Re: [PATCH 2/2] linux: Make profil_counter a compat_symbol (BZ#17726)

[Florian Weimer <fweimer at redhat dot com>]

* Adhemerval Zanella:

> On 02/08/2019 08:09, Florian Weimer wrote:
>> * Adhemerval Zanella:
>> 
>>> As indicated by Joseph's comment on BZ#17726, this symbol is most
>>> likely a historical ABI accident.  This patch make it on both arm
>>> and sparc ABIs a compat_symbol.
>>>
>>> Checked with a build arm-linux-gnueabihf, sparcv9-linux-gnu, adn
>>> sparc64-linux-gnu to see if the symbol is still present.
>>>
>>> 	* gmon/Versions (libc) [GLIBC_2.31]: New entry.
>>> 	* sysdeps/unix/sysv/linux/arm/profil-counter.h (profil_counter):
>>> 	Make a compat_symbol.
>>> 	* sysdeps/unix/sysv/linux/sparc/profil-counter.h
>>> 	(__profil_counter_global): Likewise.
>> 
>> There is this bug:
>> 
>>   <https://sourceware.org/bugzilla/show_bug.cgi?id=14043>
>> 
>> So it was already gone for over four years on arm until someone noticed.
>> I wonder if something really needs it?
>
> My take on this accidental ABI leakage is just to dropped it, it exposes
> unnecessary implementation detail, add complexity on arch-specific code
> base, it is dangerous because pollute application namespace with non
> expected symbol, and I would consider a potential bug if application are
> relying on this specific symbol semantic (and it is also non-portable).
>
> So I would vote if we just remove it instead of make it a compat symbol.

I tend to agree, but I wonder if Joseph filed this bug because it was
something that someone actually encountered as the result of a glibc
update.  In this case, it strongly points to accidental use of the
symbol.

I'm not necessarily opposed to actually removing symbols from the ABI in
general.  (I think some of us are though.)  I'm wondering if we have
evidence that we can't do it here.

Thanks,
Florian