Re: PR#502: timeout problems

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

>Looks like something wrong in Roy's latest round of timeout changes.  If I
>get a chance I'll go through them line by line looking for irix specific
>problems.  Ugh. 

I looked and couldn't find anything wrong during the period in which
the scoreboard would show SERVER_BUSY_KEEPALIVE.  Is it possible that
the server is blocking on the sync_scoreboard_image, or just failing
to update the scoreboard when it is actually in WRITE status?
I notice that there is no sync_scoreboard_image() after the first
update_child_status in the read_request loop of http_main.c, but that
should not effect shared-memory, right?

....Roy

Re: PR#502: timeout problems

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

update_child_status itself writes the new child status to disk (and does a
sync_scoreboard_image), so there shouldn't be a need to
sync_scoreboard_image.  In any event it shouldn't affect shared-memory at
all...

Dean

On Wed, 30 Apr 1997, Roy T. Fielding wrote:

> >Looks like something wrong in Roy's latest round of timeout changes.  If I
> >get a chance I'll go through them line by line looking for irix specific
> >problems.  Ugh. 
> 
> I looked and couldn't find anything wrong during the period in which
> the scoreboard would show SERVER_BUSY_KEEPALIVE.  Is it possible that
> the server is blocking on the sync_scoreboard_image, or just failing
> to update the scoreboard when it is actually in WRITE status?
> I notice that there is no sync_scoreboard_image() after the first
> update_child_status in the read_request loop of http_main.c, but that
> should not effect shared-memory, right?
> 
> ....Roy
>