[dpdk-stable] Why was "drivers/crypto: fix build with -fno-common" not picked for 19.11 stable back then?

[Kevin Traynor]

On 28/06/2021 10:13, Luca Boccassi wrote:
> On Mon, 2021-06-28 at 09:34 +0200, Christian Ehrhardt wrote:
>> Hi,
>> there is an old patch [1] which back then was not considered/taken
>> into 19.11 series.
>> This is now breaking Clang build [2].
>>
>> Since you were the Author/Maintainers back then I wanted to ask if
>> there was a reason to not pick it? If it should be picked now and if
>> there are any special things to consider for a backport?
>>
>> Chiming in/subscribing to the bug would also be appreciated.
>>
>> [1]: https://github.com/DPDK/dpdk/commit/eef9e0412a84cddf8944379ed4995314b4369370
>> [2]: https://bugs.dpdk.org/show_bug.cgi?id=733#c6
> 
> Can't remember the details about this particular one, but I think
> several fixes were backported for the -fno-common issue in 19.11
> 

I have some recollection is that as -fcommon was the default in gcc at
the time of 18.11/19.11 release, using the patch 50b03f3b8eaf
("drivers/crypto: disable gcc 10 no-common errors") to continue to keep
it building in the same way, may have been seen as more acceptable in
stable than using the same patch on main when gcc had a different default.

I had to send that patch as there was no fuller solution provided by the
crypto maintainers at the time. Then, late on it was provided and
applied to main.

So it might have been a mix of the workaround seen as good enough for
stable and the other patch coming in late, but I'm not 100% about it.