Sender: rich AT phekda DOT freeserve DOT co DOT uk
Message-ID: <395095E9.F94D70EC@phekda.freeserve.co.uk>
Date: Wed, 21 Jun 2000 11:16:09 +0100
From: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.14 i586)
X-Accept-Language: de,fr
MIME-Version: 1.0
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
CC: djgpp AT delorie DOT com, rich AT phekda DOT freeserve DOT co DOT uk
Subject: Re: Inline asm: lcall & various binutils versions
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000621071322 DOT 8230C-100000 AT is>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp AT delorie DOT com

Hello.

Eli Zaretskii wrote:
> 
> On Tue, 20 Jun 2000, Richard Dawe wrote:
> 
> > #if GAS_MAJOR >= 2 && GAS_MINOR > 8
> >                "nop   \r\n"
> > #endif        /* IFDEFTEST */
> >                "nop   \r\n");
> 
> I understand that this is just an example, because in the actual problem
> you had you will need a different condition for the versions (e.g., if
> GAS_MAJOR is 3 or more, GAS_MINOR is not important).

Yes, this is an example. For libsocket/libwin I would use:

#if GAS_MINOR == 2 && GAS_MINOR == 8 && GAS_MINORMINOR == 1
    "lcall _entrypoint"
#else
    "lcall *_entrypoint"
#endif

Note that I haven't checked that 'as --version' with binutils 2.8.1 gives
(2, 8, 1). I also assume no-one uses binutils versions earlier than 2.8.1
(reasonable, I think).

More about the example: when you define IFDEFTEST, you get three 'nop's in
the output (as seen by 'objdump -d'); otherwise, you get two 'nop's.

> > Is this satisfactory? My sed knowledge is limited (*), but this seems
> > to do the trick.
> 
> The Sed scripts can be improved slightly, but they seem to be correct.

Sure, please tell me how, when you have time.

> > [ If you put '#ifdef' in the inline assembly, then it becomes an
> > assembly comment and does nothing.
> 
> AFAIK, inline assembly doesn't go through cpp; GCC emits it in the
> form of preprocessed assembly.  So you cannot have any preprocessor
> directives inside the asm() block.

Yes, this is what I saw when using the -save-temps option on gcc. The
inline assembly appears as-is (or as-was).

Re: safety - I think we're still relying on binutils 2.8.1 to generate the
right code for 'lcall _entrypoint' (i.e. an absolutely-addressed call),
it's just that the AT&T syntax to tell gas to do an absolutely-addressed
call doesn't work (hence the current problem). So, at least we can be safe
with binutils 2.9.5.1 beta. ;)

Bye, Rich =]

-- 
Richard Dawe
[ mailto:richdawe AT bigfoot DOT com | http://www.bigfoot.com/~richdawe/ ]