71cda4f111
* remove python-shebang patch since it was integrated upstream[1] [1]: https://github.com/bazelbuild/bazel/pull/11535/files
63 lines
2.0 KiB
Python
Executable File
63 lines
2.0 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
import sys
|
|
import json
|
|
|
|
if len(sys.argv) == 1:
|
|
print("usage: ./this-script WORKSPACE", file=sys.stderr)
|
|
print("Takes the bazel WORKSPACE file and reads all archives into a json dict (by evaling it as python code)", file=sys.stderr)
|
|
print("Hail Eris.", file=sys.stderr)
|
|
sys.exit(1)
|
|
|
|
http_archives = []
|
|
|
|
# just the kw args are the dict { name, sha256, urls … }
|
|
def http_archive(**kw):
|
|
http_archives.append(kw)
|
|
# like http_file
|
|
def http_file(**kw):
|
|
http_archives.append(kw)
|
|
|
|
# this is inverted from http_archive/http_file and bundles multiple archives
|
|
def distdir_tar(**kw):
|
|
for archive_name in kw['archives']:
|
|
http_archives.append({
|
|
"name": archive_name,
|
|
"sha256": kw['sha256'][archive_name],
|
|
"urls": kw['urls'][archive_name]
|
|
})
|
|
|
|
# stubs for symbols we are not interested in
|
|
# might need to be expanded if new bazel releases add symbols to the workspace
|
|
def workspace(name): pass
|
|
def load(*args): pass
|
|
def bind(**kw): pass
|
|
def list_source_repository(**kw): pass
|
|
def new_local_repository(**kw): pass
|
|
def local_repository(**kw): pass
|
|
DOC_VERSIONS = []
|
|
def stardoc_repositories(**kw): pass
|
|
def skydoc_repositories(**kw): pass
|
|
def rules_sass_dependencies(**kw): pass
|
|
def node_repositories(**kw): pass
|
|
def sass_repositories(**kw): pass
|
|
def register_execution_platforms(*args): pass
|
|
def rbe_autoconfig(*args, **kw): pass
|
|
def rules_pkg_dependencies(*args, **kw): pass
|
|
def winsdk_configure(*args, **kw): pass
|
|
def register_local_rc_exe_toolchains(*args, **kw): pass
|
|
def register_toolchains(*args, **kw): pass
|
|
def debian_deps(): pass
|
|
def grpc_deps(): pass
|
|
def grpc_extra_deps(): pass
|
|
def bazel_skylib_workspace(): pass
|
|
|
|
# execute the WORKSPACE like it was python code in this module,
|
|
# using all the function stubs from above.
|
|
with open(sys.argv[1]) as f:
|
|
exec(f.read())
|
|
|
|
# transform to a dict with the names as keys
|
|
d = { el['name']: el for el in http_archives }
|
|
|
|
print(json.dumps(d, sort_keys=True, indent=4))
|