[torqueusers] [torquedev] Release Candidate for TORQUE 2.5.8

Ken Nielson knielson at adaptivecomputing.com
Wed Aug 24 17:29:47 MDT 2011



----- Original Message -----
> From: "Glen Beane" <glen.beane at gmail.com>
> To: "David Beer" <dbeer at adaptivecomputing.com>
> Cc: torqueusers at adaptivecomputing.com
> Sent: Wednesday, August 24, 2011 5:08:41 PM
> Subject: Re: [torqueusers] [torquedev] Release Candidate for TORQUE 2.5.8
> On Aug 24, 2011, at 6:28 PM, David Beer <dbeer at adaptivecomputing.com>
> wrote:
> 
> >
> >
> > ----- Original Message -----
> >> On Aug 24, 2011, at 5:10 PM, David Beer
> >> <dbeer at adaptivecomputing.com>
> >> wrote:
> >>
> >>>
> >>>
> >>> ----- Original Message -----
> >>>> On Aug 24, 2011, at 4:31 PM, David Beer
> >>>> <dbeer at adaptivecomputing.com>
> >>>> wrote:
> >>>>
> >>>>>
> >>>>>
> >>>>> ----- Original Message -----
> >>>>>> On Wed, Aug 24, 2011 at 1:54 PM, Ken Nielson <
> >>>>>> knielson at adaptivecomputing.com > wrote:
> >>>>>>
> >>>>>>
> >>>>>> There is a new release candidate for TORQUE available. This has
> >>>>>> the
> >>>>>> fixes for the compiler warnings with the exception of
> >>>>>> process_request.c, function get_creds line 288. dereferencing
> >>>>>> type-punned pointer will break strict-aliasing rules.
> >>>>>>
> >>>>>> in my opinion, this is a warning that shouldn't be ignored
> >>>>>> indefinitely (not that anyone has suggested it). I think it
> >>>>>> would
> >>>>>> be
> >>>>>> good to fix before the official release. I would also add
> >>>>>> -fno-strict-aliasing to my CFLAGS until it were fixed.
> >>>>>>
> >>>>>> I would volunteer to the look at the code responsible for the
> >>>>>> warning,
> >>>>>> but I can't guarantee I would have a chance any time soon.
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> torquedev mailing list
> >>>>>> torquedev at supercluster.org
> >>>>>> http://www.supercluster.org/mailman/listinfo/torquedev
> >>>>>
> >>>>> I tried to look at the error myself, but I didn't see any
> >>>>> compiler
> >>>>> warnings when I compiled the code:
> >>>>>
> >>>>> $ gcc --version
> >>>>> gcc (Ubuntu/Linaro 4.5.2-8ubuntu4) 4.5.2
> >>>>> Copyright (C) 2010 Free Software Foundation, Inc.
> >>>>> This is free software; see the source for copying conditions.
> >>>>> There
> >>>>> is NO
> >>>>> warranty; not even for MERCHANTABILITY or FITNESS FOR A
> >>>>> PARTICULAR
> >>>>> PURPOSE.
> >>>>>
> >>>>> I used to think that gcc warnings were a moving target (as
> >>>>> Michael
> >>>>> Jennings points out) because they added warnings, but it appears
> >>>>> they also add and then remove things from the warning lists.
> >>>>>
> >>>>> As far as the greater debate of whether or not gcc warnings
> >>>>> should
> >>>>> be enabled by default, I think it makes a lot of sense to only
> >>>>> have
> >>>>> developers compile with warnings enabled and let users just
> >>>>> compile
> >>>>> and run the code. When I'm a user, I only care about warnings
> >>>>> that
> >>>>> cause issues with the software. I think that this warning would
> >>>>> have
> >>>>> been caught as well except that it is only a warning for certain
> >>>>> versions of gcc.
> >>>>>
> >>>>> --
> >>>>> David Beer
> >>>>> Direct Line: 801-717-3386 | Fax: 801-717-3738
> >>>>>   Adaptive Computing
> >>>>>   1656 S. East Bay Blvd. Suite #300
> >>>>>   Provo, UT 84606
> >>>>>
> >>>>> _____________________________________________
> >>>>
> >>>>
> >>>> What are your compiler flags?
> >>>
> >>> gcc -g -W -Wall -Wno-unused-parameter -Wno-long-long -pedantic
> >>> -Werror -D_LARGEFILE64_SOURCE -o .libs/printjob printjob.o
> >>> ../lib/Libpbs/.libs/libtorque.so -Wl,--rpath -Wl,/usr/local/lib
> >>
> >> I think this warning would require -O2 or higher (or explicit
> >> strict
> >> aliasing optimization turned on)
> >
> > I don't believe that optimization levels change what gcc considers a
> > warning, but I recompiled it with -O2 and didn't receive a warning
> > that way either.
> >
> > gcc -g -O2 -W -Wall -Wno-unused-parameter -Wno-long-long -pedantic
> > -Werror -D_LARGEFILE64_SOURCE -o .libs/hostn hostn.o
> > ../lib/Libpbs/.libs/libtorque.so -Wl,--rpath -Wl,/usr/local/lib
> 
> Optimization level can affect what is considered a warning -- without
> strict aliasing optimization enabled there are no strict aliasing
> rules to break

I think this may be a GNU compiler bug. It looks like struct ucred does not get defined from bits/socket.h unless _GNU_SOURCE is defined. See my e-mail for more details.

Ken


More information about the torqueusers mailing list