Andrew Dunstan <andrew@dunslane.net> writes:
> On 05/19/2018 12:45 PM, Tom Lane wrote:
>> However ... we've got a ton of other places that use INT64_FORMAT with
>> the native printf, eg in pgbench, and frogmouth is not producing
>> warnings about those usages. So I'm confused about exactly what is
>> happening there. Andrew, do you have any insight?
> Very occasionally ;-)
> A little Googling suggested that __USE_MINGW_ANSI_STDIO might help.
> Here's what happened
> $ gcc -Wall -o testme testme.c
> testme.c: In function 'main':
> testme.c:7:5: warning: unknown conversion type character 'l' in format
> testme.c:7:5: warning: too many arguments for format
> $ ./testme
> sizeof long long = 8
> val = 734294471 (%lld) 3153770738837321131 (%I64d)
Yeah, this agrees with what we're seeing in the ecpg test failures, both
as to the warning and the wrong run-time answer. What remains unexplained
is why we don't see the same compile-time warning for uses of
printf("%lld") elsewhere in the build.
> $ gcc -Wall -D__USE_MINGW_ANSI_STDIO -o testme testme.c
[ works as expected ]
> So maybe we just need to define this on XP/mingw (shouldn't be necessary
> on anything later). I don't know what other effects it might have,
> though. Perhaps there is some other flag or define that has the same
> effect that we use in compiling pgbench etc that isn't used by ecpg?
Might be worth trying. As I mentioned in
<13103.1526749980@sss.pgh.pa.us>, it seems like it's time to jettison
any pretense of support for non-C99-compliant spellings of "%lld".
It'd be good to know whether __USE_MINGW_ANSI_STDIO works for that
purpose on ancient MinGW.
> Now recall that this animal is on serious life support. ...
> Not sure how much more effort I should put in here.
I couldn't blame you for just deciding to skip the ecpg tests on
this critter.
regards, tom lane