Re: [Mspgcc-users] long long support for printf() family (EXTRA: gcc internal error: Segmentation fault)

[Freddy S. <fr...@du...>]

Hello,

On Thu, Apr 24, 2008 at 05:22:28PM +0200, Freddy Spierenburg wrote:
> The attached patch does add this support.

Unfortunately the yesterday patch featured a grave bug. I
overlooked the _ulong variable. Hence my patch could not handle
the MSB part of the long long, pretty pointless.

In solving this bug I could take two directions. Make _ulong a
long long or invent an extra _ulonglong. For performance reason I
chose the last route. I think those few extra conditional jumps
are worth not having to do 64-bit arithmetic on all integers,
please correct me if I'm wrong. Or perhaps you all think I'm
braindead anyway, trying to do 64-bit arithmetic on a simple
16-bit CPU. :)

Next I stumbled upon a gcc problem. I got the error below, while
compiling vuprintf.c:

	stdlib/vuprintf.c: In function `vuprintf':
	stdlib/vuprintf.c:605: internal error: Segmentation fault

A little bit of investigation showed me the line below was
causing the trouble:

	notlastdigit=(_ulonglong>=base);

For some strange reason gcc has trouble comparing the long long
variable _ulonglong to the base. Unfortunately I do not have the
guts (yet) to dive into the gcc source. Does anybody have an idea
what is going on?

I changed the line into the line below and that works without
trouble:

	notlastdigit=((_ulonglong>>32)>0)||((unsigned long)_ulonglong>=base);

But the elegant department is probably raising an eyebrow or two.

Anyway, please find attached a patch that this time correctly
adds the long long support to the printf() family. Any comments
are appreciated.


-- 
$ cat ~/.signature
Freddy Spierenburg <fr...@du...>  http://freddy.snarl.nl/
GnuPG: 0x7941D1E1=C948 5851 26D2 FA5C 39F1  E588 6F17 FD5D 7941 D1E1
$ # Please read http://www.ietf.org/rfc/rfc2015.txt before complain!