[tools] Update update_test_prefix.py to handle %s after prefixes

[mtrofin]

Summary

Sometimes the check-prefixes is followed by %s, and we want to keep a
white space before it.

[pengfei]

llvm/utils/update_test_prefix.py

21

I checked it and find it didn't work. It seems \w in [] loses its function.
I also think the * after [\s] (you can use \s without []) is problematical for --check-prefixes=A,B. Besides, changing to + is still problematical for it may consume extera blank lines.
It's hard to modify regular expression correctly without lots of checks, I think you can use my previous patch as a test:

git reset --hard c22dc71b120b066c0066b8517014149a001cc2b0^1
git show c22dc71b120b066c0066b8517014149a001cc2b0 --stat | grep "llvm/test" | cut -d " " -f 2 | xargs llvm/utils/update_test_prefix.py

Check if the outputs are identical to the patch.

[mtrofin]

removed space

[mtrofin]

llvm/utils/update_test_prefix.py

21

The problem seems to have been a spurious space. I tried small unittests, like this:

import re

test = ['-check-prefixes=abc123', '-check-prefixes=abc123-def', '-check-prefixes=abc %s', """-check-prefixes=abc
        def""", '-check-prefixes=abc,def', '-check-prefixes=abc,def %s']

for t in test:
  print(t, re.sub('-?-check-prefixes=([\w-]+[\s]*)', '--check-prefix=\\1', t))

$ python test.py
('-check-prefixes=abc123', '--check-prefix=abc123')
('-check-prefixes=abc123-def', '--check-prefix=abc123-def')
('-check-prefixes=abc %s', '--check-prefix=abc %s')
('-check-prefixes=abc\n        def', '--check-prefix=abc\n        def')
('-check-prefixes=abc,def', '--check-prefix=abc,def')
('-check-prefixes=abc,def %s', '--check-prefix=abc,def %s')

[pengfei]

llvm/utils/update_test_prefix.py

21

Ah, that's it. The unittest is awesome. But the last 2 test results are wrong. The expected results are:

('-check-prefixes=abc,def', '--check-prefixes=abc,def')
('-check-prefixes=abc,def %s', '--check-prefixes=abc,def %s')

We can remove the es when there is only one prefix. You can add one more test:

('-check-prefixes=abc\n\n        def', '--check-prefix=abc\n\n        def')

to see using + instead of * is also problematic.

[mtrofin]

llvm/utils/update_test_prefix.py

21

One more try:

print(t, re.sub('-?-check-prefixes=([\w-]+)(\Z|[ \t\n])', '--check-prefix=\\1\\2', t))

('-check-prefixes=abc123', '--check-prefix=abc123')
('-check-prefixes=abc123-def', '--check-prefix=abc123-def')
('-check-prefixes=abc %s', '--check-prefix=abc %s')
('-check-prefixes=abc\n        def', '--check-prefix=abc\n        def')
('-check-prefixes=abc,def', '-check-prefixes=abc,def')
('-check-prefixes=abc,def %s', '-check-prefixes=abc,def %s')

[mtrofin]

llvm/utils/update_test_prefix.py

21

oh, and (for the test you suggested)

('-check-prefixes=abc\n\n def', '--check-prefix=abc\n\n def')

[pengfei]

llvm/utils/update_test_prefix.py

21

It looks good to me. Thanks Mircea!
I know why you are using such expression now. I missed the cases 1, 2 which aren't end up with '\n' or space, there may exist such cases in real tests.

[mtrofin]

updated with last regexp

llvm/utils/update_test_prefix.py

def remove_prefix(i, d=0):
     with open(i, 'r+') as f:
         s = f.read()
         if r:
             p = r.group(1)
             s = re.sub('=' + p + ',', '=', s)
             s = re.sub(',' + p + '([, \n])', '\\1', s)
             s = re.sub('\s+-?-check-prefix=' + p + '([ \n])', '\\1', s)
         else:
-            s = re.sub('-?-check-prefixes=([^, ]+\n)', '--check-prefix=\\1', s)
-            s = re.sub('-?-check-prefixes=([^, ]+) ', '--check-prefix=\\1', s)
+            s = re.sub('-?-check-prefixes=([\w-]+[\s]*) ', '--check-prefix=\\1', s)
             t = re.search('-?-check-(?:prefix|prefixes)=([^ ]+)\s+-?-check-(?:prefix|prefixes)=([^ ]+)', s)
             while t:
                 s = re.sub(t.group(), '--check-prefixes=' + t.group(1) + ',' + t.group(2), s)
                 t = re.search('-?-check-(?:prefix|prefixes)=([^ ]+)\s+-?-check-(?:prefix|prefixes)=([^ ]+)', s)
             s = re.sub('\s+-?-check-prefix=CHECK[ \t]*\n', '\n', s)
         f.truncate(0)
         f.seek(0)
         f.write(s)