'escape_index' is ensured to be less than 'sizeof(escape_buffer)-1'.
This guarantees enough space for one more character in the
escape_buffer. However, if we meet this condition, we go ahead and shove
a character and a null-terminator in to the string, potentially leading
to a buffer overrun.
While looking through syslog messages on OS X, came across plenty of
error/warning messages where the level was in all-caps and as a result
were not recognized. There was also a lot of false positives due to the
word error turning up as part of a longer string.
This change:
- Changes the log level regex to do a case-insensitive search.
- Ognores the string 'error', if it is part of a longer string.
Reverse lookup for pretty print should be a best effort undertaking. So,
+ Give up on EAI_AGAIN.
+ Don't block on the getnameinfo call indefinitely, resolv.conf might be
set up with a large timeout.
+ Disable lookups after the first failure.
+ Lots of OCD error checking.
The regular expression seems to match arbitrary version strings in
pretty-print view. Adding a more restrictive regular expression,
that should be able to distinguish between valid IPv4 addresses and
arbitrary version strings.
- The codes from httplib.py don't contain all the HTTP status codes and
might differ between different versions of Python installed on the
systems the binary is built on.
- Add the complete list from:
http://www.iana.org/assignments/http-status-codes/http-status-codes.xhtml
- Remove ```init_sql.py``` from the source directory.
- Remove the ```init.sql``` targets from Makefile.am