def shortlog_area_prefix(shortlog):
'''Get the prefix of a shortlog which describes its area.
This returns the "raw" prefix as it appears in the shortlog. To
canonicalize this to one of a known set of areas, use
shortlog_area() instead. If no prefix is present, returns None.
'''
# Base case for recursion.
if not shortlog:
return None
# 'Revert "foo"' should map to foo's area prefix.
if shortlog_is_revert(shortlog):
shortlog = shortlog_reverts_what(shortlog)
return shortlog_area_prefix(shortlog)
# If there is no ':', there is no area. Otherwise, the candidate
# area is the substring up to the first ':'.
if ':' not in shortlog:
return None
area, rest = [s.strip() for s in shortlog.split(':', 1)]
# subsys: foo should map to foo's area prefix, etc.
if area in ['subsys', 'include', 'api']:
return shortlog_area_prefix(rest)
return area
def _downstream_outstanding_commits(self):
# Compute a list of commit objects for outstanding
# downstream patches.
# Convert downstream and upstream revisions to SHAs.
dsha = self._dp.sha(self._dr)
usha = self._up.sha(self._ur)
# First, get a list of all downstream OOT patches. Note:
# pygit2 doesn't seem to have any ready-made rev-list
# equivalent, so call out to Project.git() to get the commit
# SHAs, then wrap them with pygit2 objects.
cp = self._dp.git('rev-list --reverse {} ^{}'.format(dsha, usha),
capture_stdout=True)
if not cp.stdout.strip():
return []
commit_shas = cp.stdout.decode('utf-8').strip().splitlines()
all_downstream_oot = [self._repo.revparse_single(c)
for c in commit_shas]
# Now filter out reverted patches and mergeups from the
# complete list of OOT patches.
downstream_out = OrderedDict()
for c in all_downstream_oot:
sha, sl = str(c.oid), commit_shortlog(c)
is_revert = shortlog_is_revert(sl) # this is just a heuristic
if len(c.parents) > 1:
if not self._include_mergeups:
# Skip all the mergeup commits.
log.dbg('** skipped mergeup {} ("{}")'.format(sha, sl),
level=log.VERBOSE_VERY)
continue
else:
is_revert = False # a merge is never a revert
if is_revert:
# If a shortlog marks a revert, delete the original commit
# from downstream_out, if it can be found.
try:
rsha = commit_reverts_what(c)
except ValueError:
# Badly formatted revert message.
# Treat as outstanding, but complain.
log.wrn(
'revert {} doesn\'t say "reverts commit <SHA>":\n{}'.
format(str(sha), textwrap.indent(c.message, '\t')))
rsha = None
if rsha in downstream_out:
log.dbg('** commit {} ("{}") was reverted in {}'.
format(rsha, commit_shortlog(downstream_out[rsha]),
sha), level=log.VERBOSE_VERY)
del downstream_out[rsha]
continue
elif rsha is not None:
# Make sure the reverted commit is in downstream history.
# (It might not be in all_downstream_oot if e.g.
# downstream reverts an upstream patch as a hotfix, and we
# shouldn't warn about that.)
is_ancestor = self._dp.git(
'merge-base --is-ancestor {} {}'.format(rsha, dsha),
capture_stdout=True).returncode == 0
if not is_ancestor:
log.wrn(('commit {} ("{}") reverts {}, '
"which isn't in downstream history").
format(sha, sl, rsha))
# Emit a warning if we have a non-revert patch with an
# incorrect sauce tag. (Again, downstream might carry reverts
# of upstream patches as hotfixes, which we shouldn't
# warn about.)
if (not shortlog_has_sauce(sl, self._downstream_sauce) and
not is_revert):
log.wrn(f'{self._dp.name}: bad or missing sauce tag: {sha} ("{sl}")')
downstream_out[sha] = c
log.dbg('** added oot patch: {} ("{}")'.format(sha, sl),
level=log.VERBOSE_VERY)
return list(downstream_out.values())
def analyze(self):
'''Analyze repository history.
If this returns without raising an exception, the return value
is a ZephyrRepoAnalysis.
'''
try:
self.repo = pygit2.Repository(self.repo_path)
except KeyError:
# pygit2 raises KeyError when the current path is not a Git
# repository.
msg = "Can't initialize Git repository at {}"
raise InvalidRepositoryError(msg.format(self.repo_path))
#
# Group all upstream commits by area, and collect patch counts.
#
upstream_new = self._new_upstream_only_commits()
upstream_commit_range = (upstream_new[0], upstream_new[-1])
upstream_area_patches = defaultdict(list)
for c in upstream_new:
area = self._check_known_area(c) or commit_area(c)
upstream_area_patches[area].append(c)
unknown_area = upstream_area_patches.get(None)
if unknown_area:
raise UnknownCommitsError(*unknown_area)
upstream_area_counts = {}
for area, patches in upstream_area_patches.items():
upstream_area_counts[area] = len(patches)
