[clang] [Utils] add --update-tests flag to llvm-lit (PR #97369)

[Henrik G. Olsson via cfe-commits]

https://github.com/hnrklssn updated https://github.com/llvm/llvm-project/pull/97369

>From d93d77e193f235d12d4de4a4b184c458508fa8df Mon Sep 17 00:00:00 2001
From: "Henrik G. Olsson" <h_olsson at apple.com>
Date: Mon, 1 Jul 2024 18:19:09 -0700
Subject: [PATCH 1/4] [Utils] add update-verify-tests.py

Adds a python script to automatically take output from a failed clang
-verify test and update the test case(s) to expect the new behaviour.
---
 clang/utils/update-verify-tests.py | 404 +++++++++++++++++++++++++++++
 1 file changed, 404 insertions(+)
 create mode 100644 clang/utils/update-verify-tests.py

diff --git a/clang/utils/update-verify-tests.py b/clang/utils/update-verify-tests.py
new file mode 100644
index 00000000000000..cfcfefc85e576a
--- /dev/null
+++ b/clang/utils/update-verify-tests.py
@@ -0,0 +1,404 @@
+import sys
+import re
+
+"""
+ Pipe output from clang's -verify into this script to have the test case updated to expect the actual diagnostic output.
+ When inserting new expected-* checks it will place them on the line before the location of the diagnostic, with an @+1,
+ or @+N for some N if there are multiple diagnostics emitted on the same line. If the current checks are using @-N for
+ this line, the new check will follow that convention also.
+ Existing checks will be left untouched as much as possible, including their location and whitespace content, to minimize
+ diffs. If inaccurate their count will be updated, or the check removed entirely.
+
+ Missing features:
+  - custom prefix support (-verify=my-prefix)
+  - multiple prefixes on the same line (-verify=my-prefix,my-other-prefix)
+  - multiple prefixes on separate RUN lines (RUN: -verify=my-prefix\nRUN: -verify my-other-prefix)
+  - regexes with expected-*-re: existing ones will be left untouched if accurate, but the script will abort if there are any
+    diagnostic mismatches on the same line.
+  - multiple checks targeting the same line are supported, but a line may only contain one check
+  - if multiple checks targeting the same line are failing the script is not guaranteed to produce a minimal diff
+
+Example usage:
+  build/bin/llvm-lit clang/test/Sema/ --no-progress-bar -v | python3 update-verify-tests.py
+"""
+
+
+class KnownException(Exception):
+    pass
+
+
+def parse_error_category(s):
+    parts = s.split("diagnostics")
+    diag_category = parts[0]
+    category_parts = parts[0].strip().strip("'").split("-")
+    expected = category_parts[0]
+    if expected != "expected":
+        raise Exception(
+            f"expected 'expected', but found '{expected}'. Custom verify prefixes are not supported."
+        )
+    diag_category = category_parts[1]
+    if "seen but not expected" in parts[1]:
+        seen = True
+    elif "expected but not seen" in parts[1]:
+        seen = False
+    else:
+        raise KnownException(f"unexpected category '{parts[1]}'")
+    return (diag_category, seen)
+
+
+diag_error_re = re.compile(r"File (\S+) Line (\d+): (.+)")
+diag_error_re2 = re.compile(r"File \S+ Line \d+ \(directive at (\S+):(\d+)\): (.+)")
+
+
+def parse_diag_error(s):
+    m = diag_error_re2.match(s)
+    if not m:
+        m = diag_error_re.match(s)
+    if not m:
+        return None
+    return (m.group(1), int(m.group(2)), m.group(3))
+
+
+class Line:
+    def __init__(self, content, line_n):
+        self.content = content
+        self.diag = None
+        self.line_n = line_n
+        self.related_diags = []
+        self.targeting_diags = []
+
+    def update_line_n(self, n):
+        if self.diag and not self.diag.line_is_absolute:
+            self.diag.orig_target_line_n += n - self.line_n
+        self.line_n = n
+        for diag in self.targeting_diags:
+            if diag.line_is_absolute:
+                diag.orig_target_line_n = n
+            else:
+                diag.orig_target_line_n = n - diag.line.line_n
+        for diag in self.related_diags:
+            if not diag.line_is_absolute:
+                pass
+
+    def render(self):
+        if not self.diag:
+            return self.content
+        assert "{{DIAG}}" in self.content
+        res = self.content.replace("{{DIAG}}", self.diag.render())
+        if not res.strip():
+            return ""
+        return res
+
+
+class Diag:
+    def __init__(
+        self,
+        diag_content,
+        category,
+        targeted_line_n,
+        line_is_absolute,
+        count,
+        line,
+        is_re,
+        whitespace_strings,
+    ):
+        self.diag_content = diag_content
+        self.category = category
+        self.orig_target_line_n = targeted_line_n
+        self.line_is_absolute = line_is_absolute
+        self.count = count
+        self.line = line
+        self.target = None
+        self.is_re = is_re
+        self.absolute_target()
+        self.whitespace_strings = whitespace_strings
+
+    def add(self):
+        if targeted_line > 0:
+            targeted_line += 1
+        elif targeted_line < 0:
+            targeted_line -= 1
+
+    def absolute_target(self):
+        if self.line_is_absolute:
+            res = self.orig_target_line_n
+        else:
+            res = self.line.line_n + self.orig_target_line_n
+        if self.target:
+            assert self.line.line_n == res
+        return res
+
+    def relative_target(self):
+        return self.absolute_target() - self.line.line_n
+
+    def render(self):
+        assert self.count >= 0
+        if self.count == 0:
+            return ""
+        line_location_s = ""
+        if self.relative_target() != 0:
+            if self.line_is_absolute:
+                line_location_s = f"@{self.absolute_target()}"
+            elif self.relative_target() > 0:
+                line_location_s = f"@+{self.relative_target()}"
+            else:
+                line_location_s = (
+                    f"@{self.relative_target()}"  # the minus sign is implicit
+                )
+        count_s = "" if self.count == 1 else f"{self.count}"
+        re_s = "-re" if self.is_re else ""
+        if self.whitespace_strings:
+            whitespace1_s = self.whitespace_strings[0]
+            whitespace2_s = self.whitespace_strings[1]
+            whitespace3_s = self.whitespace_strings[2]
+            whitespace4_s = self.whitespace_strings[3]
+        else:
+            whitespace1_s = " "
+            whitespace2_s = ""
+            whitespace3_s = ""
+            whitespace4_s = ""
+        if count_s and not whitespace3_s:
+            whitespace3_s = " "
+        return f"//{whitespace1_s}expected-{self.category}{re_s}{whitespace2_s}{line_location_s}{whitespace3_s}{count_s}{whitespace4_s}{{{{{self.diag_content}}}}}"
+
+
+expected_diag_re = re.compile(
+    r"//(\s*)expected-(note|warning|error)(-re)?(\s*)(@[+-]?\d+)?(\s*)(\d+)?(\s*)\{\{(.*)\}\}"
+)
+
+
+def parse_diag(line, filename, lines):
+    s = line.content
+    ms = expected_diag_re.findall(s)
+    if not ms:
+        return None
+    if len(ms) > 1:
+        print(
+            f"multiple diags on line {filename}:{line.line_n}. Aborting due to missing implementation."
+        )
+        sys.exit(1)
+    [
+        whitespace1_s,
+        category_s,
+        re_s,
+        whitespace2_s,
+        target_line_s,
+        whitespace3_s,
+        count_s,
+        whitespace4_s,
+        diag_s,
+    ] = ms[0]
+    if not target_line_s:
+        target_line_n = 0
+        is_absolute = False
+    elif target_line_s.startswith("@+"):
+        target_line_n = int(target_line_s[2:])
+        is_absolute = False
+    elif target_line_s.startswith("@-"):
+        target_line_n = int(target_line_s[1:])
+        is_absolute = False
+    else:
+        target_line_n = int(target_line_s[1:])
+        is_absolute = True
+    count = int(count_s) if count_s else 1
+    line.content = expected_diag_re.sub("{{DIAG}}", s)
+
+    return Diag(
+        diag_s,
+        category_s,
+        target_line_n,
+        is_absolute,
+        count,
+        line,
+        bool(re_s),
+        [whitespace1_s, whitespace2_s, whitespace3_s, whitespace4_s],
+    )
+
+
+def link_line_diags(lines, diag):
+    line_n = diag.line.line_n
+    target_line_n = diag.absolute_target()
+    step = 1 if target_line_n < line_n else -1
+    for i in range(target_line_n, line_n, step):
+        lines[i - 1].related_diags.append(diag)
+
+
+def add_line(new_line, lines):
+    lines.insert(new_line.line_n - 1, new_line)
+    for i in range(new_line.line_n, len(lines)):
+        line = lines[i]
+        assert line.line_n == i
+        line.update_line_n(i + 1)
+    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
+
+
+indent_re = re.compile(r"\s*")
+
+
+def get_indent(s):
+    return indent_re.match(s).group(0)
+
+
+def add_diag(line_n, diag_s, diag_category, lines):
+    target = lines[line_n - 1]
+    for other in target.targeting_diags:
+        if other.is_re:
+            raise KnownException(
+                "mismatching diag on line with regex matcher. Skipping due to missing implementation"
+            )
+    reverse = (
+        True
+        if [other for other in target.targeting_diags if other.relative_target() < 0]
+        else False
+    )
+
+    targeting = [
+        other for other in target.targeting_diags if not other.line_is_absolute
+    ]
+    targeting.sort(reverse=reverse, key=lambda d: d.relative_target())
+    prev_offset = 0
+    prev_line = target
+    direction = -1 if reverse else 1
+    for d in targeting:
+        if d.relative_target() != prev_offset + direction:
+            break
+        prev_offset = d.relative_target()
+        prev_line = d.line
+    total_offset = prev_offset - 1 if reverse else prev_offset + 1
+    if reverse:
+        new_line_n = prev_line.line_n + 1
+    else:
+        new_line_n = prev_line.line_n
+    assert new_line_n == line_n + (not reverse) - total_offset
+
+    new_line = Line(get_indent(prev_line.content) + "{{DIAG}}\n", new_line_n)
+    new_line.related_diags = list(prev_line.related_diags)
+    add_line(new_line, lines)
+
+    new_diag = Diag(
+        diag_s, diag_category, total_offset, False, 1, new_line, False, None
+    )
+    new_line.diag = new_diag
+    new_diag.target_line = target
+    assert type(new_diag) != str
+    target.targeting_diags.append(new_diag)
+    link_line_diags(lines, new_diag)
+
+
+updated_test_files = set()
+
+
+def update_test_file(filename, diag_errors):
+    print(f"updating test file {filename}")
+    if filename in updated_test_files:
+        print(
+            f"{filename} already updated, but got new output - expect incorrect results"
+        )
+    else:
+        updated_test_files.add(filename)
+    with open(filename, "r") as f:
+        lines = [Line(line, i + 1) for i, line in enumerate(f.readlines())]
+    for line in lines:
+        diag = parse_diag(line, filename, lines)
+        if diag:
+            line.diag = diag
+            diag.target_line = lines[diag.absolute_target() - 1]
+            link_line_diags(lines, diag)
+            lines[diag.absolute_target() - 1].targeting_diags.append(diag)
+
+    for line_n, diag_s, diag_category, seen in diag_errors:
+        if seen:
+            continue
+        # this is a diagnostic expected but not seen
+        assert lines[line_n - 1].diag
+        if diag_s != lines[line_n - 1].diag.diag_content:
+            raise KnownException(
+                f"{filename}:{line_n} - found diag {lines[line_n - 1].diag.diag_content} but expected {diag_s}"
+            )
+        if diag_category != lines[line_n - 1].diag.category:
+            raise KnownException(
+                f"{filename}:{line_n} - found {lines[line_n - 1].diag.category} diag but expected {diag_category}"
+            )
+        lines[line_n - 1].diag.count -= 1
+    diag_errors_left = []
+    diag_errors.sort(reverse=True, key=lambda t: t[0])
+    for line_n, diag_s, diag_category, seen in diag_errors:
+        if not seen:
+            continue
+        target = lines[line_n - 1]
+        other_diags = [
+            d
+            for d in target.targeting_diags
+            if d.diag_content == diag_s and d.category == diag_category
+        ]
+        other_diag = other_diags[0] if other_diags else None
+        if other_diag:
+            other_diag.count += 1
+        else:
+            diag_errors_left.append((line_n, diag_s, diag_category))
+    for line_n, diag_s, diag_category in diag_errors_left:
+        add_diag(line_n, diag_s, diag_category, lines)
+    with open(filename, "w") as f:
+        for line in lines:
+            f.write(line.render())
+
+
+def update_test_files(errors):
+    errors_by_file = {}
+    for (filename, line, diag_s), (diag_category, seen) in errors:
+        if filename not in errors_by_file:
+            errors_by_file[filename] = []
+        errors_by_file[filename].append((line, diag_s, diag_category, seen))
+    for filename, diag_errors in errors_by_file.items():
+        try:
+            update_test_file(filename, diag_errors)
+        except KnownException as e:
+            print(f"{filename} - ERROR: {e}")
+            print("continuing...")
+
+
+curr = []
+curr_category = None
+curr_run_line = None
+lines_since_run = []
+for line in sys.stdin.readlines():
+    lines_since_run.append(line)
+    try:
+        if line.startswith("RUN:"):
+            if curr:
+                update_test_files(curr)
+                curr = []
+                lines_since_run = [line]
+                curr_run_line = line
+            else:
+                for line in lines_since_run:
+                    print(line, end="")
+                    print("====================")
+                print("no mismatching diagnostics found since last RUN line")
+            continue
+        if line.startswith("error: "):
+            if "no expected directives found" in line:
+                print(
+                    f"no expected directives found for RUN line '{curr_run_line.strip()}'. Add 'expected-no-diagnostics' manually if this is intended."
+                )
+                continue
+            curr_category = parse_error_category(line[len("error: ") :])
+            continue
+
+        diag_error = parse_diag_error(line.strip())
+        if diag_error:
+            curr.append((diag_error, curr_category))
+    except Exception as e:
+        for line in lines_since_run:
+            print(line, end="")
+            print("====================")
+            print(e)
+        sys.exit(1)
+if curr:
+    update_test_files(curr)
+    print("done!")
+else:
+    for line in lines_since_run:
+        print(line, end="")
+        print("====================")
+    print("no mismatching diagnostics found")

