Andrew Dunstan writes:
> On 7/4/21 3:57 PM, Tom Lane wrote:
>> I'm now a little dubious about my claim that this would have helped find
>> any bugs. Invalidating a finished OpClassCache entry does not model any
>> real-world scenario, because as noted elsewhere in LookupOpclassInfo,
>> once such
On 7/4/21 3:57 PM, Tom Lane wrote:
> Over in [1] it is demonstrated that with CLOBBER_CACHE_ALWAYS enabled,
> initdb accounts for a full 50% of the runtime of "make check-world"
> (well, actually of the buildfarm cycle, which is not quite exactly
> that but close). Since initdb certainly doesn't
Over in [1] it is demonstrated that with CLOBBER_CACHE_ALWAYS enabled,
initdb accounts for a full 50% of the runtime of "make check-world"
(well, actually of the buildfarm cycle, which is not quite exactly
that but close). Since initdb certainly doesn't cost that much
normally, I wondered why it i
