[distcc] (fwd from mbp@sourcefrog.net) Re: distcc

[Martin Pool]

----- Forwarded message from Martin Pool <mbp at sourcefrog.net> -----

From: Martin Pool <mbp at sourcefrog.net>
Subject: Re: distcc
Date: Mon, 17 Nov 2003 14:20:56 +1100
To: Jack Pavlovsky <man at oceanography.ru>
User-Agent: Mutt/1.5.4i

On 20 Oct 2003, Jack Pavlovsky <man at oceanography.ru> wrote:
> Compiling middleman (middle-man.sf.net), I've got an error:
> gcc -c -g -O2 -pthread -Iinclude -I. -Wall -D_GNU_SOURCE -DVERSION=\"1.9\" 
> -DHAVE_CONFIG_H -D_REENTRANT pcre/get.c -o pcre/get.o
> gcc -c -g -O2 -pthread -Iinclude -I. -Wall -D_GNU_SOURCE -DVERSION=\"1.9\" 
> -DHAVE_CONFIG_H -D_REENTRANT pcre/study.c -o pcre/study.o
> gcc -o pcre/dftables -g -O2 -pthread -Iinclude -I. -Wall -D_GNU_SOURCE 
> -DVERSION=\"1.9\" pcre/dftables.c
> gcc -c -g -O2 -pthread -Iinclude -I. -Wall -D_GNU_SOURCE -DVERSION=\"1.9\" 
> -DHAVE_CONFIG_H -D_REENTRANT pcre/pcre.c -o pcre/pcre.o
> pcre/pcre.c:261:24: chartables.c: No such file or directory
> distcc[26809] ERROR: compile on ivy failed
> 
> Running make again goes thru this place (this chartables.c is
> generated w/o compilation). Any way to circumvent this? (probably
> there're several proggies that have Makefiles like that - wouldn't
> be cool to leave things compiling for night only to find out that it
> bailed out in the middle). Not sure, maybe it's only related to make
> program itself. I use GNU make 3.80.

I don't understand what you mean.  You didn't show the command that
compiles chartables.c.

> I had installed distcc, and had runned distccd --daemon manually, at first.
> It worked. Then I had installed it to my initscripts (the same command).
> After reboot, the distccd processes are shown with ps, port 3632 is listened
> to, but such error occurs. If I kill distccd, and run distccd --daemon 
> again,
> it starts to work. As I wanted automatic startup, I tried to add it to 
> xinetd.
> I added stuff to xinetd config, then restarted xinetd - it worked. But upon
> reboot it doesn't work again (which is even more strange, since distccd is
> started from xinetd only when the distcc connects to port). Kill xinetd and
> run it again - stuff works. I couldn't comprehend this. Then the wild thing
> came to mind: the only difference in the initscript-started distccd or 
> xinetd
> and those I start in my shell is the environment - namely, the trouble was
> with initscript PATH not containing /usr/local/bin (where my gcc and g++ 
> are),
> while shell PATH contained it. I have set PATH in my initscripts, before 
> distccd
> is run, to contain /usr/local/bin and it works now. I don't think anything
> could (therefore should) be done in distccd code, but the stuff MUST be
> documented as a serious caveat. E.g., in your installation guide, you should
> write that initscript PATH should contain dir with gcc. Obviously many 
> people
> have gcc in /usr/bin, so this problem wasn't spotted earlier, but it should
> be fixed now.

OK.  The log file would probably have made this obvious.  

I just added this to the Problems page.  Would this have helped you
work it out?

  http://distcc.samba.org/problems.html#gcc-not-found

-- 
Martin 
                               linux.conf.au -- Adelaide, January 2004

----- End forwarded message -----
-- 
Martin 
                               linux.conf.au -- Adelaide, January 2004