contrib: open a hole in the open().read() ban for open().close()
authorAugie Fackler <augie@google.com>
Wed, 14 Mar 2018 15:42:23 -0400
changeset 36949 35d814fe2f30
parent 36948 0585337ea787
child 36950 c18ae7a07019
contrib: open a hole in the open().read() ban for open().close() It turns out open().close() is both fine and something we occasionally do to verify something can be written. The few cases in the codebase were getting missed due to a regular expression bug (which I discussed in my previous change), but since I'm about to fix the bug, I need to fix the patterns. Differential Revision: https://phab.mercurial-scm.org/D2864
contrib/check-code.py
--- a/contrib/check-code.py	Wed Mar 14 15:39:28 2018 -0400
+++ b/contrib/check-code.py	Wed Mar 14 15:42:23 2018 -0400
@@ -319,9 +319,9 @@
      "use util.readfile() instead"),
     (r'[\s\(](open|file)\([^)]*\)\.write\(',
      "use util.writefile() instead"),
-    (r'^[\s\(]*(open(er)?|file)\([^)]*\)',
+    (r'^[\s\(]*(open(er)?|file)\([^)]*\)(?!\.close\(\))',
      "always assign an opened file to a variable, and close it afterwards"),
-    (r'[\s\(](open|file)\([^)]*\)\.',
+    (r'[\s\(](open|file)\([^)]*\)\.(?!close\(\))',
      "always assign an opened file to a variable, and close it afterwards"),
     (r'(?i)descend[e]nt', "the proper spelling is descendAnt"),
     (r'\.debug\(\_', "don't mark debug messages for translation"),