> | this overflow of stack in assembly:
> |
> | read_infos:
> | .LLFB46:
> | .loc 1 197 0
> | !#PROLOGUE# 0
> | save %sp, --2147483544, %sp
> |
> | Notice the double negative, produces by overflowing the stack pointer,
>
> I'm not very familiar with compilers internals but (I once made
> a code-generator, that's all) but how can a double negative can
> happen ?
It is basically this:
fprintf(asm_file, "\tsave\t%sp, -%d, %sp\n", size_of_stack);
So the first negative is hardcoded (because the stack decreases down
always) and the second one is just the result of SSIZE_MAX plus some
(registering saving, etc), which overflows to a negative value.
>
> | which only uses, internally, a signed int. That is plenty enough to hold
> | a normal, legal stack, but not really enough for your needs.
>
> Ok. So wouldn't it be more usefull to get a compiler message that
> is more explicit ?
I agree. I think the compiler should produce an error instead of the --.
That's probably something that each arch should handle though.
> | If you need that much (I can't see that you would), then malloc it.
>
> I don't need that much, in fact. I just need to ensure that what
> is send from one side of a client/server connection is fully
> received at the other side...
Then probably a 1024 byte buffer, and a loop to read/process it until
it's all read. You'd have to make sure the connection was non-blocking
though.
--
Debian - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/