#
# Analyze downstream portion of the tree.
#
downstream_only = self._all_downstream_only_commits()
downstream_outstanding = OrderedDict()
for c in downstream_only:
if len(c.parents) > 1:
# Skip all the mergeup commits.
continue
sl = commit_shortlog(c)
if shortlog_is_revert(sl):
# If a shortlog marks a revert, delete the original commit
# from outstanding.
what = shortlog_reverts_what(sl)
if what not in downstream_outstanding:
logging.warning(
"%s was reverted, but isn't in downstream history",
what)
continue
del downstream_outstanding[what]
else:
# Non-revert commits just get appended onto
# downstream_outstanding, keyed by shortlog to make finding
# them later in case they're reverted easier.
#
# We could try to support this by looking into the entire
# revert message to find the "This reverts commit SHA"
# text and computing reverts based on oid rather than
# shortlog. That'd be more robust, but let's not worry
# about it for now.
if sl in downstream_outstanding:
msg = 'duplicated commit shortlogs ({})'.format(sl)
raise NotImplementedError(msg)
# Emit a warning if we have a non-revert patch with an
# incorrect sauce tag. (Downstream might carry reverts
# of upstream patches as hotfixes, which we shouldn't
# warn about.)
if not shortlog_has_sauce(sl, self.downstream_sauce):
logging.warning('out of tree patch has bad sauce: %s %s',
c.oid, sl)
downstream_outstanding[sl] = c
# Compute likely merged patches.
upstream_downstream = [
c for c in upstream_new
if c.author.email.endswith(self.downstream_domain)
]
likely_merged = OrderedDict()
for downstream_sl, downstream_c in downstream_outstanding.items():
def ed(upstream_commit):
return editdistance.eval(
shortlog_no_sauce(downstream_sl, self.downstream_sauce),
commit_shortlog(upstream_commit))
matches = [
c for c in upstream_downstream
if ed(c) < self.edit_dist_threshold
]
if len(matches) != 0:
likely_merged[downstream_sl] = matches
return ZephyrRepoAnalysis(upstream_area_counts, upstream_area_patches,
upstream_commit_range,
downstream_outstanding, likely_merged)
def analyze(self):
'''Analyze repository history.
If this returns without raising an exception, the return value
is a ZephyrRepoAnalysis.
'''
try:
self.repo = pygit2.Repository(self.repo_path)
except KeyError:
# pygit2 raises KeyError when the current path is not a Git
# repository.
msg = "Can't initialize Git repository at {}"
raise InvalidRepositoryError(msg.format(self.repo_path))
#
# Group all upstream commits by area, and collect patch counts.
#
upstream_new = self._new_upstream_only_commits()
upstream_commit_range = (upstream_new[0], upstream_new[-1])
upstream_area_patches = defaultdict(list)
for c in upstream_new:
area = self._check_known_area(c) or commit_area(c)
upstream_area_patches[area].append(c)
unknown_area = upstream_area_patches.get(None)
if unknown_area:
raise UnknownCommitsError(*unknown_area)
upstream_area_counts = {}
for area, patches in upstream_area_patches.items():
upstream_area_counts[area] = len(patches)
#
# Analyze FIO portion of the tree.
#
fio_only = self._all_fio_only_commits()
fio_outstanding = OrderedDict()
for c in fio_only:
if len(c.parents) > 1:
# Skip all the mergeup commits.
continue
sl = commit_shortlog(c)
if shortlog_is_revert(sl):
# If a shortlog marks a revert, delete the original commit
# from outstanding.
what = shortlog_reverts_what(sl)
if what not in fio_outstanding:
print('WARNING: {} was reverted,'.format(what),
"but isn't present in FIO history",
file=sys.stderr)
continue
del fio_outstanding[what]
else:
# Non-revert commits just get appended onto
# fio_outstanding, keyed by shortlog to make finding
# them later in case they're reverted easier.
#
# We could try to support this by looking into the entire
# revert message to find the "This reverts commit SHA"
# text and computing reverts based on oid rather than
# shortlog. That'd be more robust, but let's not worry
# about it for now.
if sl in fio_outstanding:
msg = 'duplicated commit shortlogs ({})'.format(sl)
raise NotImplementedError(msg)
fio_outstanding[sl] = c
# Compute likely merged patches.
upstream_fio = [c for c in upstream_new if commit_is_fio(c)]
likely_merged = OrderedDict()
for fio_sl, fio_c in fio_outstanding.items():
def ed(upstream_commit):
return editdistance.eval(shortlog_no_sauce(fio_sl),
commit_shortlog(upstream_commit))
matches = [c for c in upstream_fio if
ed(c) < self.edit_dist_threshold]
if len(matches) != 0:
likely_merged[fio_sl] = matches
return ZephyrRepoAnalysis(upstream_area_counts,
upstream_area_patches,
upstream_commit_range,
fio_outstanding,
likely_merged)
