Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On 2021-Aug-22, Tom Lane wrote:
>> Thomas Munro <thomas.munro@gmail.com> writes:
>>> After [1], seawasp blew up[2]. I tested the following fix on LLVM 13
>>> and 14 (main branch ~2 days ago). Better ideas welcome.
>>>
>>> - if (F.getAttributes().hasFnAttribute(llvm::Attribute::NoInline))
>>> +#if LLVM_VERSION_MAJOR < 14
>>> +#define hasFnAttr hasFnAttribute
>>> +#endif
>>> +
>>> + if (F.getAttributes().hasFnAttr(llvm::Attribute::NoInline))
>> Seems like either we should push back on pointless renaming, or else
>> that we're not really supposed to be accessing this non-stable API.
>> I have no idea which of those situations applies ... but if the LLVM
>> guys are randomly renaming methods that *are* supposed to be
>> user-visible, they need re-education.
> Did anything happen? Seawasp is still red ...
I stand by my opinion that Thomas' patch is a kluge rather than something
we should accept as a long-term answer. However, in the interests of
keeping the buildfarm green, maybe we should commit it until we have a
better solution. It looks like seawasp is only building HEAD, so I think
we could commit this in HEAD only.
regards, tom lane
