scripts/markdown-badges.py: Generate sorted table for build status

Organize the build badges in a tree/compiler table to make it easier
to examine quickly for common patterns. E.g. "Oh, looks like ToT Clang
broke all the kernels", or "Oh, something is -next broke every Clang
version", etc.

Signed-off-by: Kees Cook <keescook@chromium.org>

scripts/markdown-badges.py

@@ -0,0 +1,81 @@
+#!/usr/bin/env python3
+import os, glob, re
+from pkg_resources import parse_version
+
+# Figure out where we to find the workflow definitions.
+ci_root=os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
+
+# Construct the markdown for a specific workflow badge.
+def svg(workflow):
+	if workflow == None:
+		return "   "
+	workflow_url=f"https://github.com/clangbuiltlinux/continuous-integration2/actions/workflows/{workflow}.yml"
+	return f"[![{workflow} build status]({workflow_url}/badge.svg)]({workflow_url})"
+
+
+print("Copy and paste the output below into README.md:\n")
+
+# Be able to location and extract the name of a workflow.
+name_re = re.compile(r'^name: (.*) \(([^\)]+)\)$')
+# Quick "basename $arg .yml" regular expression.
+base_re = re.compile(r'^.*/([^/]+)\.yml$')
+
+# Find all the tuxsuite workflows.
+trees = dict()
+for yml in glob.glob(f"{ci_root}/.github/workflows/*.yml"):
+	tuxsuite = False
+	tree = None
+	compiler = None
+	for line in open(yml):
+		m = name_re.search(line)
+		if m:
+			tree = m.group(1)
+			compiler = m.group(2)
+			continue
+		if 'tuxsuite' in line:
+			tuxsuite = True
+			break
+	if not tuxsuite:
+		continue
+	# Found a tuxsuite workflow with no "name:" field?!
+	if tree == None or compiler == None:
+		raise ValueError(f"{yml}: missing 'name:'")
+
+	m = base_re.search(yml)
+	base = m.group(1)
+	trees.setdefault(tree, dict())
+	trees[tree][compiler] = base
+
+# Construct the list of all compilers seen by any tree.
+compilers = set()
+for tree in trees:
+	compilers.update(trees[tree].keys())
+# Sort the columns with latest Clang on the left.
+columns = sorted(compilers, key=parse_version, reverse=True)
+
+# To stabilize the size of the the SVGs to make the table as presentable
+# as possible, we must do our best to keep the table columns the same
+# width. To that end, figure out the widest column name.
+# Note that the SVGs themselves vary in size, but we can only do so much.
+max_width = len(max(columns, key=len))
+
+# - Make every column the same width
+# - Force whitespace to non-breaking (to keep the cell from eliding spaces)
+# - Replace "-" with non-breaking-dash (to keep the name from wrapping)
+print("|     | " + " | ".join([" " * (max_width - len(col)) + col.replace('-', '‑') for col in columns]) + " |")
+# Align tree name to the right for readability, and center the badges.
+print("| ---: |" + " :---: |" * len(columns))
+
+# Sort so latest trees are at the top.
+rows = sorted(trees, key=parse_version, reverse=True)
+# Manually override some trees to the top, since they're unversioned.
+rows.remove('next')
+rows.remove('mainline')
+rows.insert(0, 'mainline')
+rows.insert(0, 'next')
+
+# Emit the svg markdown for each tree/compiler combo, if it exists.
+for tree in rows:
+	# Keep names from wrapping.
+	row = tree.replace('-', '‑')
+	print(f"| {row} | " + " | ".join([svg(trees[tree].get(compiler, None)) for compiler in columns]) + " |")