push_directory(user_config_file, project_dir, artifact_dir, artifact_digest, queue):
	225	+
	226	+def _test_push_directory(user_config_file, project_dir, artifact_dir, tmpdir,
	227	+ artifact_digest, queue):
220	228	# Fake minimal context
221	229	context = Context()
222	230	context.load(config=user_config_file)
223	231	context.artifactdir = artifact_dir
224	232	context.set_message_handler(message_handler)
	233	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
225	234
226	235	# Load the project manually
227	236	project = Project(project_dir, context)
...	...	@@ -273,7 +282,7 @@ def test_push_message(cli, tmpdir, datafiles):
273	282	# See https://github.com/grpc/grpc/blob/master/doc/fork_support.md for details
274	283	process = multiprocessing.Process(target=_queue_wrapper,
275	284	args=(_test_push_message, queue, user_config_file,
276		- project_dir, artifact_dir))
	285	+ project_dir, artifact_dir, tmpdir))
277	286
278	287	try:
279	288	# Keep SIGINT blocked in the child process
...	...	@@ -291,13 +300,16 @@ def test_push_message(cli, tmpdir, datafiles):
291	300	size_bytes=message_size)
292	301	assert share.has_object(message_digest)
293	302
	303	+ assert os.listdir(os.path.join(str(tmpdir), 'cache', 'tmp')) == []
	304	+
294	305
295		-def _test_push_message(user_config_file, project_dir, artifact_dir, queue):
	306	+def _test_push_message(user_config_file, project_dir, artifact_dir, tmpdir, queue):
296	307	# Fake minimal context
297	308	context = Context()
298	309	context.load(config=user_config_file)
299	310	context.artifactdir = artifact_dir
300	311	context.set_message_handler(message_handler)
	312	+ context.tmpdir = os.path.join(str(tmpdir), 'cache', 'tmp')
301	313
302	314	# Load the project manually
303	315	project = Project(project_dir, context)

tests/integration/pullbuildtrees.py

---

...	...	@@ -23,6 +23,7 @@ def default_state(cli, tmpdir, share):
23	23	'artifacts': {'url': share.repo, 'push': False},
24	24	'artifactdir': os.path.join(str(tmpdir), 'artifacts'),
25	25	'cache': {'pull-buildtrees': False},
	26	+ 'tmpdir': os.path.join(str(tmpdir), 'cache', 'tmp'),
26	27	})
27	28
28	29
...	...	@@ -79,6 +80,9 @@ def test_pullbuildtrees(cli, tmpdir, datafiles, integration_cache):
79	80	assert os.path.isdir(buildtreedir)
80	81	default_state(cli, tmpdir, share1)
81	82
	83	+ # Check tmpdir for downloads is cleared
	84	+ assert os.listdir(os.path.join(integration_cache, 'tmp')) == []
	85	+
82	86	# Pull artifact with pullbuildtrees set in user config, then assert
83	87	# that pulling with the same user config doesn't creates a pull job,
84	88	# or when buildtrees cli flag is set.
...	...	@@ -91,6 +95,9 @@ def test_pullbuildtrees(cli, tmpdir, datafiles, integration_cache):
91	95	assert element_name not in result.get_pulled_elements()
92	96	default_state(cli, tmpdir, share1)
93	97
	98	+ # Check tmpdir for downloads is cleared
	99	+ assert os.listdir(os.path.join(integration_cache, 'tmp')) == []
	100	+
94	101	# Pull artifact with default config and buildtrees cli flag set, then assert
95	102	# that pulling with pullbuildtrees set in user config doesn't create a pull
96	103	# job.
...	...	@@ -101,6 +108,9 @@ def test_pullbuildtrees(cli, tmpdir, datafiles, integration_cache):
101	108	assert element_name not in result.get_pulled_elements()
102	109	default_state(cli, tmpdir, share1)
103	110
	111	+ # Check tmpdir for downloads is cleared
	112	+ assert os.listdir(os.path.join(integration_cache, 'tmp')) == []
	113	+
104	114	# Assert that a partial build element (not containing a populated buildtree dir)
105	115	# can't be pushed to an artifact share, then assert that a complete build element
106	116	# can be. This will attempt a partial pull from share1 and then a partial push

tests/testutils/runcli.py

---

...	...	@@ -526,7 +526,8 @@ def cli_integration(tmpdir, integration_cache):
526	526	# to avoid downloading the huge base-sdk repeatedly
527	527	fixture.configure({
528	528	'sourcedir': os.path.join(integration_cache, 'sources'),
529		- 'artifactdir': os.path.join(integration_cache, 'artifacts')
	529	+ 'artifactdir': os.path.join(integration_cache, 'artifacts'),
	530	+ 'tmpdir': os.path.join(integration_cache, 'tmp')
530	531	})
531	532
532	533	return fixture
...	...	@@ -573,6 +574,8 @@ def configured(directory, config=None):
573	574	config['builddir'] = os.path.join(directory, 'build')
574	575	if not config.get('artifactdir', False):
575	576	config['artifactdir'] = os.path.join(directory, 'artifacts')
	577	+ if not config.get('tmpdir', False):
	578	+ config['tmpdir'] = os.path.join(directory, 'tmp')
576	579	if not config.get('logdir', False):
577	580	config['logdir'] = os.path.join(directory, 'logs')
578	581

—
View it on GitLab.
You're receiving this email because of your account on gitlab.com. If you'd like to receive fewer emails, you can adjust your notification settings.