Re: refreshIsCachable Suggested Patch

Probably correct on some compilers. However, this is far from the only
place where it is assumed the compiled manages a sequence point at ? and
only evaluates the correct alternative.

According to all C documentation I have been able to find this should be
the case, but it has been indicated that not all compilers follows this
rule in all cases. If they don't I think the compiler is broken until
proven othervise.

I.e. to all my knowledge the following two code fragments SHOULD be
identical in the absense of sideefects in the expression d.

  d = a ? b : c;

  if (a)
    d = b;
  else
    d = c;

Or more expressive, if you write

   a ? b() : c()

   then b is called IFF a is true, and c called IFF a is false

   a ? b->c : d

   then b is dereferenced as part of the evaluation of b->c IFF a is
true. If a is false then only d is evaluated and b is never
dereferenced.

   a ? b++ : c++

   then the sideeffect b++ is executed IFF a is true, and c++ IFF a is
false.

   b=1;
   a[b++] ? b : c

   should evaluate to 2 (b++) if a[1] is true, c otherwise.

If you have any C alnguage information which contradicts this then I
would be very glad to know.

--
Henrik Nordstrom
Squid Hacker
Zhang Luwei wrote:
> 
> Hi all,
> 
>         I am Zhang Luwei again, the Masters Research Student from the
> National University of Singapore (NUS). I have just looked upon the patch
> at the url below for refreshIsCachable method:
> 
> http://squid.sourceforge.net/hno/patches/squid-2.1.PATCH2.dont_cache_unrefreshable.patch
> 
> In the code under the method refreshIsCachable, we have:
> 
> rep = entry->mem_obj ? entry->mem_obj->reply : NULL;
> 
> Would like to suggest a better patch. My suggestion is based on the
> assembly theory. The patch above, when transformed into assembly language,
> may cause instability in the proxy server.
> My suggestion to a better way to write the line will be:
> 
> if (entry->mem_obj && entry->mem_obj->reply && (entry->mem_obj->reply->content_length == 0))
>         return 0;
> 
> Hope my suggestion helps,
> Hope to hear from all of you soon.
> 
> Best regards,
> Truly yours,
> Luwei