tracing: ignore any IOErrors when writing to pipe

When the pager forks off the main process, we can end up with the pipe
closed prematurely. Rather than break hg entirely when that happens
and tracing is active, just let lingering events disappear as needed.

Differential Revision: https://phab.mercurial-scm.org/D4445

hgdemandimport/tracing.py

@@ -1,34 +1,44 b''
 # Support code for event tracing in Mercurial. Lives in demandimport
 # so it can also be used in demandimport.
 #
 # Copyright 2018 Google LLC.
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 from __future__ import absolute_import
 
 import contextlib
 import os
 
 _pipe = None
 _checked = False
 
 @contextlib.contextmanager
 def log(whencefmt, *whenceargs):
     global _pipe, _session, _checked
     if _pipe is None:
         if _checked:
             yield
             return
         _checked = True
         if 'HGCATAPULTSERVERPIPE' not in os.environ:
             yield
             return
         _pipe = open(os.environ['HGCATAPULTSERVERPIPE'], 'w', 1)
         _session = os.environ.get('HGCATAPULTSESSION', 'none')
     whence = whencefmt % whenceargs
     try:
+        # Both writes to the pipe are wrapped in try/except to ignore
+        # errors, as we can see mysterious errors in here if the pager
+        # is active. Presumably other conditions could trigger
+        # problems too.
+        try:
         _pipe.write('START %s %s\n' % (_session, whence))
+        except IOError:
+            pass
         yield
     finally:
+        try:
         _pipe.write('END %s %s\n' % (_session, whence))
+        except IOError:
+            pass