>From e9919beb705213398ad4faaf4d2b70d6579bdf5a Mon Sep 17 00:00:00 2001
From: "Henrik G. Olsson" <h_olsson at apple.com>
Date: Tue, 27 Aug 2024 16:51:35 -0700
Subject: [PATCH 2/4] [Utils] Add testing for update-verify-tests, fix bugs

Fixes various line offset bugs. Fixes a bug where whitespace could be
emitted between 'expected-[category]' and the line location specifier,
which clang does not parse.

Now handles the cases where the file does not emit any diagnostics, or
previously contained 'expected-no-diagnostics'.
Also does further work to minimise diffs by replacing diagnostics checks
with a new one, even if the previous check was in a location that the
script would not normally emit. This is useful for the case where a
check is on the same line as the emitted diagnostic, and the diagnostic
message changes. Instead of removing the previous diagnostic and
inserting a new one with @+1, the old diagnostic text is replaced with
the new one.
---
 .../Inputs/duplicate-diag.c                   |   8 +
 .../Inputs/duplicate-diag.c.expected          |   8 +
 .../Inputs/infer-indentation.c                |   8 +
 .../Inputs/infer-indentation.c.expected       |  11 +
 .../Inputs/leave-existing-diags.c             |  11 +
 .../Inputs/leave-existing-diags.c.expected    |  12 ++
 .../Inputs/multiple-errors.c                  |   6 +
 .../Inputs/multiple-errors.c.expected         |   9 +
 .../multiple-missing-errors-same-line.c       |   8 +
 ...ltiple-missing-errors-same-line.c.expected |  13 ++
 .../update-verify-tests/Inputs/no-checks.c    |   3 +
 .../Inputs/no-checks.c.expected               |   4 +
 .../update-verify-tests/Inputs/no-diags.c     |   5 +
 .../Inputs/no-diags.c.expected                |   5 +
 .../Inputs/no-expected-diags.c                |   4 +
 .../Inputs/no-expected-diags.c.expected       |   4 +
 .../Inputs/update-same-line.c                 |   4 +
 .../Inputs/update-same-line.c.expected        |   4 +
 .../Inputs/update-single-check.c              |   4 +
 .../Inputs/update-single-check.c.expected     |   4 +
 .../update-verify-tests/duplicate-diag.test   |   4 +
 .../infer-indentation.test                    |   3 +
 .../leave-existing-diags.test                 |   4 +
 .../utils/update-verify-tests/lit.local.cfg   |  25 +++
 .../update-verify-tests/multiple-errors.test  |   3 +
 .../multiple-missing-errors-same-line.test    |   3 +
 .../utils/update-verify-tests/no-checks.test  |   3 +
 .../utils/update-verify-tests/no-diags.test   |   4 +
 .../no-expected-diags.test                    |   4 +
 .../update-verify-tests/update-same-line.test |   4 +
 .../update-single-check.test                  |   3 +
 clang/utils/update-verify-tests.py            | 197 ++++++++++++------
 32 files changed, 325 insertions(+), 67 deletions(-)
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/infer-indentation.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/infer-indentation.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/multiple-errors.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/multiple-errors.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-checks.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-checks.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-diags.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-diags.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/update-same-line.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/update-same-line.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/update-single-check.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/update-single-check.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/duplicate-diag.test
 create mode 100644 clang/test/utils/update-verify-tests/infer-indentation.test
 create mode 100644 clang/test/utils/update-verify-tests/leave-existing-diags.test
 create mode 100644 clang/test/utils/update-verify-tests/lit.local.cfg
 create mode 100644 clang/test/utils/update-verify-tests/multiple-errors.test
 create mode 100644 clang/test/utils/update-verify-tests/multiple-missing-errors-same-line.test
 create mode 100644 clang/test/utils/update-verify-tests/no-checks.test
 create mode 100644 clang/test/utils/update-verify-tests/no-diags.test
 create mode 100644 clang/test/utils/update-verify-tests/no-expected-diags.test
 create mode 100644 clang/test/utils/update-verify-tests/update-same-line.test
 create mode 100644 clang/test/utils/update-verify-tests/update-single-check.test

