CI Build & Test: Show doctest failures, warnings as annotations in the 'Files changed' tab

.github/workflows/build.yml

@@ -197,15 +197,15 @@ jobs:
       - name: Test all files (sage -t --all --long)
         if: (success() || failure()) && steps.build.outcome == 'success'
         run: |
-          ../sage -python -m pip install coverage
-          ../sage -python -m coverage run ./bin/sage-runtests --all --long -p2 --random-seed=286735480429121101562228604801325644303
-        working-directory: ./worktree-image/src
+          ./sage -python -m pip install coverage
+          ./sage -python -m coverage run --rcfile=src/tox.ini src/bin/sage-runtests --all --long -p2 --format github --random-seed=286735480429121101562228604801325644303
+        working-directory: ./worktree-image
 
       - name: Prepare coverage results
         if: (success() || failure()) && steps.build.outcome == 'success'
         run: |
-          ./venv/bin/python3 -m coverage combine src/.coverage/
-          ./venv/bin/python3 -m coverage xml
+          ./sage -python -m coverage combine --rcfile=src/tox.ini
+          ./sage -python -m coverage xml --rcfile=src/tox.ini
           mkdir -p coverage-report
           mv coverage.xml coverage-report/
         working-directory: ./worktree-image

src/bin/sage-runtests

@@ -34,6 +34,8 @@ if __name__ == "__main__":
     what.add_argument("--installed", action="store_true", default=False, help="test all installed modules of the Sage library")
     parser.add_argument("--logfile", type=argparse.FileType('a'), metavar="FILE", help="log all output to FILE")
 
+    parser.add_argument("--format", choices=["sage", "github"], default="sage",
+                        help="set format of error messages and warnings")
     parser.add_argument("-l", "--long", action="store_true", default=False, help="include lines with the phrase 'long time'")
     parser.add_argument("-s", "--short", dest="target_walltime", nargs='?',
                         type=int, default=-1, const=300, metavar="SECONDS",

src/sage/doctest/control.py

@@ -139,6 +139,7 @@ def __init__(self, **kwds):
         self.show_skipped = False
         self.target_walltime = -1
         self.baseline_stats_path = None
+        self.format = "sage"
 
         # sage-runtests contains more optional tags. Technically, adding
         # auto_optional_tags here is redundant, since that is added

src/sage/doctest/forker.py

@@ -1235,16 +1235,57 @@ def _failure_header(self, test, example, message='Failed example:'):
         """
         out = [self.DIVIDER]
         with OriginalSource(example):
-            if test.filename:
-                if test.lineno is not None and example.lineno is not None:
-                    lineno = test.lineno + example.lineno + 1
+            if self.options.format == 'sage':
+                if test.filename:
+                    if test.lineno is not None and example.lineno is not None:
+                        lineno = test.lineno + example.lineno + 1
+                    else:
+                        lineno = '?'
+                    out.append('File "%s", line %s, in %s' %
+                               (test.filename, lineno, test.name))
+                else:
+                    out.append('Line %s, in %s' % (example.lineno + 1, test.name))
+                out.append(message)
+            elif self.options.format == 'github':
+                # https://docs.github.com/en/actions/using-workflows/workflow-commands-for-github-actions#using-workflow-commands-to-access-toolkit-functions
+                if message.startswith('Warning: '):
+                    command = f'::warning title={message}'
+                    message = message[len('Warning: '):]
+                elif self.baseline.get('failed', False):
+                    command = f'::notice title={message}'
+                    message += ' [failed in baseline]'
                 else:
-                    lineno = '?'
-                out.append('File "%s", line %s, in %s' %
-                           (test.filename, lineno, test.name))
+                    command = f'::error title={message}'
+                if extra := getattr(example, 'extra', None):
+                    message += f': {extra}'
+                if test.filename:
+                    command += f',file={test.filename}'
+                    if test.lineno is not None and example.lineno is not None:
+                        lineno = test.lineno + example.lineno + 1
+                        command += f',line={lineno}'
+                    lineno = None
+                else:
+                    command += f',line={example.lineno + 1}'
+                #
+                # Urlencoding trick for multi-line annotations
+                # https://github.com/actions/starter-workflows/issues/68#issuecomment-581479448
+                #
+                # This only affects the display in the workflow Summary, after clicking "Show more";
+                # the message needs to be long enough so that "Show more" becomes available.
+                # https://github.com/actions/toolkit/issues/193#issuecomment-1867084340
+                #
+                # Unfortunately, this trick does not make the annotations in the diff view multi-line.
+                #
+                if '\n' in message:
+                    message = message.replace('\n', '%0A')
+                    # The actual threshold for "Show more" to appear depends on the window size.
+                    show_more_threshold = 500
+                    if (pad := show_more_threshold - len(message)) > 0:
+                        message += ' ' * pad
+                command += f'::{message}'
+                out.append(command)
             else:
-                out.append('Line %s, in %s' % (example.lineno + 1, test.name))
-            out.append(message)
+                raise ValueError(f'unknown format option: {self.options.format}')
             source = example.source
             out.append(doctest._indent(source))
             return '\n'.join(out)
@@ -1417,6 +1458,7 @@ def report_failure(self, out, test, example, got, globs):
         """
         if not self.options.initial or self.no_failure_yet:
             self.no_failure_yet = False
+            example.extra = f'Got: {got}'
             returnval = doctest.DocTestRunner.report_failure(self, out, test, example, got)
             if self.options.debug:
                 self._fakeout.stop_spoofing()
@@ -1567,6 +1609,8 @@ def report_unexpected_exception(self, out, test, example, exc_info):
         """
         if not self.options.initial or self.no_failure_yet:
             self.no_failure_yet = False
+
+            example.extra = "Exception raised:\n" + "".join(traceback.format_exception(*exc_info))
             returnval = doctest.DocTestRunner.report_unexpected_exception(self, out, test, example, exc_info)
             if self.options.debug:
                 self._fakeout.stop_spoofing()

src/tox.ini

@@ -299,3 +299,7 @@ source = sage
 concurrency = multiprocessing
 data_file = .coverage/.coverage
 disable_warnings = no-data-collected
+
+[coverage:report]
+ignore_errors = True
+skip_empty = True