On 02/23/2018 10:29 AM, Alexander Bokovoy wrote:
On pe, 23 helmi 2018, Jeff Law wrote:
> On 02/23/2018 09:58 AM, Randy Barlow wrote:
>> Greetings gcc maintainers!
>>
>> A FESCo issue[0] has been filed due to the dropping of 389-ds-base and
>> freeipa on 32 bit arches for Fedora 28. This was done without a change
>> request being filed, so FESCo is trying to decide how best to handle it.
>>
>> It seems there are some concerns about whether the C tooling correctly
>> handles some cases for 32-bit arches that led to the decision by the IPA
>> maintainers to drop support for 32-bit. FESCo would like to ask for the
>> GCC maintainers' input on the issue.
>>
>> Thanks in advance for your feedback!
> GCC, binutils, etc all support 32-bit arches just fine. THe current
> thinking is there are serious concurrency concerns in the freeipa code
> which are leading to the failures they were seeing on 32 bit platforms
> (i686 in particular).
To be clear there are concerns in 389-ds code that lead to data
corruptions on i686. They aren't seen on other 32-bit platforms Fedora
supports according to 389-ds developers.
Sorry to mis-characterize the problem.
It's the 389-ds code with the
concurrency concerns, not freeipa.
The fact that the 389-ds code works on other architectures does not
allow us to draw any conclusions at this point. It really needs to go
through a root cause analysis to determine exactly why it is not working
correctly. That will in turn tell us if it's the 389-ds code or
something in the compiler (or elsewhere) that is the problem.
Jeff