Re: [patch] ARM: -mdisable-got adapted for gcc-3.0.2

[Adrian von Bidder <avbidder at acter dot ch>]

Nick Clifton wrote:
> 
> Hi Adrian,
> 
> > The first hunk of the patch is against config.sub. I have submitted it
> > to config-patches, and it has been approved; it's included here since I
> > work with 3.0.2 release.
> 
> Great - one question though - do you have a gcc copyright assignment
> on file with the FSF ?

Not yet. The one on binutils is pending (on my desk), too. I see I'll
definitely have to do the paperwork soon. (As we are a Linux based
company here I don't see any problems from my boss, so it's just the
matter of actually sending the form...)

> I am looking through your patch and it seems OK so far, although I
> have a few more tests to run.

I'm quite sure it's full of bugs ... I'm hacking gcc 'blindly' right now
- adapting the patch from Vladim Lebedev, guessing at what the whole
thing actually does. So I'm grateful for any input at all.

I guess the bit in legitimize_pic_address is quite close to what I want
it to do - though it does extra handling only for SYMBOL_REF.  Is extra
handling needed for LABEL_REF, too? (Vladim's patch had something, but
I'm /think/ it would only get executed when flag_pic is not set in
legitimize_pic_address, i.e. never)

But I'm absolutely not sure
 - whether I may just use SYMBOL_REF_FLAG as I did in
ENCODE_SECTION_INFO (I guess if I must not do that, I shall have to go
the hard way, changing symbol names around)
 - how to do the distinction between symbols/labels in the text segment
and symbols/labels in the data segment properly. If I can do this within
legitimize_pic_address, I wouldn't need to touch ENCODE_SECTION_INFO at
all.

greets from Zürich
-- vbi