[sharp-discuss] sharp-discuss Digest, Vol 41, Issue 2

[Clemens Vonrhein]

Hi Mark,

thanks for the feedback - and sorry for the late answer (and
appearance of your post on the list: it was stuck in a queue).

On Sat, Aug 16, 2008 at 12:44:44AM +0200, Mark Brooks wrote:
> Hi,
>     I also had a bit of a struggle installing SHARP on Ubuntu, (Hardy Heron
> currently, but Gutsy also previously), but got it working eventually.
> 
> The default httpd binary core dumped, but the answer for me was to move the
> httpd binary (unpacked from "helpers_server.linux.tar.gz") to somewhere
> else, then use the httpd.SuSE-9.3 binary
> e.g.:
> 
> %mv httpd httpd.distrib
> %mv httpd.SuSE-9.3 httpd
> 
> ...then use the kill_server/restart_server scripts to start the server.
> I have Sharp running on port 8080, and the vanilla Apache2 from Ubuntu
> running on 80, and they don't get in the way of each other.

Yes: it is fairly easy to have lots of different httpd (of different
flavour/version) running in parallel. Apart from the port number (we
use 8080 because it is an unprivileged port, i.e. > 1024) the location
of the log files the httpd wants to write should also be unique (for
SHARP/autoSHARP this is in $BDG_home/sushi/logs).

> I guess that recompiling either Apache2 or Apache 1.3.X is overkill, and
> instead you should concentrate on getting the available Apache 1.3 binaries
> distributed with SHARP to run.

Ah - but that is the problem: we can distribute several httpd binaries
for a variety of Linux distributions, but this will only be covering
the distros we actually have access to.

The problem is that the httpd is so close to the core of a Linux
system, that it is a combination of 32/64bit, kernel, glibc version
etc that seems to determine if the binary will run or
not. Furthermore, the whole design of the Apache httpd is intended to
have it very dynamic - to load only modules that are needed (again
something very closely tied to the actual OS). So compiling a truly
static binary for httpd that will run on all machines (or at least on
all 32bit or 64bit machines) seems near impossible.

There was a time, when a single httpd binary (compiled on a RedHat 7.3
system) was running happily on all systems out there at the time
... but this is a few years ago unfortunately.

So at the moment it seems that the easiest is to just recompile the
official Apache source code on the machine the SHARP/autoSHARP server
should be running. It is sure to give a working binary.

> I don't have access to the machine in question right now, but I
> think that Apache should give a logfile when it dies.

Ideally it does: but if you're running a binary from a different Linux
system it might just crash (and not get to the point where it can
write a nice, meaningful message into
$BDG_home/sushi/lgs/error_log). I've seen cases, where running

  % $BDG_home/helpers/linux/httpd -v

runs fine but when trying to use such a binary for actually starting
the server it crashes ...

> Good luck,

Thanks - as a side-note: the next CCP4 release (6.1.0 - or the current
pre-release versions) contains a CCP4i interface to at least
autoSHARP ... no httpd required there ;-)

Cheers

Clemens

-- 

***************************************************************
* Clemens Vonrhein, Ph.D.     vonrhein AT GlobalPhasing DOT com
*
*  Global Phasing Ltd.
*  Sheraton House, Castle Park 
*  Cambridge CB3 0AX, UK
*--------------------------------------------------------------
* BUSTER Development Group      (http://www.globalphasing.com)
***************************************************************