diff --git a/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c b/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c
new file mode 100644
index 00000000000000..8c7e46c6eca9c1
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c
@@ -0,0 +1,8 @@
+void foo() {
+    // expected-error at +1{{use of undeclared identifier 'a'}}
+    a = 2; a = 2;
+    b = 2; b = 2;
+    // expected-error at +1 3{{use of undeclared identifier 'c'}}
+    c = 2; c = 2;
+    // expected-error 2{{asdf}}
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c.expected b/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c.expected
new file mode 100644
index 00000000000000..6214ff382f4495
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/duplicate-diag.c.expected
@@ -0,0 +1,8 @@
+void foo() {
+    // expected-error at +1 2{{use of undeclared identifier 'a'}}
+    a = 2; a = 2;
+    // expected-error at +1 2{{use of undeclared identifier 'b'}}
+    b = 2; b = 2;
+    // expected-error at +1 2{{use of undeclared identifier 'c'}}
+    c = 2; c = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c b/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c
new file mode 100644
index 00000000000000..0210ac35fd5cd1
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c
@@ -0,0 +1,8 @@
+void foo() {
+         //     expected-error at +1    2      {{use of undeclared identifier 'a'}}
+    a = 2; a = 2; b = 2; b = 2; c = 2;
+         //     expected-error at +1    2      {{asdf}}
+    d = 2;
+    e = 2; f = 2;                 //     expected-error    2      {{use of undeclared identifier 'e'}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c.expected b/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c.expected
new file mode 100644
index 00000000000000..5c5aaeeef97acf
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/infer-indentation.c.expected
@@ -0,0 +1,11 @@
+void foo() {
+         //     expected-error at +3          {{use of undeclared identifier 'c'}}
+         //     expected-error at +2    2      {{use of undeclared identifier 'b'}}
+         //     expected-error at +1    2      {{use of undeclared identifier 'a'}}
+    a = 2; a = 2; b = 2; b = 2; c = 2;
+         //     expected-error at +1          {{use of undeclared identifier 'd'}}
+    d = 2;
+    //     expected-error at +1          {{use of undeclared identifier 'f'}}
+    e = 2; f = 2;                 //     expected-error          {{use of undeclared identifier 'e'}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c b/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c
new file mode 100644
index 00000000000000..1aa8d088e97273
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c
@@ -0,0 +1,11 @@
+void foo() {
+    a = 2;
+    // expected-error at -1{{use of undeclared identifier 'a'}}
+    b = 2;// expected-error{{use of undeclared identifier 'b'}}
+    c = 2;
+    // expected-error at 5{{use of undeclared identifier 'c'}}
+    d = 2; // expected-error-re{{use of {{.*}} identifier 'd'}}
+
+    e = 2; // error to trigger mismatch
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c.expected b/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c.expected
new file mode 100644
index 00000000000000..6b621061bbfbbd
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/leave-existing-diags.c.expected
@@ -0,0 +1,12 @@
+void foo() {
+    a = 2;
+    // expected-error at -1{{use of undeclared identifier 'a'}}
+    b = 2;// expected-error{{use of undeclared identifier 'b'}}
+    c = 2;
+    // expected-error at 5{{use of undeclared identifier 'c'}}
+    d = 2; // expected-error-re{{use of {{.*}} identifier 'd'}}
+
+    // expected-error at +1{{use of undeclared identifier 'e'}}
+    e = 2; // error to trigger mismatch
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c b/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c
new file mode 100644
index 00000000000000..e230e0a337bf49
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c
@@ -0,0 +1,6 @@
+void foo() {
+    a = 2;
+    b = 2;
+
+    c = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c.expected b/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c.expected
new file mode 100644
index 00000000000000..27dc1f30a26faf
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/multiple-errors.c.expected
@@ -0,0 +1,9 @@
+void foo() {
+    // expected-error at +1{{use of undeclared identifier 'a'}}
+    a = 2;
+    // expected-error at +1{{use of undeclared identifier 'b'}}
+    b = 2;
+
+    // expected-error at +1{{use of undeclared identifier 'c'}}
+    c = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c b/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c
new file mode 100644
index 00000000000000..03f723d44bbe82
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c
@@ -0,0 +1,8 @@
+void foo() {
+    a = 2; b = 2; c = 2;
+}
+
+void bar() {
+    x = 2; y = 2; z = 2;
+    // expected-error at -1{{use of undeclared identifier 'x'}}
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c.expected b/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c.expected
new file mode 100644
index 00000000000000..24b57f4353d95d
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/multiple-missing-errors-same-line.c.expected
@@ -0,0 +1,13 @@
+void foo() {
+    // expected-error at +3{{use of undeclared identifier 'c'}}
+    // expected-error at +2{{use of undeclared identifier 'b'}}
+    // expected-error at +1{{use of undeclared identifier 'a'}}
+    a = 2; b = 2; c = 2;
+}
+
+void bar() {
+    x = 2; y = 2; z = 2;
+    // expected-error at -1{{use of undeclared identifier 'x'}}
+    // expected-error at -2{{use of undeclared identifier 'y'}}
+    // expected-error at -3{{use of undeclared identifier 'z'}}
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-checks.c b/clang/test/utils/update-verify-tests/Inputs/no-checks.c
new file mode 100644
index 00000000000000..8fd1f7cd333705
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-checks.c
@@ -0,0 +1,3 @@
+void foo() {
+    bar = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-checks.c.expected b/clang/test/utils/update-verify-tests/Inputs/no-checks.c.expected
new file mode 100644
index 00000000000000..e80548fbe50f2c
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-checks.c.expected
@@ -0,0 +1,4 @@
+void foo() {
+    // expected-error at +1{{use of undeclared identifier 'bar'}}
+    bar = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-diags.c b/clang/test/utils/update-verify-tests/Inputs/no-diags.c
new file mode 100644
index 00000000000000..66d169be439402
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-diags.c
@@ -0,0 +1,5 @@
+void foo() {
+    // expected-error at +1{{asdf}}
+    int a = 2;
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-diags.c.expected b/clang/test/utils/update-verify-tests/Inputs/no-diags.c.expected
new file mode 100644
index 00000000000000..05230284945702
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-diags.c.expected
@@ -0,0 +1,5 @@
+// expected-no-diagnostics
+void foo() {
+    int a = 2;
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c b/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c
new file mode 100644
index 00000000000000..78b72e1357da76
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c
@@ -0,0 +1,4 @@
+// expected-no-diagnostics
+void foo() {
+    a = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c.expected b/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c.expected
new file mode 100644
index 00000000000000..d948ffce56189a
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/no-expected-diags.c.expected
@@ -0,0 +1,4 @@
+void foo() {
+    // expected-error at +1{{use of undeclared identifier 'a'}}
+    a = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/update-same-line.c b/clang/test/utils/update-verify-tests/Inputs/update-same-line.c
new file mode 100644
index 00000000000000..5278ce0c57c319
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/update-same-line.c
@@ -0,0 +1,4 @@
+void foo() {
+    bar = 2;     //   expected-error       {{asdf}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/update-same-line.c.expected b/clang/test/utils/update-verify-tests/Inputs/update-same-line.c.expected
new file mode 100644
index 00000000000000..8ba47f788319b1
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/update-same-line.c.expected
@@ -0,0 +1,4 @@
+void foo() {
+    bar = 2;     //   expected-error       {{use of undeclared identifier 'bar'}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/update-single-check.c b/clang/test/utils/update-verify-tests/Inputs/update-single-check.c
new file mode 100644
index 00000000000000..20b011bfc3d77e
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/update-single-check.c
@@ -0,0 +1,4 @@
+void foo() {
+    // expected-error at +1{{asdf}}
+    bar = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/Inputs/update-single-check.c.expected b/clang/test/utils/update-verify-tests/Inputs/update-single-check.c.expected
new file mode 100644
index 00000000000000..e80548fbe50f2c
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/update-single-check.c.expected
@@ -0,0 +1,4 @@
+void foo() {
+    // expected-error at +1{{use of undeclared identifier 'bar'}}
+    bar = 2;
+}
diff --git a/clang/test/utils/update-verify-tests/duplicate-diag.test b/clang/test/utils/update-verify-tests/duplicate-diag.test
new file mode 100644
index 00000000000000..3163ce46199c3f
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/duplicate-diag.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/duplicate-diag.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/duplicate-diag.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
+
diff --git a/clang/test/utils/update-verify-tests/infer-indentation.test b/clang/test/utils/update-verify-tests/infer-indentation.test
new file mode 100644
index 00000000000000..6ba2f5d9d505bf
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/infer-indentation.test
@@ -0,0 +1,3 @@
+# RUN: cp %S/Inputs/infer-indentation.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/infer-indentation.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
diff --git a/clang/test/utils/update-verify-tests/leave-existing-diags.test b/clang/test/utils/update-verify-tests/leave-existing-diags.test
new file mode 100644
index 00000000000000..cde690ef715a67
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/leave-existing-diags.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/leave-existing-diags.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/leave-existing-diags.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
+
diff --git a/clang/test/utils/update-verify-tests/lit.local.cfg b/clang/test/utils/update-verify-tests/lit.local.cfg
new file mode 100644
index 00000000000000..a0b6afccc25010
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/lit.local.cfg
@@ -0,0 +1,25 @@
+import lit.util
+
+# python 2.7 backwards compatibility
+try:
+    from shlex import quote as shell_quote
+except ImportError:
+    from pipes import quote as shell_quote
+
+if config.standalone_build:
+    # These tests require the update-verify-tests.py script from the clang
+    # source tree, so skip these tests if we are doing standalone builds.
+    config.unsupported = True
+else:
+    config.suffixes = [".test"]
+
+    script_path = os.path.join(
+        config.clang_src_dir, "utils", "update-verify-tests.py"
+    )
+    python = shell_quote(config.python_executable)
+    config.substitutions.append(
+        (
+            "%update-verify-tests",
+            "%s %s" % (python, shell_quote(script_path)),
+        )
+    )
diff --git a/clang/test/utils/update-verify-tests/multiple-errors.test b/clang/test/utils/update-verify-tests/multiple-errors.test
new file mode 100644
index 00000000000000..1332ef365dc863
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/multiple-errors.test
@@ -0,0 +1,3 @@
+# RUN: cp %S/Inputs/multiple-errors.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/multiple-errors.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
diff --git a/clang/test/utils/update-verify-tests/multiple-missing-errors-same-line.test b/clang/test/utils/update-verify-tests/multiple-missing-errors-same-line.test
new file mode 100644
index 00000000000000..a9c21cd77e192b
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/multiple-missing-errors-same-line.test
@@ -0,0 +1,3 @@
+# RUN: cp %S/Inputs/multiple-missing-errors-same-line.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/multiple-missing-errors-same-line.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
diff --git a/clang/test/utils/update-verify-tests/no-checks.test b/clang/test/utils/update-verify-tests/no-checks.test
new file mode 100644
index 00000000000000..f6ea91fa552be4
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/no-checks.test
@@ -0,0 +1,3 @@
+# RUN: cp %S/Inputs/no-checks.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/no-checks.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
diff --git a/clang/test/utils/update-verify-tests/no-diags.test b/clang/test/utils/update-verify-tests/no-diags.test
new file mode 100644
index 00000000000000..464fe8894253b6
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/no-diags.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/no-diags.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/no-diags.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
+
diff --git a/clang/test/utils/update-verify-tests/no-expected-diags.test b/clang/test/utils/update-verify-tests/no-expected-diags.test
new file mode 100644
index 00000000000000..75235f17a64a29
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/no-expected-diags.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/no-expected-diags.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/no-expected-diags.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
+
diff --git a/clang/test/utils/update-verify-tests/update-same-line.test b/clang/test/utils/update-verify-tests/update-same-line.test
new file mode 100644
index 00000000000000..324768eae5faac
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/update-same-line.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/update-same-line.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/update-same-line.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
+
diff --git a/clang/test/utils/update-verify-tests/update-single-check.test b/clang/test/utils/update-verify-tests/update-single-check.test
new file mode 100644
index 00000000000000..2cb1ae3bcbd3b8
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/update-single-check.test
@@ -0,0 +1,3 @@
+# RUN: cp %S/Inputs/update-single-check.c %t.c && not %clang_cc1 -verify %t.c 2>&1 | %update-verify-tests
+# RUN: diff -u %S/Inputs/update-single-check.c.expected %t.c
+# RUN: %clang_cc1 -verify %t.c
diff --git a/clang/utils/update-verify-tests.py b/clang/utils/update-verify-tests.py
index cfcfefc85e576a..f40ce15e449f22 100644
--- a/clang/utils/update-verify-tests.py
+++ b/clang/utils/update-verify-tests.py
@@ -28,6 +28,8 @@ class KnownException(Exception):
 
 
 def parse_error_category(s):
+    if "no expected directives found" in line:
+        return None
     parts = s.split("diagnostics")
     diag_category = parts[0]
     category_parts = parts[0].strip().strip("'").split("-")
@@ -64,21 +66,10 @@ def __init__(self, content, line_n):
         self.content = content
         self.diag = None
         self.line_n = line_n
-        self.related_diags = []
         self.targeting_diags = []
 
     def update_line_n(self, n):
-        if self.diag and not self.diag.line_is_absolute:
-            self.diag.orig_target_line_n += n - self.line_n
         self.line_n = n
-        for diag in self.targeting_diags:
-            if diag.line_is_absolute:
-                diag.orig_target_line_n = n
-            else:
-                diag.orig_target_line_n = n - diag.line.line_n
-        for diag in self.related_diags:
-            if not diag.line_is_absolute:
-                pass
 
     def render(self):
         if not self.diag:
@@ -95,16 +86,17 @@ def __init__(
         self,
         diag_content,
         category,
-        targeted_line_n,
+        parsed_target_line_n,
         line_is_absolute,
         count,
         line,
         is_re,
         whitespace_strings,
+        is_from_source_file,
     ):
         self.diag_content = diag_content
         self.category = category
-        self.orig_target_line_n = targeted_line_n
+        self.parsed_target_line_n = parsed_target_line_n
         self.line_is_absolute = line_is_absolute
         self.count = count
         self.line = line
@@ -112,25 +104,50 @@ def __init__(
         self.is_re = is_re
         self.absolute_target()
         self.whitespace_strings = whitespace_strings
+        self.is_from_source_file = is_from_source_file
+
+    def decrement_count(self):
+        self.count -= 1
+        assert self.count >= 0
+
+    def increment_count(self):
+        assert self.count >= 0
+        self.count += 1
 
