Description:
In the past there have been a number of bugs discovered where an error message returned by the server was corrupted:
* http://bugs.mysql.com/bug.php?id=42685
* http://bugs.mysql.com/bug.php?id=47412
RQG already automatically checks for this and will terminate a trial when corruption is detected (characters outside printable ASCII range). This is great, however there are a some messages in the MySQL error messages file which could trigger this detection incorrectly. It is likely the result of worldwide programmars adding error message (even if English) outside the standard ASCII range.
The main benefit of fixing this woud be that this would no loger terminate RQG runs incorrectly, and furthermore that any tools which relay these messages to a GUI would more correctly display these error messages.
Fixing this up also has a very very small optimization benefit - error messages (return to client and/or written to logs) will be smaller - 1 byte instead of 3 bytes. For often seen messages this makes a very small difference.
How to repeat:
Finding the characters can be done with a good editor (VIM has the "ga" shortcut which shows you the chacter value) or in a more manual fashion:
[roel@localhost share]$ cat errmsg-utf8.txt | grep "[ \t]*eng" | sed 's/[ \t]//g' | sed 's/\(.\)/\n\1/g' | sort -u | egrep -v "[-:';%\"\!\?=/_\#a-zA-Z0-9
\@]" > out.txt
[roel@localhost share]$ vi out.txt <---- remove some leftover valid chars
[roel@localhost share]$ cat out.txt <---- I have added character codes
?= 47196
z = 382
a = 945
? = 916
e = 949
? = 951
? = 942
? = 953
? = 943
? = 954
? = 955
ยต = 956
? = 957
? = 959
? = 928
p = 960
? = 961
s = 963
t = 964
? = 967
? = 974
Suggested fix:
Then just find a good/easy way to replace them (in editor or small script).
© 2024, Oracle Corporation and/or its affiliates