Bumping minimum GCC from 5.x to 6.x?

[erik.joelsson at oracle.com]

On 2021-09-01 10:01, Andrew Hughes wrote:
> On 15:17 Mon 30 Aug     , Magnus Ihse Bursie wrote:
>> Hi all,
>>
>> There is an open request[1] to bump the minimum GCC version from 5 to 6.
>> We've traditionally been very conservative with supporting old GCC
>> versions, but GCC 6 was released in 2016 (compared to GCC 5 in 2015),
>> and I think this sounds conservative enough.
>>
>> Would it create a lot of problems for anyone if we raised the minimum
>> version?
>>
>> /Magnus
>>
>> [1] https://bugs.openjdk.java.net/browse/JDK-8256977
>>
> Is this just for HEAD (i.e. OpenJDK 18 on)?  I don't think it's
> appropriate to change the requirements for older released versions.

The intention is definitely just to make this change in mainline (and 
then it will be down to each update maintainer if they want to backport 
it, but I see little value doing that myself).

/Erik

> I think what compilers are supported tends to be determined less by
> dictat and more by need.  If there are active maintainers building
> with GCC < 6, then they will report build failures and (ideally)
> fixes.  If no-one is actively building on a particular setup, it will
> often start to fail without any realising, through no real fault of
> anyone making changes.
>
> For example, we had a recent change in 8u that broke older systems (a
> C library rather than compiler issue in this case), and this was
> resolved by people reporting the failure. It wasn't obvious in a code
> review of the original patch, as it was to do with a macro being
> undefined in certain circumstances, so I can't see that it would have
> been caught unless someone was actively maintaining builds on those
> older platforms.
>
> Thanks,