-    def add(self):
-        if targeted_line > 0:
-            targeted_line += 1
-        elif targeted_line < 0:
-            targeted_line -= 1
+    def unset_target(self):
+        assert self.target is not None
+        self.target.targeting_diags.remove(self)
+        self.target = None
+
+    def set_target(self, target):
+        if self.target:
+            self.unset_target()
+        self.target = target
+        self.target.targeting_diags.append(self)
 
     def absolute_target(self):
-        if self.line_is_absolute:
-            res = self.orig_target_line_n
-        else:
-            res = self.line.line_n + self.orig_target_line_n
         if self.target:
-            assert self.line.line_n == res
-        return res
+            return self.target.line_n
+        if self.line_is_absolute:
+            return self.parsed_target_line_n
+        return self.line.line_n + self.parsed_target_line_n
 
     def relative_target(self):
         return self.absolute_target() - self.line.line_n
 
+    def take(self, other_diag):
+        assert self.count == 0
+        assert other_diag.count > 0
+        assert other_diag.target == self.target
+        assert not other_diag.line_is_absolute
+        assert not other_diag.is_re and not self.is_re
+        self.line_is_absolute = False
+        self.diag_content = other_diag.diag_content
+        self.count = other_diag.count
+        self.category = other_diag.category
+        self.count = other_diag.count
+        other_diag.count = 0
+
     def render(self):
         assert self.count >= 0
         if self.count == 0:
@@ -151,19 +168,25 @@ def render(self):
             whitespace1_s = self.whitespace_strings[0]
             whitespace2_s = self.whitespace_strings[1]
             whitespace3_s = self.whitespace_strings[2]
-            whitespace4_s = self.whitespace_strings[3]
         else:
             whitespace1_s = " "
             whitespace2_s = ""
             whitespace3_s = ""
-            whitespace4_s = ""
-        if count_s and not whitespace3_s:
-            whitespace3_s = " "
-        return f"//{whitespace1_s}expected-{self.category}{re_s}{whitespace2_s}{line_location_s}{whitespace3_s}{count_s}{whitespace4_s}{{{{{self.diag_content}}}}}"
-
+        if count_s and not whitespace2_s:
+            whitespace2_s = " " # required to parse correctly
+        elif not count_s and whitespace2_s == " ":
+            """ Don't emit a weird extra space.
+            However if the whitespace is something other than the
+            standard single space, let it be to avoid disrupting manual formatting.
+            The existence of a non-empty whitespace2_s implies this was parsed with
+            a count > 1 and then decremented, otherwise this whitespace would have
+            been parsed as whitespace3_s.
+            """
+            whitespace2_s = ""
+        return f"//{whitespace1_s}expected-{self.category}{re_s}{line_location_s}{whitespace2_s}{count_s}{whitespace3_s}{{{{{self.diag_content}}}}}"
 
 expected_diag_re = re.compile(
-    r"//(\s*)expected-(note|warning|error)(-re)?(\s*)(@[+-]?\d+)?(\s*)(\d+)?(\s*)\{\{(.*)\}\}"
+    r"//(\s*)expected-(note|warning|error)(-re)?(@[+-]?\d+)?(?:(\s*)(\d+))?(\s*)\{\{(.*)\}\}"
 )
 
 
@@ -173,19 +196,17 @@ def parse_diag(line, filename, lines):
     if not ms:
         return None
     if len(ms) > 1:
-        print(
+        raise KnownException(
             f"multiple diags on line {filename}:{line.line_n}. Aborting due to missing implementation."
         )
-        sys.exit(1)
     [
         whitespace1_s,
         category_s,
         re_s,
-        whitespace2_s,
         target_line_s,
-        whitespace3_s,
+        whitespace2_s,
         count_s,
-        whitespace4_s,
+        whitespace3_s,
         diag_s,
     ] = ms[0]
     if not target_line_s:
@@ -211,18 +232,11 @@ def parse_diag(line, filename, lines):
         count,
         line,
         bool(re_s),
-        [whitespace1_s, whitespace2_s, whitespace3_s, whitespace4_s],
+        [whitespace1_s, whitespace2_s, whitespace3_s],
+        True,
     )
 
 
-def link_line_diags(lines, diag):
-    line_n = diag.line.line_n
-    target_line_n = diag.absolute_target()
-    step = 1 if target_line_n < line_n else -1
-    for i in range(target_line_n, line_n, step):
-        lines[i - 1].related_diags.append(diag)
-
-
 def add_line(new_line, lines):
     lines.insert(new_line.line_n - 1, new_line)
     for i in range(new_line.line_n, len(lines)):
@@ -231,6 +245,14 @@ def add_line(new_line, lines):
         line.update_line_n(i + 1)
     assert all(line.line_n == i + 1 for i, line in enumerate(lines))
 
+def remove_line(old_line, lines):
+    lines.remove(old_line)
+    for i in range(old_line.line_n - 1, len(lines)):
+        line = lines[i]
+        assert line.line_n == i + 2
+        line.update_line_n(i + 1)
+    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
+
 
 indent_re = re.compile(r"\s*")
 
@@ -238,8 +260,11 @@ def add_line(new_line, lines):
 def get_indent(s):
     return indent_re.match(s).group(0)
 
+def orig_line_n_to_new_line_n(line_n, orig_lines):
+    return orig_lines[line_n - 1].line_n
 
-def add_diag(line_n, diag_s, diag_category, lines):
+def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines):
+    line_n = orig_line_n_to_new_line_n(orig_line_n, orig_lines)
     target = lines[line_n - 1]
     for other in target.targeting_diags:
         if other.is_re:
@@ -272,18 +297,42 @@ def add_diag(line_n, diag_s, diag_category, lines):
     assert new_line_n == line_n + (not reverse) - total_offset
 
     new_line = Line(get_indent(prev_line.content) + "{{DIAG}}\n", new_line_n)
-    new_line.related_diags = list(prev_line.related_diags)
     add_line(new_line, lines)
 
+    whitespace_strings = prev_line.diag.whitespace_strings if prev_line.diag else None
     new_diag = Diag(
-        diag_s, diag_category, total_offset, False, 1, new_line, False, None
+        diag_s, diag_category, total_offset, False, 1, new_line, False, whitespace_strings, False,
     )
     new_line.diag = new_diag
-    new_diag.target_line = target
-    assert type(new_diag) != str
-    target.targeting_diags.append(new_diag)
-    link_line_diags(lines, new_diag)
+    new_diag.set_target(target)
+
+def remove_dead_diags(lines):
+    for line in lines:
+        if not line.diag or line.diag.count != 0:
+            continue
+        if line.render() == "":
+            remove_line(line, lines)
+        else:
+            assert line.diag.is_from_source_file
+            for other_diag in line.targeting_diags:
+                if other_diag.is_from_source_file or other_diag.count == 0 or other_diag.category != line.diag.category:
+                    continue
+                if other_diag.is_re or line.diag.is_re:
+                    continue
+                line.diag.take(other_diag)
+                remove_line(other_diag.line, lines)
+
+def has_live_diags(lines):
+    for line in lines:
+        if line.diag and line.diag.count > 0:
+            return True
+    return False
 
+def get_expected_no_diags_line_n(lines):
+    for line in lines:
+        if "expected-no-diagnostics" in line.content:
+            return line.line_n
+    return None
 
 updated_test_files = set()
 
@@ -298,13 +347,14 @@ def update_test_file(filename, diag_errors):
         updated_test_files.add(filename)
     with open(filename, "r") as f:
         lines = [Line(line, i + 1) for i, line in enumerate(f.readlines())]
+    orig_lines = list(lines)
+    expected_no_diags_line_n = get_expected_no_diags_line_n(orig_lines)
+
     for line in lines:
         diag = parse_diag(line, filename, lines)
         if diag:
             line.diag = diag
-            diag.target_line = lines[diag.absolute_target() - 1]
-            link_line_diags(lines, diag)
-            lines[diag.absolute_target() - 1].targeting_diags.append(diag)
+            diag.set_target(lines[diag.absolute_target() - 1])
 
     for line_n, diag_s, diag_category, seen in diag_errors:
         if seen:
@@ -319,13 +369,13 @@ def update_test_file(filename, diag_errors):
             raise KnownException(
                 f"{filename}:{line_n} - found {lines[line_n - 1].diag.category} diag but expected {diag_category}"
             )
-        lines[line_n - 1].diag.count -= 1
+        lines[line_n - 1].diag.decrement_count()
     diag_errors_left = []
     diag_errors.sort(reverse=True, key=lambda t: t[0])
     for line_n, diag_s, diag_category, seen in diag_errors:
         if not seen:
             continue
-        target = lines[line_n - 1]
+        target = orig_lines[line_n - 1]
         other_diags = [
             d
             for d in target.targeting_diags
@@ -333,13 +383,17 @@ def update_test_file(filename, diag_errors):
         ]
         other_diag = other_diags[0] if other_diags else None
         if other_diag:
-            other_diag.count += 1
+            other_diag.increment_count()
         else:
-            diag_errors_left.append((line_n, diag_s, diag_category))
-    for line_n, diag_s, diag_category in diag_errors_left:
-        add_diag(line_n, diag_s, diag_category, lines)
+            add_diag(line_n, diag_s, diag_category, lines, orig_lines)
+    remove_dead_diags(lines)
+    has_diags = has_live_diags(lines)
     with open(filename, "w") as f:
+        if not has_diags and expected_no_diags_line_n is None:
+            f.write("// expected-no-diagnostics\n")
         for line in lines:
+            if has_diags and line.line_n == expected_no_diags_line_n:
+                continue
             f.write(line.render())
 
 
@@ -361,6 +415,7 @@ def update_test_files(errors):
 curr_category = None
 curr_run_line = None
 lines_since_run = []
+skip_to_next_file = False
 for line in sys.stdin.readlines():
     lines_since_run.append(line)
     try:
@@ -374,20 +429,28 @@ def update_test_files(errors):
                 for line in lines_since_run:
                     print(line, end="")
                     print("====================")
-                print("no mismatching diagnostics found since last RUN line")
+                if lines_since_run:
+                    print("no mismatching diagnostics found since last RUN line")
+            skip_to_next_file = False
+            continue
+        if skip_to_next_file:
             continue
         if line.startswith("error: "):
-            if "no expected directives found" in line:
-                print(
-                    f"no expected directives found for RUN line '{curr_run_line.strip()}'. Add 'expected-no-diagnostics' manually if this is intended."
-                )
-                continue
             curr_category = parse_error_category(line[len("error: ") :])
             continue
 
         diag_error = parse_diag_error(line.strip())
         if diag_error:
             curr.append((diag_error, curr_category))
+    except KnownException as e:
+            print(f"Error while parsing: {e}")
+            if curr:
+                print("skipping to next file")
+            curr = []
+            curr_category = None
+            curr_run_line = None
+            lines_since_run = []
+            skip_to_next_file = True
     except Exception as e:
         for line in lines_since_run:
             print(line, end="")

