Policy for removing target support

[Jan Beulich]

On 11.09.2025 10:48, Rainer Orth wrote:
> Hi Jan,
> 
>>> true, but I still wonder what's the point between 10 and 14 years after
>>> the fact?  I can fully understand conservatism in this respect, as is
>>> evident by me supporting old OS versions in GCC for quite some time, but
>>> at some point you have to make a cut.  I still maintain that untested
>>> code is useless if not harmful by suggesting support that doesn't
>>> actually work.
>>
>> First, despite the growing testsuite, lots and lots of code are untested
> 
> ... which is no reason to have more if it can be avoided ;-)
> 
>> Then I'm still feeling like you're mixing the "building on" and "building
>> for" aspects. IOW I still don't understand what parts of untested code
>> you'd like to remove as a result on "building on" no longer being possible.
>> Or else why you brought up the "building on" aspect in the first place.
> 
> But what's the point of even building cross-binutils if the target
> itself is no longer used and hasn't been for a long time?
> 
>> As to the "building for", my earlier question still didn't get a clear
>> answer (or maybe what seemed clear to you didn't end up clear for me): As
>> indicated, I'm having trouble matching the version numbers (e.g. 10) with
>> what we have in the two configure scripts in question. Which in turn
>> makes it impossible for me to understand what parts you think can be
>> dropped there.
> 
> Right now, I'd be dropping primarily the bfd/config.bfd and
> ld/configure.tgt snippets pertaining to pre-Solaris 10 versions.

I understand that; what I don't understand is which of the entries this
covers. We don't have anything named e.g. "solaris9" in binutils.

Jan

>  That's
> not too much, but would could considerably simplify the script by not
> littering them with unused parts.  There may be more, but at least it's
> a start.  It's equivalent to previous such efforts in both GCC and GDB:
> once the configure support is gone, you no longer need to worry about
> those old versions (which haven't been considered anyway, but nobody can
> tell if they still work since they are untested) and remove code
> specific to them if you come across it.
> 
> As I said, I very well understand the point about being conservative,
> but there should be a limit and I try to understand what that is.
> 
> 	Rainer
>