RFC: Drop --disable-plugins to always enable plugins

[Jan Beulich]

On 06.08.2025 14:56, H.J. Lu wrote:
> On Wed, Aug 6, 2025 at 5:24 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 06.08.2025 14:19, H.J. Lu wrote:
>>> On Tue, Aug 5, 2025 at 11:10 PM Alan Modra <amodra@gmail.com> wrote:
>>>>
>>>> On Tue, Aug 05, 2025 at 10:01:59PM -0700, H.J. Lu wrote:
>>>>> On Tue, Aug 5, 2025, 9:34 PM Alan Modra <amodra@gmail.com> wrote:
>>>>>
>>>>>> On Tue, Aug 05, 2025 at 09:06:29PM -0700, H.J. Lu wrote:
>>>>>>> BTW, should we drop --disable-plugins to always enable plugins?
>>>>>>> Binutils should still work with compilers which don't support plugin.
>>>>>>
>>>>>> bfd/configure.ac disables plugins for various targets, so I think we're
>>>>>> stuck with BFD_SUPPORTS_PLUGINS == 0.  Otherwise something would need
>>>>>> to be done for hpux, vms and aix to fix their ar support.
>>>>>>
>>>>>
>>>>> Did you mean host? If a host doesn't support dlopen,
>>>>> it can be treated as if dlopen fails.
>>>>
>>>> No, target.  When BFD_SUPPORTS_PLUGINS ar and other binutils use
>>>> a default target of "plugin".  plugin_vec uses the generic archive
>>>> support, while their target vec uses another archive format.
>>>
>>> Do we enable plugin with --enable-targets=all on hpux?
>>>
>>> Do we disable plugin on Linux when hpux is also enabled?
>>>
>>> I enabled plugin to build cross-compiling to hpux on Linux.  There is
>>> no test failure.
>>>
>>> Can we require the generic archive support if input has plugin?
>>
>> Why would we want / need to force in plugin-related code when that's
>> entirely unused in some builds / configurations?
> 
> Plugin is supported on most (> 90%?) platforms/targerts.

And on any of them people could choose to turn it off via said configure
option, aiui.

> BFD_SUPPORTS_PLUGINS check makes code less maintainable.

That's true for many other optional things, too, I suppose. That's the
price to pay for offering people flexibility.

Jan