>From dd702ce2be48dcf225755d7228b7993b698e888a Mon Sep 17 00:00:00 2001
From: "Henrik G. Olsson" <h_olsson at apple.com>
Date: Tue, 27 Aug 2024 17:42:50 -0700
Subject: [PATCH 3/4] [Utils] Add custom prefix support to update-verify-tests

Custom prefixes can now be provided with --prefix. The default remains
'expected'.
---
 .../Inputs/non-default-prefix.c               |   5 +
 .../Inputs/non-default-prefix.c.expected      |   5 +
 .../non-default-prefix.test                   |   4 +
 clang/utils/update-verify-tests.py            | 153 ++++++++++--------
 4 files changed, 98 insertions(+), 69 deletions(-)
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c
 create mode 100644 clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c.expected
 create mode 100644 clang/test/utils/update-verify-tests/non-default-prefix.test

diff --git a/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c b/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c
new file mode 100644
index 00000000000000..3d63eaf0f1b878
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c
@@ -0,0 +1,5 @@
+void foo() {
+    a = 2; // check-error{{asdf}}
+           // expected-error at -1{ignored}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c.expected b/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c.expected
new file mode 100644
index 00000000000000..a877f86922123d
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/Inputs/non-default-prefix.c.expected
@@ -0,0 +1,5 @@
+void foo() {
+    a = 2; // check-error{{use of undeclared identifier 'a'}}
+           // expected-error at -1{ignored}}
+}
+
diff --git a/clang/test/utils/update-verify-tests/non-default-prefix.test b/clang/test/utils/update-verify-tests/non-default-prefix.test
new file mode 100644
index 00000000000000..e581755a6e6038
--- /dev/null
+++ b/clang/test/utils/update-verify-tests/non-default-prefix.test
@@ -0,0 +1,4 @@
+# RUN: cp %S/Inputs/non-default-prefix.c %t.c && not %clang_cc1 -verify=check %t.c 2>&1 | %update-verify-tests --prefix check
+# RUN: diff -u %S/Inputs/non-default-prefix.c.expected %t.c
+# RUN: %clang_cc1 -verify=check %t.c
+
diff --git a/clang/utils/update-verify-tests.py b/clang/utils/update-verify-tests.py
index f40ce15e449f22..dd83a9418ddf5f 100644
--- a/clang/utils/update-verify-tests.py
+++ b/clang/utils/update-verify-tests.py
@@ -1,5 +1,6 @@
 import sys
 import re
+import argparse
 
 """
  Pipe output from clang's -verify into this script to have the test case updated to expect the actual diagnostic output.
@@ -10,7 +11,6 @@
  diffs. If inaccurate their count will be updated, or the check removed entirely.
 
  Missing features:
-  - custom prefix support (-verify=my-prefix)
   - multiple prefixes on the same line (-verify=my-prefix,my-other-prefix)
   - multiple prefixes on separate RUN lines (RUN: -verify=my-prefix\nRUN: -verify my-other-prefix)
   - regexes with expected-*-re: existing ones will be left untouched if accurate, but the script will abort if there are any
@@ -27,16 +27,16 @@ class KnownException(Exception):
     pass
 
 
-def parse_error_category(s):
-    if "no expected directives found" in line:
+def parse_error_category(s, prefix):
+    if "no expected directives found" in s:
         return None
     parts = s.split("diagnostics")
     diag_category = parts[0]
     category_parts = parts[0].strip().strip("'").split("-")
     expected = category_parts[0]
-    if expected != "expected":
+    if expected != prefix:
         raise Exception(
-            f"expected 'expected', but found '{expected}'. Custom verify prefixes are not supported."
+            f"expected prefix '{prefix}', but found '{expected}'. Multiple verify prefixes are not supported."
         )
     diag_category = category_parts[1]
     if "seen but not expected" in parts[1]:
@@ -84,6 +84,7 @@ def render(self):
 class Diag:
     def __init__(
         self,
+        prefix,
         diag_content,
         category,
         parsed_target_line_n,
@@ -94,6 +95,7 @@ def __init__(
         whitespace_strings,
         is_from_source_file,
     ):
+        self.prefix = prefix
         self.diag_content = diag_content
         self.category = category
         self.parsed_target_line_n = parsed_target_line_n
@@ -183,14 +185,14 @@ def render(self):
             been parsed as whitespace3_s.
             """
             whitespace2_s = ""
-        return f"//{whitespace1_s}expected-{self.category}{re_s}{line_location_s}{whitespace2_s}{count_s}{whitespace3_s}{{{{{self.diag_content}}}}}"
+        return f"//{whitespace1_s}{self.prefix}-{self.category}{re_s}{line_location_s}{whitespace2_s}{count_s}{whitespace3_s}{{{{{self.diag_content}}}}}"
 
 expected_diag_re = re.compile(
-    r"//(\s*)expected-(note|warning|error)(-re)?(@[+-]?\d+)?(?:(\s*)(\d+))?(\s*)\{\{(.*)\}\}"
+    r"//(\s*)([a-zA-Z]+)-(note|warning|error)(-re)?(@[+-]?\d+)?(?:(\s*)(\d+))?(\s*)\{\{(.*)\}\}"
 )
 
 
-def parse_diag(line, filename, lines):
+def parse_diag(line, filename, lines, prefix):
     s = line.content
     ms = expected_diag_re.findall(s)
     if not ms:
@@ -201,6 +203,7 @@ def parse_diag(line, filename, lines):
         )
     [
         whitespace1_s,
+        check_prefix,
         category_s,
         re_s,
         target_line_s,
@@ -209,6 +212,8 @@ def parse_diag(line, filename, lines):
         whitespace3_s,
         diag_s,
     ] = ms[0]
+    if check_prefix != prefix:
+        return None
     if not target_line_s:
         target_line_n = 0
         is_absolute = False
@@ -225,6 +230,7 @@ def parse_diag(line, filename, lines):
     line.content = expected_diag_re.sub("{{DIAG}}", s)
 
     return Diag(
+        prefix,
         diag_s,
         category_s,
         target_line_n,
@@ -263,7 +269,7 @@ def get_indent(s):
 def orig_line_n_to_new_line_n(line_n, orig_lines):
     return orig_lines[line_n - 1].line_n
 
-def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines):
+def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines, prefix):
     line_n = orig_line_n_to_new_line_n(orig_line_n, orig_lines)
     target = lines[line_n - 1]
     for other in target.targeting_diags:
@@ -301,7 +307,7 @@ def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines):
 
     whitespace_strings = prev_line.diag.whitespace_strings if prev_line.diag else None
     new_diag = Diag(
-        diag_s, diag_category, total_offset, False, 1, new_line, False, whitespace_strings, False,
+        prefix, diag_s, diag_category, total_offset, False, 1, new_line, False, whitespace_strings, False,
     )
     new_line.diag = new_diag
     new_diag.set_target(target)
@@ -328,16 +334,16 @@ def has_live_diags(lines):
             return True
     return False
 
-def get_expected_no_diags_line_n(lines):
+def get_expected_no_diags_line_n(lines, prefix):
     for line in lines:
-        if "expected-no-diagnostics" in line.content:
+        if f"{prefix}-no-diagnostics" in line.content:
             return line.line_n
     return None
 
 updated_test_files = set()
 
 
-def update_test_file(filename, diag_errors):
+def update_test_file(filename, diag_errors, prefix):
     print(f"updating test file {filename}")
     if filename in updated_test_files:
         print(
@@ -348,10 +354,10 @@ def update_test_file(filename, diag_errors):
     with open(filename, "r") as f:
         lines = [Line(line, i + 1) for i, line in enumerate(f.readlines())]
     orig_lines = list(lines)
-    expected_no_diags_line_n = get_expected_no_diags_line_n(orig_lines)
+    expected_no_diags_line_n = get_expected_no_diags_line_n(orig_lines, prefix)
 
     for line in lines:
-        diag = parse_diag(line, filename, lines)
+        diag = parse_diag(line, filename, lines, prefix)
         if diag:
             line.diag = diag
             diag.set_target(lines[diag.absolute_target() - 1])
@@ -385,7 +391,7 @@ def update_test_file(filename, diag_errors):
         if other_diag:
             other_diag.increment_count()
         else:
-            add_diag(line_n, diag_s, diag_category, lines, orig_lines)
+            add_diag(line_n, diag_s, diag_category, lines, orig_lines, prefix)
     remove_dead_diags(lines)
     has_diags = has_live_diags(lines)
     with open(filename, "w") as f:
@@ -397,7 +403,7 @@ def update_test_file(filename, diag_errors):
             f.write(line.render())
 
 
-def update_test_files(errors):
+def update_test_files(errors, prefix):
     errors_by_file = {}
     for (filename, line, diag_s), (diag_category, seen) in errors:
         if filename not in errors_by_file:
@@ -405,63 +411,72 @@ def update_test_files(errors):
         errors_by_file[filename].append((line, diag_s, diag_category, seen))
     for filename, diag_errors in errors_by_file.items():
         try:
-            update_test_file(filename, diag_errors)
+            update_test_file(filename, diag_errors, prefix)
         except KnownException as e:
             print(f"{filename} - ERROR: {e}")
             print("continuing...")
 
