BIND 10 #267: python log shouldn't die with OSError

[BIND 10 Development]

#267: python log shouldn't die with OSError
--------------------------+-------------------------------------------------
      Reporter:  jinmei   |        Owner:  stephen                    
          Type:  defect   |       Status:  reviewing                  
      Priority:  major    |    Milestone:  06. 4th Incremental Release
     Component:  logging  |   Resolution:                             
      Keywords:           |    Sensitive:  0                          
Estimatedhours:  0.0      |        Hours:                             
      Billable:  1        |   Totalhours:                             
      Internal:  0        |  
--------------------------+-------------------------------------------------
Changes (by zzchen_pku):

  * owner:  zzchen_pku => stephen


Comment:

 Replying to [comment:16 stephen]:

 > '''src/lib/python/isc/log/log.py'''
 > ''!__init!__'' and ''shouldRollover()''
 > Are we absolutely sure that "filename" contains a directory component at
 this point?  If given as just a bare filename (e.g. "log.txt"), then
 os.path.split() will return the empty string as its first element and the
 subsequent call to os.makedirs() will fail.
 Added NSFileLogHandler._get_abs_file_path() to get absolute file path.
 > Remark: "dir" (used to hold the result of os.path.split()) is the name
 of a built-in Python function. Its use as a variable is OK, but it is
 highlighted when I look at the file with an editor that does syntax
 colouring.
 Done, should avoid use build-in python fuction as variable name.
 > I don't think that the comments have quite the level of detail that
 Jinmei wanted.  I think the essence of the question is why create the
 directories?  Most other applications will fail if the directory does not
 exist.
 > Remark: I think that the location of the directory creation is OK - when
 the handler is initialized and whenever there is a possibility that the
 logging system will write to a different file.  Changes made via a call to
 update_config() will be picked up in the next call to shouldRollover().
 (Incidentally, a comment to that effect in update_config() would be
 useful.)
 Updated.
 > '''src/lib/python/isc/log/tests/log_test.py'''
 > ''General''
 Done.
 > ''Class !TestSysLogHandler''
 > test_emit(): This only checks that the emit() call does not fail - it
 does not check that the record has been sent to the syslog stream.
 The destination of syslog stream depends on OS type and syslog.conf, so it
 is difficult to trace and check the log message. Do you have any good
 suggestions?
 Thank you for your comments.

-- 
Ticket URL: <http://bind10.isc.org/ticket/267#comment:17>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development