Re: [PATCH v7 2/8] Remove __morecore and __default_morecore

[Florian Weimer via Libc-alpha <libc-alpha@sourceware.org>]

* Siddhesh Poyarekar:

> On 7/9/21 11:57 AM, Florian Weimer via Libc-alpha wrote:
>> * Siddhesh Poyarekar:
>> 
>>> diff --git a/malloc/hooks.c b/malloc/hooks.c
>>> index 45c91d6502..4aa6dadcff 100644
>>> --- a/malloc/hooks.c
>>> +++ b/malloc/hooks.c
>>> @@ -20,6 +20,8 @@
>>>   #if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_34)
>>>   void weak_variable (*__after_morecore_hook) (void) = NULL;
>>>   compat_symbol (libc, __after_morecore_hook, __after_morecore_hook, GLIBC_2_0);
>>> +void *(*__morecore)(ptrdiff_t);
>>> +compat_symbol (libc, __morecore, __morecore, GLIBC_2_0);
>>>   #endif
>> This fails to assemble on ARC (and elsewhere with GCC 9 and
>> earlier):
>> /tmp/ccEXL6bz.s: Assembler messages:
>> /tmp/ccEXL6bz.s: Error: `__morecore@GLIBC_2.32' can't be versioned to common sym
>> bol '__morecore'
>> You need to add __attribute__ ((morecore)).
>
> I could do this (I know you mean nocommon there)...
>
>> I posted a generic fix for this:
>>    Force building with -fno-common
>>    <https://sourceware.org/pipermail/libc-alpha/2021-July/128716.html>
>> But until then, every instance needs to be fixed manually.
>
> ... or I could review your generic fix instead.  It seems like a much
> better cleanup than having to specify the attribute on everything that 
> changes.  What do you think?

Do we want this change for glibc 2.34 at this point?  I wasn't sure.

Carlos, I guess this is a question for you.

Thanks,
Florian