httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Randy Terbush <ra...@zyzzyva.com>
Subject Re: graceful restarts, take 5
Date Tue, 22 Apr 1997 17:35:11 GMT

I agree completely. I've been giving this argument some thought 
this morning. I think Dean has sorted out some issues that continue 
to plague busy servers. Same thing with Roy's timeout changes. We 
need to include it despite the end result of holding off the 1.2.


> I've just tested this patch and then read it line-by-line and can
> honestly say that I wouldn't run a big (ISP type) server without it.
> In fact, after testing it versus the current snapshot, I think we need
> to either include this (with change noted below) in 1.2 or must remove
> all trace of the graceful restart stuff before 1.2, since the current
> code is much worse than I thought before testing it.
> 
> Aside from that, the risk of this patch breaking things is about
> the same as the timeout patch we already committed, since the core
> of both involve getting the signal handling right.  In other words,
> I'd rather delay 1.2b9 for this than let it slide.
> 
> I do have one incidental change in alloc.c:
> 
> >***************
> >*** 201,206 ****
> >--- 199,210 ----
> >  
> >    /* Nope. */
> >  
> >+   if (min_size + BLOCK_MINFREE < BLOCK_MINALLOC) {
> >+     min_size = BLOCK_MINALLOC;
> >+   } else {
> >+     min_size += BLOCK_MINALLOC;
> >+   }
> >+ 
> >    return malloc_block (min_size);
> >  }
> 
> I am pretty sure the above should be
> 
>      if (min_size + BLOCK_MINFREE < BLOCK_MINALLOC) {
>        min_size = BLOCK_MINALLOC;
>      } else {
>        min_size += BLOCK_MINFREE;
>      }
> 
> (i.e., the last BLOCK_MINALLOC is wrong).
> 
> .....Roy




Mime
View raw message