[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha-dev
Subject:    Re: [Linux-ha-dev] Re: flock [Was: Re: FreeBSD compatability]
From:       Alan Robertson <alanr () unix ! sh>
Date:       2001-05-22 13:03:04
[Download RAW message or body]

David Lee wrote:
> 
> On Thu, 17 May 2001, Alan Robertson wrote:
> 
> > David Lee wrote:
> > > [...]
> > > Shouldn't something at least be automatic?  On Solaris (also *BSD?) there
> > > is no "flock()", and the compilation failures are hard:
> >
> > Fcntl() is the "standard" way to do this.
> >
> > You can't use flock() if it isn't there...
> >
> > You can't use fcntl() if it isn't ther
> >
> > It seems a shame to throw away the flock() code, since it's already there.
> >
> > If both are present, *normally* you prefer the flock - but GFS can't yet
> >       use it.
> >
> > I probably don't *care* if GFS can't yet use it - but someone else might...
> >
> > So, the following method is probably sufficient:
> >
> >       Perform an automatic configure check for the two locking methods,
> >               and always use fcntl() if both are present
> >
> > The following method is "perfect":
> >
> >       Perform an automatic configure check for the two locking methods,
> >               and use fcntl() if both are present unless overridden by
> >               a configure option of --disable-fcntl-locking.
> >
> > Does this sound right?
> 
> Seems OK.  I've opted for your "sufficient" method, but you're welcome
> to enhance it for your "perfect" one.
> 
> Warning: It is completely and utterly untested, although the "HAVE_FCNTL"
> side does at least compile on Solaris.
> 
> One note of caution though.  There is something about the overall build
> that causes the end-user's "config.h" not to be rebuilt.  This is needed
> to pick up the new "HAVE_FLOCK" and "HAVE_FCNTL" defines.  So it is
> important for us, at this stage of the game, to ensure that this happens
> in our own particular CVS fetch/update:
> 1. A clean CVS fetch should be OK (but excessive!);
> 2. A "./bootstrap" seems not sufficient;
> 3. I had to ensure "config.h" was logically removed (I did a "mv" sideways
>    rather than a "rm" (a) just in case (b) so I could "diff" them...)

I just do a "make clean" before I do a bootstrap.  That seems to do it.

I incorporated your fix into CVS.  While I was there, I cleaned up the C++
comments in the code, and a couple of other minor things.

I also give a #error if neither of the locking methods is available.

	-- Alan Robertson
	   alanr@unix.sh
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.community.tummy.com
http://lists.community.tummy.com/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/

[prev in list] [next in list] [prev in thread] [next in thread]