Thread: GETTIMEOFDAY_1ARG change

Hello together,

i would like to add the following changes to the code so that postgres
doesn�t have any problems if we compile it on machines that have
gettimeofday with 1 ARG and therefore don�t need 'struct timezone xxx'
I�m working on a System where timezone is defined in another way.

1. nabstime.c
Datum
timeofday(PG_FUNCTION_ARGS)
{
...
#ifndef GETTIMEOFDAY_1ARGstruct timezone tpz;
#endif
...
}

2.postgres.c
ResetUsage(void)
{
#ifndef GETTIMEOFDAY_1ARGstruct timezone tz;
#endif
getrusage(RUSAGE_SELF, &Save_r);gettimeofday(&Save_t, &tz);ResetBufferUsage();
/*      ResetTupleCount(); */
}


void
ShowUsage(void)
{
...
#ifndef GETTIMEOFDAY_1ARGstruct timezone tz;
#endif
...
}

3. postmaster.c
static int
ServerLoop(void)
{
...
#ifndef GETTIMEOFDAY_1ARGstruct timezone tz;
#endif
...
}

4. vacuum.c
void
vac_init_rusage(VacRUsage *ru0)
{
#ifndef GETTIMEOFDAY_1ARGstruct timezone tz;
#endif
getrusage(RUSAGE_SELF, &ru0->ru);gettimeofday(&ru0->tv, &tz);
}


Thanks

Ulrich Neumann

> i would like to add the following changes to the code so that postgres
> doesn´t have any problems if we compile it on machines that have
> gettimeofday with 1 ARG and therefore don´t need 'struct timezone xxx'
> I´m working on a System where timezone is defined in another way.

What system? How is timezone defined for that system? Is it something
completely new and different, or a variant which we already handle in
other places but not for this case?

...
> 2.postgres.c
> ResetUsage(void)
> {
> #ifndef GETTIMEOFDAY_1ARG
>         struct timezone tz;
> #endif
> 
>         getrusage(RUSAGE_SELF, &Save_r);
>         gettimeofday(&Save_t, &tz);
...

So what "one argument" does gettimeofday() have? Where does "tz" come
from if it is not defined here? Does it become a global variable? Where
is it declared?

afaik the nabstime.c usage of gettimeofday() has been in the PostgreSQL
code for quite a while, so I'm suprised that this is a problem on the
new mystery platform ;)
                   - Thomas

Thomas Lockhart <lockhart@fourpalms.org> writes:
> afaik the nabstime.c usage of gettimeofday() has been in the PostgreSQL
> code for quite a while, so I'm suprised that this is a problem on the
> new mystery platform ;)

I imagine he's merely unhappy about seeing "unused variable" warnings.

I'm unconvinced that's worth cleaning up, and certainly wouldn't hold
up 7.2 release for it ...
        regards, tom lane