Обсуждение: Couple of minor fixes

Attached match makes a couple of minor changes to get rid of some
annoying compiler warnings:

1) backend/port/win32/sema.c declares a variable inline with the code.
IIRC, that's only allowed in C++. Patch moves declaration to the top of
the function.

2) timezone/localtime.c, transtime() has a const arcument in the
implementation but non-const in declaration.

Definitly minor, but still annoying ;-)

I'm also seeing a lot of signed/unsigned mismatch, but IIRC it was said
at some point that we don't care about those.

//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
> Attached match makes a couple of minor changes to get rid of some
> annoying compiler warnings:

> 1) backend/port/win32/sema.c declares a variable inline with the code.
> IIRC, that's only allowed in C++. Patch moves declaration to the top of
> the function.

Done.

> 2) timezone/localtime.c, transtime() has a const arcument in the
> implementation but non-const in declaration.

I did this the other way (make implementation match declaration).
I think the coding was in fact correct per spec, because const-ness
of an argument value is not part of the function signature.

> I'm also seeing a lot of signed/unsigned mismatch, but IIRC it was said
> at some point that we don't care about those.

In HEAD?  They're all cleaned up according to the compilers I use.

            regards, tom lane

> > 2) timezone/localtime.c, transtime() has a const arcument in the
> > implementation but non-const in declaration.
>
> I did this the other way (make implementation match declaration).
> I think the coding was in fact correct per spec, because
> const-ness of an argument value is not part of the function signature.

Ok.


> > I'm also seeing a lot of signed/unsigned mismatch, but IIRC it was
> > said at some point that we don't care about those.
>
> In HEAD?  They're all cleaned up according to the compilers I use.

Yup.
More digging then I guess. Probably a #define that's messe dup
somewhere.

//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
>>> I'm also seeing a lot of signed/unsigned mismatch, but IIRC it was
>>> said at some point that we don't care about those.
>>
>> In HEAD?  They're all cleaned up according to the compilers I use.

> Yup.
> More digging then I guess. Probably a #define that's messe dup
> somewhere.

Maybe the issue is in Windows-specific code?  I have no idea what the
warning status is for that stuff.  In all the stuff I regularly compile,
there are only a couple of warnings and they are flex's fault (so
there's not much we can do about 'em).  I think it's good to get rid of
noise warnings as much as possible --- otherwise they can make you miss
important ones.

            regards, tom lane

> >>> I'm also seeing a lot of signed/unsigned mismatch, but
> IIRC it was
> >>> said at some point that we don't care about those.
> >>
> >> In HEAD?  They're all cleaned up according to the compilers I use.
>
> > Yup.
> > More digging then I guess. Probably a #define that's messe dup
> > somewhere.
>
> Maybe the issue is in Windows-specific code?  I have no idea
> what the warning status is for that stuff.  In all the stuff
> I regularly compile, there are only a couple of warnings and
> they are flex's fault (so there's not much we can do about
> 'em).  I think it's good to get rid of noise warnings as much
> as possible --- otherwise they can make you miss important ones.

Probably. I'll see what I can find. It could also be my build being
broken (I've hacked up the build system pretty bad right now), but I
think I used to see this before as well. Will get back to you on this
;-)

//Magnus

> > > 2) timezone/localtime.c, transtime() has a const arcument in the
> > > implementation but non-const in declaration.
> >
> > I did this the other way (make implementation match declaration).
> > I think the coding was in fact correct per spec, because
> const-ness of
> > an argument value is not part of the function signature.
>
> Ok.

Came across another one of those. schemacmds.h has:
extern void AlterSchemaOwner_oid(const Oid schemaOid, Oid newOwnerId);

but schemacmds.c has:
void
AlterSchemaOwner_oid(Oid oid, Oid newOwnerId)


//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
> Came across another one of those. schemacmds.h has:
> extern void AlterSchemaOwner_oid(const Oid schemaOid, Oid newOwnerId);
> but schemacmds.c has:
> void
> AlterSchemaOwner_oid(Oid oid, Oid newOwnerId)

Got it, thanks.

            regards, tom lane