-
-curr = []
-curr_category = None
-curr_run_line = None
-lines_since_run = []
-skip_to_next_file = False
-for line in sys.stdin.readlines():
-    lines_since_run.append(line)
-    try:
-        if line.startswith("RUN:"):
-            if curr:
-                update_test_files(curr)
+def check_expectations(tool_output, prefix):
+    curr = []
+    curr_category = None
+    curr_run_line = None
+    lines_since_run = []
+    skip_to_next_file = False
+    for line in tool_output:
+        lines_since_run.append(line)
+        try:
+            if line.startswith("RUN:"):
+                if curr:
+                    update_test_files(curr, prefix)
+                    curr = []
+                    lines_since_run = [line]
+                    curr_run_line = line
+                else:
+                    for line in lines_since_run:
+                        print(line, end="")
+                        print("====================")
+                    if lines_since_run:
+                        print("no mismatching diagnostics found since last RUN line")
+                skip_to_next_file = False
+                continue
+            if skip_to_next_file:
+                continue
+            if line.startswith("error: "):
+                curr_category = parse_error_category(line[len("error: ") :], prefix)
+                continue
+    
+            diag_error = parse_diag_error(line.strip())
+            if diag_error:
+                curr.append((diag_error, curr_category))
+        except KnownException as e:
+                print(f"Error while parsing: {e}")
+                if curr:
+                    print("skipping to next file")
                 curr = []
-                lines_since_run = [line]
-                curr_run_line = line
-            else:
-                for line in lines_since_run:
-                    print(line, end="")
-                    print("====================")
-                if lines_since_run:
-                    print("no mismatching diagnostics found since last RUN line")
-            skip_to_next_file = False
-            continue
-        if skip_to_next_file:
-            continue
-        if line.startswith("error: "):
-            curr_category = parse_error_category(line[len("error: ") :])
-            continue
-
-        diag_error = parse_diag_error(line.strip())
-        if diag_error:
-            curr.append((diag_error, curr_category))
-    except KnownException as e:
-            print(f"Error while parsing: {e}")
-            if curr:
-                print("skipping to next file")
-            curr = []
-            curr_category = None
-            curr_run_line = None
-            lines_since_run = []
-            skip_to_next_file = True
-    except Exception as e:
+                curr_category = None
+                curr_run_line = None
+                lines_since_run = []
+                skip_to_next_file = True
+        except Exception as e:
+            for line in lines_since_run:
+                print(line, end="")
+                print("====================")
+                print(e)
+            sys.exit(1)
+    if curr:
+        update_test_files(curr, prefix)
+        print("done!")
+    else:
         for line in lines_since_run:
             print(line, end="")
             print("====================")
-            print(e)
-        sys.exit(1)
-if curr:
-    update_test_files(curr)
-    print("done!")
-else:
-    for line in lines_since_run:
-        print(line, end="")
-        print("====================")
-    print("no mismatching diagnostics found")
+        print("no mismatching diagnostics found")
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--prefix", default="expected", help="The prefix passed to -verify")
+    args = parser.parse_args()
+    check_expectations(sys.stdin.readlines(), args.prefix)
+
+if __name__ == "__main__":
+    main()

>From 7395ef2db8c43f41cdbed70a4fa8c06579a078fe Mon Sep 17 00:00:00 2001
From: "Henrik G. Olsson" <h_olsson at apple.com>
Date: Wed, 11 Sep 2024 17:05:43 -0700
Subject: [PATCH 4/4] [Utils] Separate update-verify-tests script from core

The script no longer attempts to be RUN line aware. Feeding raw llvm-lit
output into it _may_ still work, but given a large enough number of
test cases to update, the risk that some test case uses unsupported
features grows pretty large. This will instead be handled by adding a
lit integration to automatically invoke the core for each failing test
output. The lit integration will be landed as a separate change.
---
 clang/utils/UpdateVerifyTests/core.py | 452 +++++++++++++++++++++++++
 clang/utils/update-verify-tests.py    | 462 +-------------------------
 2 files changed, 461 insertions(+), 453 deletions(-)
 create mode 100644 clang/utils/UpdateVerifyTests/core.py

diff --git a/clang/utils/UpdateVerifyTests/core.py b/clang/utils/UpdateVerifyTests/core.py
new file mode 100644
index 00000000000000..d1350cdbb698b6
--- /dev/null
+++ b/clang/utils/UpdateVerifyTests/core.py
@@ -0,0 +1,452 @@
+import sys
+import re
+
+DEBUG = False
+
+
+def dprint(*args):
+    if DEBUG:
+        print(*args, file=sys.stderr)
+
+
+class KnownException(Exception):
+    pass
+
+
+def parse_error_category(s, prefix):
+    if "no expected directives found" in s:
+        return None
+    parts = s.split("diagnostics")
+    diag_category = parts[0]
+    category_parts = parts[0].strip().strip("'").split("-")
+    expected = category_parts[0]
+    if expected != prefix:
+        raise Exception(
+            f"expected prefix '{prefix}', but found '{expected}'. Multiple verify prefixes are not supported."
+        )
+    diag_category = category_parts[1]
+    if "seen but not expected" in parts[1]:
+        seen = True
+    elif "expected but not seen" in parts[1]:
+        seen = False
+    else:
+        raise KnownException(f"unexpected category '{parts[1]}'")
+    return (diag_category, seen)
+
+
+diag_error_re = re.compile(r"File (\S+) Line (\d+): (.+)")
+diag_error_re2 = re.compile(r"File \S+ Line \d+ \(directive at (\S+):(\d+)\): (.+)")
+
+
+def parse_diag_error(s):
+    m = diag_error_re2.match(s)
+    if not m:
+        m = diag_error_re.match(s)
+    if not m:
+        return None
+    return (m.group(1), int(m.group(2)), m.group(3))
+
+
+class Line:
+    def __init__(self, content, line_n):
+        self.content = content
+        self.diag = None
+        self.line_n = line_n
+        self.targeting_diags = []
+
+    def update_line_n(self, n):
+        self.line_n = n
+
+    def render(self):
+        if not self.diag:
+            return self.content
+        assert "{{DIAG}}" in self.content
+        res = self.content.replace("{{DIAG}}", self.diag.render())
+        if not res.strip():
+            return ""
+        return res
+
+
+class Diag:
+    def __init__(
+        self,
+        prefix,
+        diag_content,
+        category,
+        parsed_target_line_n,
+        line_is_absolute,
+        count,
+        line,
+        is_re,
+        whitespace_strings,
+        is_from_source_file,
+    ):
+        self.prefix = prefix
+        self.diag_content = diag_content
+        self.category = category
+        self.parsed_target_line_n = parsed_target_line_n
+        self.line_is_absolute = line_is_absolute
+        self.count = count
+        self.line = line
+        self.target = None
+        self.is_re = is_re
+        self.absolute_target()
+        self.whitespace_strings = whitespace_strings
+        self.is_from_source_file = is_from_source_file
+
+    def decrement_count(self):
+        self.count -= 1
+        assert self.count >= 0
+
+    def increment_count(self):
+        assert self.count >= 0
+        self.count += 1
+
+    def unset_target(self):
+        assert self.target is not None
+        self.target.targeting_diags.remove(self)
+        self.target = None
+
+    def set_target(self, target):
+        if self.target:
+            self.unset_target()
+        self.target = target
+        self.target.targeting_diags.append(self)
+
+    def absolute_target(self):
+        if self.target:
+            return self.target.line_n
+        if self.line_is_absolute:
+            return self.parsed_target_line_n
+        return self.line.line_n + self.parsed_target_line_n
+
+    def relative_target(self):
+        return self.absolute_target() - self.line.line_n
+
+    def take(self, other_diag):
+        assert self.count == 0
+        assert other_diag.count > 0
+        assert other_diag.target == self.target
+        assert not other_diag.line_is_absolute
+        assert not other_diag.is_re and not self.is_re
+        self.line_is_absolute = False
+        self.diag_content = other_diag.diag_content
+        self.count = other_diag.count
+        self.category = other_diag.category
+        self.count = other_diag.count
+        other_diag.count = 0
+
+    def render(self):
+        assert self.count >= 0
+        if self.count == 0:
+            return ""
+        line_location_s = ""
+        if self.relative_target() != 0:
+            if self.line_is_absolute:
+                line_location_s = f"@{self.absolute_target()}"
+            elif self.relative_target() > 0:
+                line_location_s = f"@+{self.relative_target()}"
+            else:
+                line_location_s = (
+                    f"@{self.relative_target()}"  # the minus sign is implicit
+                )
+        count_s = "" if self.count == 1 else f"{self.count}"
+        re_s = "-re" if self.is_re else ""
+        if self.whitespace_strings:
+            whitespace1_s = self.whitespace_strings[0]
+            whitespace2_s = self.whitespace_strings[1]
+            whitespace3_s = self.whitespace_strings[2]
+        else:
+            whitespace1_s = " "
+            whitespace2_s = ""
+            whitespace3_s = ""
+        if count_s and not whitespace2_s:
+            whitespace2_s = " "  # required to parse correctly
+        elif not count_s and whitespace2_s == " ":
+            """Don't emit a weird extra space.
