Fail cc_ast_dump when AST dump encounters errors, add missing depende…

…ncies to API AST dump

PR #2991 introduced a regression where the API AST dump did not have access to
all the required includes. However, an incomplete AST was still being generated
despite the missing include error. Add the missing dependencies so that the
correct types are generated and enable pipefail for the ast dump/gzip shell
commands so that the cc_ast_dump action fails if there are any errors being
reported.

build/cc_ast_dump.bzl

@@ -37,7 +37,9 @@ def _cc_ast_dump_impl(ctx):
     inputs = depset(direct = [ctx.file.src], transitive = [cc_toolchain.all_files] + [cc_info.compilation_context.headers])
     env = cc_common.get_environment_variables(feature_configuration = feature_configuration, action_name = CPP_COMPILE_ACTION_NAME, variables = variables)
 
-    command = " ".join([executable] + arguments + ["|", "gzip", "-3", "-", ">", ctx.outputs.out.path])
+    # Enable pipefail so that the action fails when there are e.g. missing include errors, otherwise
+    # we'd just continue despite the error and end up with an incomplete AST.
+    command = " ".join(["set -euo pipefail;"] + [executable] + arguments + ["|", "gzip", "-3", "-", ">", ctx.outputs.out.path])
 
     # run_shell until https://github.com/bazelbuild/bazel/issues/5511 is fixed
     ctx.actions.run_shell(

src/workerd/tools/BUILD.bazel

@@ -20,6 +20,8 @@ cc_ast_dump(
     deps = [
         "//src/workerd/api:html-rewriter",
         "//src/workerd/api:hyperdrive",
+        "//src/workerd/api:memory-cache",
+        "//src/workerd/api:r2",
         "//src/workerd/api/node",
         "//src/workerd/io",
         "//src/workerd/jsg",