Trac #31036: fixdoctests fails on multiline tests

{{{#!diff
diff --git a/src/sage/geometry/polyhedron/backend_normaliz.py
b/src/sage/geometry/polyhedron/backend_normaliz.py
index f47d07f..eb01e1740c 100644
--- a/src/sage/geometry/polyhedron/backend_normaliz.py
+++ b/src/sage/geometry/polyhedron/backend_normaliz.py
@@ -103,7 +103,8 @@ class Polyhedron_normaliz(Polyhedron_base):
     EXAMPLES::

         sage: p = Polyhedron(vertices=[(0,0),(1,0),(0,1)],
rays=[(1,1)],   # optional - pynormaliz
-        ....:                lines=[], backend='normaliz')
+        A 2-dimensional polyhedron in QQ^2 defined as the convex hull
of 3 vertices and 1 ray
+        ....:                lines=[], backend='normaliz'); p
         sage: TestSuite(p).run()
# optional - pynormaliz
}}}

I just added the printing of `p` and run `sage --fixdoctests`.
`fixdoctests` assumed that all tests have only one line so far.

In addition we now fix error testing according to our conventions:

Before running `fixdoctests`

{{{#!diff
     EXAMPLES::

+        sage: raise ValueError
         sage: rt2 = AA(sqrt(2)); rt2
         1.414213562373095?
         sage: rt3 = AA(sqrt(3)); rt3
}}}

After running `fixdoctests`:

{{{#!diff
     EXAMPLES::

+        sage: raise ValueError
+        Traceback (most recent call last):
+        ...
+        ValueError
         sage: rt2 = AA(sqrt(2)); rt2
         1.414213562373095?
         sage: rt3 = AA(sqrt(3)); rt3
}}}

Before running `fixdoctests`:

{{{
+        sage: raise TypeError("the error has changed")
+        Traceback (most recent call last):
+        ...
+        ValueError
}}}

After runninng `fixdoctests`:

{{{#!diff
     EXAMPLES::

+        sage: raise TypeError("the error has changed")
+        Traceback (most recent call last):
+        ...
+        TypeError: the error has changed
         sage: rt2 = AA(sqrt(2)); rt2
         1.414213562373095?
         sage: rt3 = AA(sqrt(3)); rt3
}}}

If a non-empty doctest raises an error after changes, then `fixdoctests`
needs to run twice (probably not very likely usecase).

URL: https://trac.sagemath.org/31036
Reported by: gh-kliem
Ticket author(s): Jonathan Kliem
Reviewer(s): Markus Wageringel

Author: Release Manager

src/bin/sage-fixdoctests

@@ -62,15 +62,31 @@ for block in doctests:
     comma = block.find(', in ')            # we try to extract the line number which gives the test failure
     if line == -1 or comma==-1: continue   # but if something goes wrong we give up
     line_num=eval(block[line+5:comma])
+
+    # Take care of multiline examples.
+    if 'Expected' in block:
+        i1 = block.index('Failed example')
+        i2 = block.index('Expected')
+        example_len = block[i1:i2].count('\n') - 1
+        line_num += example_len - 1
+
     if 'Expected nothing' in block:
         exp = block.find('Expected nothing')
         block='\n'+block[exp+17:]  # so that split('\nGot:\n') does not fail below
     elif 'Expected:' in block:
         exp = block.find('Expected:\n')
         block=block[exp+10:]
+    elif 'Exception raised' in block:
+        exp = block.find('Exception raised')
+        block='\nGot:\n'+block[exp+17:]
     else:
         continue
-    if block[-21:]=='Got:\n    <BLANKLINE>\n':
+    # Error testing.
+    if 'Traceback (most recent call last):' in block:
+        expected, got = block.split('\nGot:\n')
+        got = got.splitlines()
+        got = ['    Traceback (most recent call last):', '    ...', got[-1]]
+    elif block[-21:]=='Got:\n    <BLANKLINE>\n':
         expected=block[:-22]
         got=['']
     else: