Re: A symbol version patch for glibc 2.x compatibility

["H . J . Lu" <hjl at valinux dot com>]

On Sat, Nov 25, 2000 at 12:52:06PM -0800, David O'Brien wrote:
> On Sat, Nov 25, 2000 at 10:40:06AM -0800, H . J . Lu wrote:
> > On Sat, Nov 25, 2000 at 03:53:32AM -0800, David O'Brien wrote:
> > > > I don't understand why David kept asking for Linux to change.
> > > 
> > > Because I'd like to follow something written down in a specification or
> > > standard.  
> > 
> > Please do. 
> 
> I would, but there isn't anything written in a standard about how to
> determine the syscall API in an ELF binary.
> 
> > Both Ulrich and I have told you that your interpretation of gABI is
> > incorrect.
> 
> Only now.  I was somewhat led me astray with these emails:
> 
>     Date: Tue, 11 Apr 2000 20:50:59 -0700
>     From: "H . J . Lu" <hjl@lucon.org>
>     To: David O'Brien <obrien@FreeBSD.org>
>     Cc: binutils@sourceware.cygnus.com
>     Subject: Re: Advice on the prefered way to brand ELF binaries as
>     needed on FreeBSD
> 
>     On Tue, Apr 11, 2000 at 08:35:03PM -0700, David O'Brien wrote:
>     > FreeBSD brands all ELF files so the image loader knows what type of
>     > binary it is and use the proper compatibility layer if the binary
>     > is not a native FreeBSD one.
>     ..snip..
>     Why invent new thing when EI_OSABI is available?
> 
> This is the main email that lead me down the path thinking EI_OSABI was
> usable as a syscall API branding method.
> 
> 
>     Date: Tue, 2 May 2000 20:44:48 -0700
>     From: "H . J . Lu" <hjl@lucon.org>
>     To: "David O'Brien" <obrien@NUXI.com>
>     Subject: Re: Changes to include/elf/common.h
>     Message-ID: <20000502204448.A20372@lucon.org>
> 
>     If the EI_OSABI field is zero, it should be treated as SVR4. That is
>     the idea how it should be used.
> 
> You really meant "generic ELF" I guess....
> 

It is OK for you to use it on FreeBSD. But it is not OK for you to
ask Linux to change.

> 
>     Date: Tue, 2 May 2000 21:25:38 -0700
>     From: "H . J . Lu" <hjl@lucon.org>
>     To: "David O'Brien" <obrien@NUXI.com>
>     Subject: Re: Changes to include/elf/common.h
>     Message-ID: <20000502212538.A20536@lucon.org>
> 
>     > If toolchain developers are unwilling to use the EI_OSABI field,
>     > why was it introduced?
> 
>     The NOTE solution was developed before EI_OSABI from SCO/HP.
> 
> Here you did not tell my what I was trying to use EI_OSABI for was wrong.
> 
> 
> > > names the section ".note.ABI-tag", NetBSD ".note.ident".  I made FreeBSD
> > > follow Linux with ".note.ABI-tag".  Or does one search thru all the
> > > PT_NOTE type sections looking for a type of 0x01?  Search all the PT_NOTE
> > > sections looking for a name string you recognize?
> > > 
> > 
> > We can make a proposal to gABI.
> 
> What method would what would you suggest.  Any above or another
> algorithm?
>  

I don't know. And I don't know if gABI/psABI people care about the
static binaries. As I undersand, static binaries are almost not
supported under Solaris 2.7 and above.


-- 
H.J. Lu (hjl@valinux.com)