Tom Lane wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
>
>> Maybe we need to abandon trying to map float8 results exactly in the
>> resultmap file, and just let pg_regress pick the best fit as we do with
>> some other tests.
>>
>
> I thought about that too but it seems a very bad idea. small-is-zero is
> distinctly "less correct" than the regular output, and I don't think we
> want pg_regress to be blindly accepting it as OK on any platform.
>
> Perhaps we could stick a version check into the resultmap lookup? It'd
> likely have been painful on the shell script implementation but now that
> the code is in C I think we have lots of flexibility. There's no need
> to feel bound by the historical resultmap format.
>
> However this is all premature unless we can verify that "cgywin's strtod()
> complains about float underflow after version so-and-so". Do they
> publish a detailed change log?
>
>
>
Yes, good points. One other thought I had was that we could have
pg_regress always allow a fallback to the canonical result file. So in
this case it would try float8-small-is-zero.out and
float8-small-is-zero_1.out and finally float8.out (the first would
succeed on eel, the last on cassowary, I am speculating). In general
this would allow a configuration to become more correct painlessly.
cheers
andrew