To clarify this problem a little more, the Apache source code tends to use
MAX_STRING_LEN whenever it needs to define a string of unknown lengh. After looking
at several locations where MAX_STRING_LEN is used, the actual length of the string may
be unknown, but will certainly not need 8k of stack space. What makes this a problem
is when multiple character strings are defined within a single function all with a
size of MAX_STRING_LEN and then that function is called recursively. On OS's like
NetWare where the stack size is fixed length, this will blow the stack rather quickly.
One instance of this that we found is in the function send_parsed_content() where
three variables are being declared with a size of MAX_STRING_LEN amounting to 24k.
When a server-side include operation is performed, this function is called recusively
through the use of a sub-request thus chewing up 48k+. Is it really necessary for
these variables to be 8k each and are there other places in the code where apache is
defining variables of 8k size when something significantly less will do? This same
problem seems to also exist in the Apache 2.0 code. Are there other OS's that must
depend on a fixed stack size or is this just a NetWare problem?
Brad
>>> [EMAIL PROTECTED] Friday, July 13, 2001 2:18:32 PM >>>
Question regarding MAX_STRING_LEN
What is the reasoning behind the defining the max string len to 8k?
/* The default string lengths */
#define MAX_STRING_LEN HUGE_STRING_LEN
#define HUGE_STRING_LEN 8192
And do we need to use it for the following:
char error_str[MAX_STRING_LEN];
char time_str[MAX_STRING_LEN];
NetWare does not automatically grow the stack size when needed.
Thanks Rich
