Re: Floppy problems after long uptime
On Thu, 18 Feb 1999, Paul Slootman wrote:
> The negative numbers made me suspect the worst, confirmed by a look into
> the source: jiffies are being stored into "int" values... My uptime was
> 30 days then (no, I haven't rebooted; I'm waiting to see whether the
> problem disappears again after another 30 days :-).
> Is the problem _this_ obvious, or am I missing something else? This is
> with 2.0.36, but the 2.2.1 source looks very similar.
> Note that in the line:
> now=2655075170 last interrupt=-1639977453 ...
> if you cast 2655075170 into an int, you get -1639892126.
It could very well be that obvious. I found a similar problem with the
future version of ping (as well as ping6 for IPv6) where half of the code
properly used sizeof(struct timeval) to get the proper amount of chars to
use to store the time, but the other half just assumed that the size of
the struct was 8.
Seems like alot of revision has been done, but an equal amount of things
was overlooked in both kernel and application source code quite often.
I could work up a diff for you for 2.2.1 if you'd like (and change the
glibc 2.1 package as well, when I can)....