httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: HEUREKA! (was: Update about my core dumps on mainframe)
Date Sat, 13 Dec 1997 19:55:16 GMT
On Sat, 13 Dec 1997, Martin Kraemer wrote:

> On Fri, Dec 12, 1997 at 04:02:31PM -0800, Dean Gaudet wrote:
> > I'll see if I can find anything in mod_mime.  I suspect the problem is
> > actually in the core.
> 
> Dean,
>  Finally, I found the bug! The server is up and running happily with an
> almost complete of modules, and serves fine.

Woo!

>  Thanks a lot for all the help you gave me with this. Actually, it was
> a silly oversight I never thought of: the EBCDIC char set is not contigous,
> and thus ('z'-'a'+1) != 26, an assumption which was made in mod_mime's
> hash() macro. So all this time I had a memory overwrite after mod_mime's
> init_mime() was called. I will have to dig thru the rest of apache to
> see if a similar assumption is used at other corners of the code.
> The appended patch changes nothing on an ASCII machine, but fixes the
> core dump on machines with character sets different from ascii (as long
> as 'a' < 'z', that is...).

There's another piece of code I was looking at... in util.c which does 'a'
and 'z' comparisons which should be rewritten using isalnum().

> So, after a little more testing (and adding the chunking charset
> protocol conversion), I could post the diffs to the list, maybe the
> coming week? It would be the first mainframe port which is standard part
> of apache. No fear, the differences are minor.

+1 on the patch you've included anyway, it's a nice trick.  I wouldn't
mind having one ebcdic port in apache.

Dean


Mime
  • Unnamed multipart/mixed (inline, None, 0 bytes)
View raw message