Re: New failures in linker testsuite for arm-elf targets

[Nick Clifton <nickc at redhat dot com>]

Hi Christophe,

> [maybe you'd want me to post this on the mailing list? if so, let me know]

Yes.  (I have added the list as a CC to this reply, so we can just carry 
on from here.

> >   In case you are not already looking at this, please could you try
> >   building an ELF targeted ARM toolchain (as opposed to an EABI one,
> >   eg arm-elf or arm-netbsdelf) and then trying the linker testsuite
> >   with your new farcall tests in it.  If you get the same results as
> >   me then you will see that there are a whole bunch of new failures
> >   there.  Failures which are not present when tested with an arm-eabi
> >   toolchain...

> I do confirm that there are new failures with the arm-elf target.
> On my side, I also noticed some failures without my patch:
> arm-pic-veneer, "preempt thumb symbol", script-type, "EABI attribute 
> merging", "EABI attribute merging 2"
>
> The issues of the 1st two seem to also explain the failures with the 
> tests I added or modified. Indeed, with arm-elf, it seems to me that the 
> attributes sections from the ELF file are not present, so some of the 
> information used to support far calls is not available.

Exactly.  In fact I think that this is the way to go. ie to say that 
your farcall support is only for EABI ARM targets and not other ELF 
based ARM targets.
(There are other ways of implementing farcall support for the ARM.  I 
wrote a patch for GCC ages ago which implements it, but it was never 
accepted into the mainline sources).

> I am not sure about how to proceed.

I suggest a simple solution - rearrange the order of the tests in 
arm-elf.exp so that the farcall support ones are all in the 
"armeabitests" list.  Add a comment explaining that the farcall support 
is only EABI ARM (to arm-elf.exp and also ld.texinfo) and test to make 
sure that both arm-eabi and arm-elf toolchains now have reasonable 
results from the linker testsuite.

Cheers
  Nick