+            However if the whitespace is something other than the
+            standard single space, let it be to avoid disrupting manual formatting.
+            The existence of a non-empty whitespace2_s implies this was parsed with
+            a count > 1 and then decremented, otherwise this whitespace would have
+            been parsed as whitespace3_s.
+            """
+            whitespace2_s = ""
+        return f"//{whitespace1_s}{self.prefix}-{self.category}{re_s}{line_location_s}{whitespace2_s}{count_s}{whitespace3_s}{{{{{self.diag_content}}}}}"
+
+
+expected_diag_re = re.compile(
+    r"//(\s*)([a-zA-Z]+)-(note|warning|error)(-re)?(@[+-]?\d+)?(?:(\s*)(\d+))?(\s*)\{\{(.*)\}\}"
+)
+
+
+def parse_diag(line, filename, lines, prefix):
+    s = line.content
+    ms = expected_diag_re.findall(s)
+    if not ms:
+        return None
+    if len(ms) > 1:
+        raise KnownException(
+            f"multiple diags on line {filename}:{line.line_n}. Aborting due to missing implementation."
+        )
+    [
+        whitespace1_s,
+        check_prefix,
+        category_s,
+        re_s,
+        target_line_s,
+        whitespace2_s,
+        count_s,
+        whitespace3_s,
+        diag_s,
+    ] = ms[0]
+    if check_prefix != prefix:
+        return None
+    if not target_line_s:
+        target_line_n = 0
+        is_absolute = False
+    elif target_line_s.startswith("@+"):
+        target_line_n = int(target_line_s[2:])
+        is_absolute = False
+    elif target_line_s.startswith("@-"):
+        target_line_n = int(target_line_s[1:])
+        is_absolute = False
+    else:
+        target_line_n = int(target_line_s[1:])
+        is_absolute = True
+    count = int(count_s) if count_s else 1
+    line.content = expected_diag_re.sub("{{DIAG}}", s)
+
+    return Diag(
+        prefix,
+        diag_s,
+        category_s,
+        target_line_n,
+        is_absolute,
+        count,
+        line,
+        bool(re_s),
+        [whitespace1_s, whitespace2_s, whitespace3_s],
+        True,
+    )
+
+
+def add_line(new_line, lines):
+    lines.insert(new_line.line_n - 1, new_line)
+    for i in range(new_line.line_n, len(lines)):
+        line = lines[i]
+        assert line.line_n == i
+        line.update_line_n(i + 1)
+    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
+
+
+def remove_line(old_line, lines):
+    lines.remove(old_line)
+    for i in range(old_line.line_n - 1, len(lines)):
+        line = lines[i]
+        assert line.line_n == i + 2
+        line.update_line_n(i + 1)
+    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
+
+
+indent_re = re.compile(r"\s*")
+
+
+def get_indent(s):
+    return indent_re.match(s).group(0)
+
+
+def orig_line_n_to_new_line_n(line_n, orig_lines):
+    return orig_lines[line_n - 1].line_n
+
+
+def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines, prefix):
+    line_n = orig_line_n_to_new_line_n(orig_line_n, orig_lines)
+    target = lines[line_n - 1]
+    for other in target.targeting_diags:
+        if other.is_re:
+            raise KnownException(
+                "mismatching diag on line with regex matcher. Skipping due to missing implementation"
+            )
+    reverse = (
+        True
+        if [other for other in target.targeting_diags if other.relative_target() < 0]
+        else False
+    )
+
+    targeting = [
+        other for other in target.targeting_diags if not other.line_is_absolute
+    ]
+    targeting.sort(reverse=reverse, key=lambda d: d.relative_target())
+    prev_offset = 0
+    prev_line = target
+    direction = -1 if reverse else 1
+    for d in targeting:
+        if d.relative_target() != prev_offset + direction:
+            break
+        prev_offset = d.relative_target()
+        prev_line = d.line
+    total_offset = prev_offset - 1 if reverse else prev_offset + 1
+    if reverse:
+        new_line_n = prev_line.line_n + 1
+    else:
+        new_line_n = prev_line.line_n
+    assert new_line_n == line_n + (not reverse) - total_offset
+
+    new_line = Line(get_indent(prev_line.content) + "{{DIAG}}\n", new_line_n)
+    add_line(new_line, lines)
+
+    whitespace_strings = prev_line.diag.whitespace_strings if prev_line.diag else None
+    new_diag = Diag(
+        prefix,
+        diag_s,
+        diag_category,
+        total_offset,
+        False,
+        1,
+        new_line,
+        False,
+        whitespace_strings,
+        False,
+    )
+    new_line.diag = new_diag
+    new_diag.set_target(target)
+
+
+def remove_dead_diags(lines):
+    for line in lines:
+        if not line.diag or line.diag.count != 0:
+            continue
+        if line.render() == "":
+            remove_line(line, lines)
+        else:
+            assert line.diag.is_from_source_file
+            for other_diag in line.targeting_diags:
+                if (
+                    other_diag.is_from_source_file
+                    or other_diag.count == 0
+                    or other_diag.category != line.diag.category
+                ):
+                    continue
+                if other_diag.is_re or line.diag.is_re:
+                    continue
+                line.diag.take(other_diag)
+                remove_line(other_diag.line, lines)
+
+
+def has_live_diags(lines):
+    for line in lines:
+        if line.diag and line.diag.count > 0:
+            return True
+    return False
+
+
+def get_expected_no_diags_line_n(lines, prefix):
+    for line in lines:
+        if f"{prefix}-no-diagnostics" in line.content:
+            return line.line_n
+    return None
+
+
+def update_test_file(filename, diag_errors, prefix, updated_test_files):
+    dprint(f"updating test file {filename}")
+    if filename in updated_test_files:
+        raise KnownException(f"{filename} already updated, but got new output")
+    else:
+        updated_test_files.add(filename)
+    with open(filename, "r") as f:
+        lines = [Line(line, i + 1) for i, line in enumerate(f.readlines())]
+    orig_lines = list(lines)
+    expected_no_diags_line_n = get_expected_no_diags_line_n(orig_lines, prefix)
+
+    for line in lines:
+        diag = parse_diag(line, filename, lines, prefix)
+        if diag:
+            line.diag = diag
+            diag.set_target(lines[diag.absolute_target() - 1])
+
+    for line_n, diag_s, diag_category, seen in diag_errors:
+        if seen:
+            continue
+        # this is a diagnostic expected but not seen
+        assert lines[line_n - 1].diag
+        if diag_s != lines[line_n - 1].diag.diag_content:
+            raise KnownException(
+                f"{filename}:{line_n} - found diag {lines[line_n - 1].diag.diag_content} but expected {diag_s}"
+            )
+        if diag_category != lines[line_n - 1].diag.category:
+            raise KnownException(
+                f"{filename}:{line_n} - found {lines[line_n - 1].diag.category} diag but expected {diag_category}"
+            )
+        lines[line_n - 1].diag.decrement_count()
+    diag_errors_left = []
+    diag_errors.sort(reverse=True, key=lambda t: t[0])
+    for line_n, diag_s, diag_category, seen in diag_errors:
+        if not seen:
+            continue
+        target = orig_lines[line_n - 1]
+        other_diags = [
+            d
+            for d in target.targeting_diags
+            if d.diag_content == diag_s and d.category == diag_category
+        ]
+        other_diag = other_diags[0] if other_diags else None
+        if other_diag:
+            other_diag.increment_count()
+        else:
+            add_diag(line_n, diag_s, diag_category, lines, orig_lines, prefix)
+    remove_dead_diags(lines)
+    has_diags = has_live_diags(lines)
+    with open(filename, "w") as f:
+        if not has_diags and expected_no_diags_line_n is None:
+            f.write("// expected-no-diagnostics\n")
+        for line in lines:
+            if has_diags and line.line_n == expected_no_diags_line_n:
+                continue
+            f.write(line.render())
+
+
+def update_test_files(errors, prefix):
+    errors_by_file = {}
+    for (filename, line, diag_s), (diag_category, seen) in errors:
+        if filename not in errors_by_file:
+            errors_by_file[filename] = []
+        errors_by_file[filename].append((line, diag_s, diag_category, seen))
+    updated_test_files = set()
+    for filename, diag_errors in errors_by_file.items():
+        try:
+            update_test_file(filename, diag_errors, prefix, updated_test_files)
+        except KnownException as e:
+            return f"Error in update-verify-tests while updating {filename}: {e}"
+    updated_files = list(updated_test_files)
+    assert updated_files
+    if len(updated_files) == 1:
+        return f"updated file {updated_files[0]}"
+    updated_files_s = "\n\t".join(updated_files)
+    return "updated files:\n\t{updated_files_s}"
+
+
+def check_expectations(tool_output, prefix):
+    """
+    The entry point function.
+    Called by the stand-alone update-verify-tests.py as well as litplugin.py.
+    """
+    curr = []
+    curr_category = None
+    try:
+        for line in tool_output:
+            if line.startswith("error: "):
+                curr_category = parse_error_category(line[len("error: ") :], prefix)
+                continue
+
+            diag_error = parse_diag_error(line.strip())
+            if diag_error:
+                curr.append((diag_error, curr_category))
+            else:
+                dprint("no match")
+                dprint(line.strip())
+    except KnownException as e:
+        return f"Error in update-verify-tests while parsing tool output: {e}"
+    if curr:
+        return update_test_files(curr, prefix)
+    else:
+        return "no mismatching diagnostics found"
diff --git a/clang/utils/update-verify-tests.py b/clang/utils/update-verify-tests.py
index dd83a9418ddf5f..e2874a8c049ef3 100644
--- a/clang/utils/update-verify-tests.py
+++ b/clang/utils/update-verify-tests.py
@@ -1,6 +1,6 @@
 import sys
-import re
 import argparse
+from UpdateVerifyTests.core import check_expectations
 
