Re: [STATUS] Thu May 8 08:31:41 PDT 1997

["Roy T. Fielding" <fi...@kiwi.ICS.UCI.EDU>]

>  * Dean's [PATCH] PR#502: timeout problems
>    <Pi...@twinlark.arctic.org>
>    Status: Dean +0
>    User reported it didn't help with IdentityCheck on.  Ugh.

I looked at this patch and it is doing the wrong thing anyway.
Instead of saving the last alarm length in a global variable, it
should just save the value returned by alarm() and the handler
returned by signal -- if the former is greater than zero, then
restore handler and the remaining alarm time when done.

So, -1 on this patch anyway.

....Roy

Re: [STATUS] Thu May 8 08:31:41 PDT 1997

[Dean Gaudet <dg...@arctic.org>]

I wasn't sure if saving the return value of alarm() would be portable,
ditto for the return value of signal.  Given the time frame that we're
trying to fix this in I went with a solution that would port.  I also know
it wasn't perfect, and said that... but it doesn't work anyhow.

Dean

On Fri, 9 May 1997, Roy T. Fielding wrote:

> >  * Dean's [PATCH] PR#502: timeout problems
> >    <Pi...@twinlark.arctic.org>
> >    Status: Dean +0
> >    User reported it didn't help with IdentityCheck on.  Ugh.
> 
> I looked at this patch and it is doing the wrong thing anyway.
> Instead of saving the last alarm length in a global variable, it
> should just save the value returned by alarm() and the handler
> returned by signal -- if the former is greater than zero, then
> restore handler and the remaining alarm time when done.
> 
> So, -1 on this patch anyway.
> 
> ....Roy
>