Andres Freund <andres@anarazel.de> writes:
> On 2020-03-28 23:50:32 -0400, Tom Lane wrote:
>> ... Curiously, its sibling
>> skate is *not* failing, despite being on the same machine and compiler.
> Hm. There's some difference in code-gen specific options.
Yeah, I confirmed that on my VM install too: using our typical codegen
options (just -O2 -g), the regression tests pass, matching skate's
results. It fell over after I matched snapper's CFLAGS, CPPFLAGS,
and LDFLAGS. I didn't try to break things down more finely as to
which option(s) trigger the bad code, since it looks like it's probably
some purely-internal compiler state ...
> What options were you using? Reproducing snapper as exactly as possible?
Yeah, see above.
> If you still have the environment it might make sense to check wether
> it's related to one of the other options. But otherwise I wouldn't be
> against the proposal.
I could, but it's mighty slow, so I don't especially want to try all 2^N
combinations. Do you have any specific cases in mind?
(I guess we can exclude LDFLAGS, since the assembly output is visibly
bad.)
regards, tom lane