 """
  Pipe output from clang's -verify into this script to have the test case updated to expect the actual diagnostic output.
@@ -19,464 +19,20 @@
   - if multiple checks targeting the same line are failing the script is not guaranteed to produce a minimal diff
 
 Example usage:
-  build/bin/llvm-lit clang/test/Sema/ --no-progress-bar -v | python3 update-verify-tests.py
+  clang -verify [file] | python3 update-verify-tests.py
+  clang -verify=check [file] | python3 update-verify-tests.py --prefix check
 """
 
 
-class KnownException(Exception):
-    pass
-
-
-def parse_error_category(s, prefix):
-    if "no expected directives found" in s:
-        return None
-    parts = s.split("diagnostics")
-    diag_category = parts[0]
-    category_parts = parts[0].strip().strip("'").split("-")
-    expected = category_parts[0]
-    if expected != prefix:
-        raise Exception(
-            f"expected prefix '{prefix}', but found '{expected}'. Multiple verify prefixes are not supported."
-        )
-    diag_category = category_parts[1]
-    if "seen but not expected" in parts[1]:
-        seen = True
-    elif "expected but not seen" in parts[1]:
-        seen = False
-    else:
-        raise KnownException(f"unexpected category '{parts[1]}'")
-    return (diag_category, seen)
-
-
-diag_error_re = re.compile(r"File (\S+) Line (\d+): (.+)")
-diag_error_re2 = re.compile(r"File \S+ Line \d+ \(directive at (\S+):(\d+)\): (.+)")
-
-
-def parse_diag_error(s):
-    m = diag_error_re2.match(s)
-    if not m:
-        m = diag_error_re.match(s)
-    if not m:
-        return None
-    return (m.group(1), int(m.group(2)), m.group(3))
-
-
-class Line:
-    def __init__(self, content, line_n):
-        self.content = content
-        self.diag = None
-        self.line_n = line_n
-        self.targeting_diags = []
-
-    def update_line_n(self, n):
-        self.line_n = n
-
-    def render(self):
-        if not self.diag:
-            return self.content
-        assert "{{DIAG}}" in self.content
-        res = self.content.replace("{{DIAG}}", self.diag.render())
-        if not res.strip():
-            return ""
-        return res
-
-
-class Diag:
-    def __init__(
-        self,
-        prefix,
-        diag_content,
-        category,
-        parsed_target_line_n,
-        line_is_absolute,
-        count,
-        line,
-        is_re,
-        whitespace_strings,
-        is_from_source_file,
-    ):
-        self.prefix = prefix
-        self.diag_content = diag_content
-        self.category = category
-        self.parsed_target_line_n = parsed_target_line_n
-        self.line_is_absolute = line_is_absolute
-        self.count = count
-        self.line = line
-        self.target = None
-        self.is_re = is_re
-        self.absolute_target()
-        self.whitespace_strings = whitespace_strings
-        self.is_from_source_file = is_from_source_file
-
-    def decrement_count(self):
-        self.count -= 1
-        assert self.count >= 0
-
-    def increment_count(self):
-        assert self.count >= 0
-        self.count += 1
-
-    def unset_target(self):
-        assert self.target is not None
-        self.target.targeting_diags.remove(self)
-        self.target = None
-
-    def set_target(self, target):
-        if self.target:
-            self.unset_target()
-        self.target = target
-        self.target.targeting_diags.append(self)
-
-    def absolute_target(self):
-        if self.target:
-            return self.target.line_n
-        if self.line_is_absolute:
-            return self.parsed_target_line_n
-        return self.line.line_n + self.parsed_target_line_n
-
-    def relative_target(self):
-        return self.absolute_target() - self.line.line_n
-
-    def take(self, other_diag):
-        assert self.count == 0
-        assert other_diag.count > 0
-        assert other_diag.target == self.target
-        assert not other_diag.line_is_absolute
-        assert not other_diag.is_re and not self.is_re
-        self.line_is_absolute = False
-        self.diag_content = other_diag.diag_content
-        self.count = other_diag.count
-        self.category = other_diag.category
-        self.count = other_diag.count
-        other_diag.count = 0
-
-    def render(self):
-        assert self.count >= 0
-        if self.count == 0:
-            return ""
-        line_location_s = ""
-        if self.relative_target() != 0:
-            if self.line_is_absolute:
-                line_location_s = f"@{self.absolute_target()}"
-            elif self.relative_target() > 0:
-                line_location_s = f"@+{self.relative_target()}"
-            else:
-                line_location_s = (
-                    f"@{self.relative_target()}"  # the minus sign is implicit
-                )
-        count_s = "" if self.count == 1 else f"{self.count}"
-        re_s = "-re" if self.is_re else ""
-        if self.whitespace_strings:
-            whitespace1_s = self.whitespace_strings[0]
-            whitespace2_s = self.whitespace_strings[1]
-            whitespace3_s = self.whitespace_strings[2]
-        else:
-            whitespace1_s = " "
-            whitespace2_s = ""
-            whitespace3_s = ""
-        if count_s and not whitespace2_s:
-            whitespace2_s = " " # required to parse correctly
-        elif not count_s and whitespace2_s == " ":
-            """ Don't emit a weird extra space.
-            However if the whitespace is something other than the
-            standard single space, let it be to avoid disrupting manual formatting.
-            The existence of a non-empty whitespace2_s implies this was parsed with
-            a count > 1 and then decremented, otherwise this whitespace would have
-            been parsed as whitespace3_s.
-            """
-            whitespace2_s = ""
-        return f"//{whitespace1_s}{self.prefix}-{self.category}{re_s}{line_location_s}{whitespace2_s}{count_s}{whitespace3_s}{{{{{self.diag_content}}}}}"
-
-expected_diag_re = re.compile(
-    r"//(\s*)([a-zA-Z]+)-(note|warning|error)(-re)?(@[+-]?\d+)?(?:(\s*)(\d+))?(\s*)\{\{(.*)\}\}"
-)
-
-
-def parse_diag(line, filename, lines, prefix):
-    s = line.content
-    ms = expected_diag_re.findall(s)
-    if not ms:
-        return None
-    if len(ms) > 1:
-        raise KnownException(
-            f"multiple diags on line {filename}:{line.line_n}. Aborting due to missing implementation."
-        )
-    [
-        whitespace1_s,
-        check_prefix,
-        category_s,
-        re_s,
-        target_line_s,
-        whitespace2_s,
-        count_s,
-        whitespace3_s,
-        diag_s,
-    ] = ms[0]
-    if check_prefix != prefix:
-        return None
-    if not target_line_s:
-        target_line_n = 0
-        is_absolute = False
-    elif target_line_s.startswith("@+"):
-        target_line_n = int(target_line_s[2:])
-        is_absolute = False
-    elif target_line_s.startswith("@-"):
-        target_line_n = int(target_line_s[1:])
-        is_absolute = False
-    else:
-        target_line_n = int(target_line_s[1:])
-        is_absolute = True
-    count = int(count_s) if count_s else 1
-    line.content = expected_diag_re.sub("{{DIAG}}", s)
-
-    return Diag(
-        prefix,
-        diag_s,
-        category_s,
-        target_line_n,
-        is_absolute,
-        count,
-        line,
-        bool(re_s),
-        [whitespace1_s, whitespace2_s, whitespace3_s],
-        True,
-    )
-
-
-def add_line(new_line, lines):
-    lines.insert(new_line.line_n - 1, new_line)
-    for i in range(new_line.line_n, len(lines)):
-        line = lines[i]
-        assert line.line_n == i
-        line.update_line_n(i + 1)
-    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
-
-def remove_line(old_line, lines):
-    lines.remove(old_line)
-    for i in range(old_line.line_n - 1, len(lines)):
-        line = lines[i]
-        assert line.line_n == i + 2
-        line.update_line_n(i + 1)
-    assert all(line.line_n == i + 1 for i, line in enumerate(lines))
-
-
-indent_re = re.compile(r"\s*")
-
-
-def get_indent(s):
-    return indent_re.match(s).group(0)
-
-def orig_line_n_to_new_line_n(line_n, orig_lines):
-    return orig_lines[line_n - 1].line_n
-
-def add_diag(orig_line_n, diag_s, diag_category, lines, orig_lines, prefix):
-    line_n = orig_line_n_to_new_line_n(orig_line_n, orig_lines)
-    target = lines[line_n - 1]
-    for other in target.targeting_diags:
-        if other.is_re:
-            raise KnownException(
-                "mismatching diag on line with regex matcher. Skipping due to missing implementation"
-            )
-    reverse = (
-        True
-        if [other for other in target.targeting_diags if other.relative_target() < 0]
-        else False
-    )
-
-    targeting = [
-        other for other in target.targeting_diags if not other.line_is_absolute
-    ]
-    targeting.sort(reverse=reverse, key=lambda d: d.relative_target())
-    prev_offset = 0
-    prev_line = target
-    direction = -1 if reverse else 1
-    for d in targeting:
-        if d.relative_target() != prev_offset + direction:
-            break
-        prev_offset = d.relative_target()
-        prev_line = d.line
-    total_offset = prev_offset - 1 if reverse else prev_offset + 1
-    if reverse:
-        new_line_n = prev_line.line_n + 1
-    else:
-        new_line_n = prev_line.line_n
-    assert new_line_n == line_n + (not reverse) - total_offset
-
-    new_line = Line(get_indent(prev_line.content) + "{{DIAG}}\n", new_line_n)
-    add_line(new_line, lines)
-
-    whitespace_strings = prev_line.diag.whitespace_strings if prev_line.diag else None
-    new_diag = Diag(
-        prefix, diag_s, diag_category, total_offset, False, 1, new_line, False, whitespace_strings, False,
-    )
-    new_line.diag = new_diag
-    new_diag.set_target(target)
-
-def remove_dead_diags(lines):
-    for line in lines:
-        if not line.diag or line.diag.count != 0:
-            continue
-        if line.render() == "":
-            remove_line(line, lines)
-        else:
-            assert line.diag.is_from_source_file
-            for other_diag in line.targeting_diags:
-                if other_diag.is_from_source_file or other_diag.count == 0 or other_diag.category != line.diag.category:
-                    continue
-                if other_diag.is_re or line.diag.is_re:
-                    continue
-                line.diag.take(other_diag)
-                remove_line(other_diag.line, lines)
-
-def has_live_diags(lines):
-    for line in lines:
-        if line.diag and line.diag.count > 0:
-            return True
-    return False
-
-def get_expected_no_diags_line_n(lines, prefix):
-    for line in lines:
-        if f"{prefix}-no-diagnostics" in line.content:
-            return line.line_n
-    return None
-
-updated_test_files = set()
-
-
-def update_test_file(filename, diag_errors, prefix):
-    print(f"updating test file {filename}")
-    if filename in updated_test_files:
-        print(
-            f"{filename} already updated, but got new output - expect incorrect results"
-        )
-    else:
-        updated_test_files.add(filename)
-    with open(filename, "r") as f:
-        lines = [Line(line, i + 1) for i, line in enumerate(f.readlines())]
-    orig_lines = list(lines)
-    expected_no_diags_line_n = get_expected_no_diags_line_n(orig_lines, prefix)
-
-    for line in lines:
-        diag = parse_diag(line, filename, lines, prefix)
-        if diag:
-            line.diag = diag
-            diag.set_target(lines[diag.absolute_target() - 1])
-
-    for line_n, diag_s, diag_category, seen in diag_errors:
-        if seen:
-            continue
-        # this is a diagnostic expected but not seen
-        assert lines[line_n - 1].diag
-        if diag_s != lines[line_n - 1].diag.diag_content:
-            raise KnownException(
-                f"{filename}:{line_n} - found diag {lines[line_n - 1].diag.diag_content} but expected {diag_s}"
-            )
-        if diag_category != lines[line_n - 1].diag.category:
-            raise KnownException(
-                f"{filename}:{line_n} - found {lines[line_n - 1].diag.category} diag but expected {diag_category}"
-            )
-        lines[line_n - 1].diag.decrement_count()
-    diag_errors_left = []
-    diag_errors.sort(reverse=True, key=lambda t: t[0])
-    for line_n, diag_s, diag_category, seen in diag_errors:
-        if not seen:
-            continue
-        target = orig_lines[line_n - 1]
-        other_diags = [
-            d
-            for d in target.targeting_diags
-            if d.diag_content == diag_s and d.category == diag_category
-        ]
-        other_diag = other_diags[0] if other_diags else None
-        if other_diag:
-            other_diag.increment_count()
-        else:
-            add_diag(line_n, diag_s, diag_category, lines, orig_lines, prefix)
-    remove_dead_diags(lines)
-    has_diags = has_live_diags(lines)
-    with open(filename, "w") as f:
-        if not has_diags and expected_no_diags_line_n is None:
-            f.write("// expected-no-diagnostics\n")
-        for line in lines:
-            if has_diags and line.line_n == expected_no_diags_line_n:
-                continue
-            f.write(line.render())
-
-
-def update_test_files(errors, prefix):
-    errors_by_file = {}
-    for (filename, line, diag_s), (diag_category, seen) in errors:
-        if filename not in errors_by_file:
-            errors_by_file[filename] = []
-        errors_by_file[filename].append((line, diag_s, diag_category, seen))
-    for filename, diag_errors in errors_by_file.items():
-        try:
-            update_test_file(filename, diag_errors, prefix)
-        except KnownException as e:
-            print(f"{filename} - ERROR: {e}")
-            print("continuing...")
-
-def check_expectations(tool_output, prefix):
-    curr = []
-    curr_category = None
-    curr_run_line = None
-    lines_since_run = []
-    skip_to_next_file = False
-    for line in tool_output:
-        lines_since_run.append(line)
-        try:
-            if line.startswith("RUN:"):
-                if curr:
-                    update_test_files(curr, prefix)
-                    curr = []
-                    lines_since_run = [line]
-                    curr_run_line = line
-                else:
-                    for line in lines_since_run:
-                        print(line, end="")
-                        print("====================")
-                    if lines_since_run:
-                        print("no mismatching diagnostics found since last RUN line")
-                skip_to_next_file = False
-                continue
-            if skip_to_next_file:
-                continue
-            if line.startswith("error: "):
-                curr_category = parse_error_category(line[len("error: ") :], prefix)
-                continue
-    
-            diag_error = parse_diag_error(line.strip())
-            if diag_error:
-                curr.append((diag_error, curr_category))
-        except KnownException as e:
-                print(f"Error while parsing: {e}")
-                if curr:
-                    print("skipping to next file")
-                curr = []
-                curr_category = None
-                curr_run_line = None
-                lines_since_run = []
-                skip_to_next_file = True
-        except Exception as e:
-            for line in lines_since_run:
-                print(line, end="")
-                print("====================")
-                print(e)
-            sys.exit(1)
-    if curr:
-        update_test_files(curr, prefix)
-        print("done!")
-    else:
-        for line in lines_since_run:
-            print(line, end="")
-            print("====================")
-        print("no mismatching diagnostics found")
-
 def main():
     parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument("--prefix", default="expected", help="The prefix passed to -verify")
+    parser.add_argument(
+        "--prefix", default="expected", help="The prefix passed to -verify"
+    )
     args = parser.parse_args()
-    check_expectations(sys.stdin.readlines(), args.prefix)
+    output = check_expectations(sys.stdin.readlines(), args.prefix)
+    print(output)
+
 
 if __name__ == "__main__":
     main()