httpd performance vs. debuggability

[Richard Hubbell <ri...@yahoo.com>]

Hello all,

Are there any tricks for improving the debuggability of httpd without the detriment to performance?  -g just about cuts capacity in half.

I was thinking maybe if we could build just the module that we think is the offending module with -g and the rest non-debug that it might get us what we want.  But is that feasible; will it even work?
Thanks.


      

Re: httpd performance vs. debuggability

[Niklas Edmundsson <ni...@acc.umu.se>]

On Fri, 13 Jun 2008, William A. Rowe, Jr. wrote:

>> Are there any tricks for improving the debuggability of httpd without the 
>> detriment to performance?  -g just about cuts capacity in half.
>
> -g should do no such thing.  -g -O0 would, obviously, but some people are
> patient enough to struggle through an optimized crash.

Our site-default compile has CFLAGS='-O2 -march=i686 -g' and seems to 
work fine. Do grep CFLAGS config.log to see what configure actually 
used.

/Nikke
-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
  Niklas Edmundsson, Admin @ {acc,hpc2n}.umu.se      |     nikke@acc.umu.se
---------------------------------------------------------------------------
  "I'm SO gonna bite you" - Harmony
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

Re: httpd performance vs. debuggability

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Richard Hubbell wrote:
> Hello all,
> 
> Are there any tricks for improving the debuggability of httpd without the detriment to performance?  -g just about cuts capacity in half.

-g should do no such thing.  -g -O0 would, obviously, but some people are
patient enough to struggle through an optimized crash.

If you seriously see such a problem, create the binaries optimized but
with -g, then strip them before you deploy them.

But DON'T throw away the originals - you can use them to then work through
any dump you create through gdb.  The fact that the running binaries were
stripped doesn't matter.