[lsb-discuss] LSB -- next 4.1

Dallman, John john.dallman at siemens.com
Thu May 28 08:08:35 PDT 2009

Well, for any given release of a Linux distribution, there is a 
compiler that is in some senses part of the OS - the one that 
was used to build that release. 

I absolutely agree that installing another compiler because an 
application needs its support libraries is unacceptable. 

My problem comes if I ship a libgomp.so accompanying my product
(which is a library, not an application) and it needs some non-
LSB interface. Suddenly, I'm no longer LSB-compliant. And at 
present, libgomp can't be LSB-compliant owing to syscall(). 

Even with an LSB-compliant libgomp, I need to be rather careful 
about where I tell my customers who write applications to install
it. Some creative genius will doubtless try to replace the 
copy in /usr/lib64, rather than keeping it as part of their app. 

John Dallman
Parasolid Porting Engineer

Siemens PLM Software
46 Regent Street, Cambridge, CB2 1DP
United Kingdom
Tel: +44-1223-371554
john.dallman at siemens.com

> -----Original Message-----
> From: Robert Schweikert [mailto:rschweikert at novell.com]
> Sent: Thursday, May 28, 2009 3:08 PM
> To: Dallman, John
> Cc: Wichmann, Mats D; LSB
> Subject: Re: [lsb-discuss] LSB -- next 4.1
> IMHO the compiler is NOT part of the OS. In addition as an application
> or library provider I would think that having to install a compiler to
> use the app or library is undesirable. Why would one put the user
> through this? From a user perspective I would want to install the app
> or
> library and just use it without having to jump through hoops and
> install
> a compiler in addition just to get the correct RTE.
> Unlike the STL, OpenMP is not part of a given language, it is more
> an aspect.
> I suppose if everyone agrees, i.e. all OpenMP library implementers, on
> the interfaces an OpenMP library should provide it might be a lot
> easier
> to consider for the LSB. With a bit of scripting one could at least
> easily compare the interfaces provided by libgomp, libguide, and
> Portland Group's version. That would cover the major compiler vendors.
> In any event having shippable implementations will lead to a faster
> success than waiting for inclusion of OpenMP in the LSB, I think.
> Robert
> Dallman, John wrote:
> > Robert Schweikert [mailto:rschweikert at novell.com] wrote:
> >
> >> IMHO we want to have the various openMP RTE libraries be LSB
> compliant
> >> such that they can be shipped.
> >
> > We'd feel extremely nervous about doing that. Run-time libraries
> > that seem much more like parts of the OS than of an application to
> me.
> > Feel free to explain why I'm wrong.
> >
> --
> Robert Schweikert                           MAY THE SOURCE BE WITH YOU
> Software Engineer Consultant                          LINUX
> rschweikert at novell.com
> 781-464-8147
> Novell
> Making IT Work As One

More information about the lsb-discuss mailing list