Re: Let's remove DSM_IMPL_NONE.

On 26.06.18 09:10, Kyotaro HORIGUCHI wrote:
> --- a/src/bin/initdb/initdb.c
> +++ b/src/bin/initdb/initdb.c
> @@ -984,6 +984,16 @@ test_config_settings(void)
>                  ok_buffers = 0;
>  
>  
> +    /*
> +     * Server doesn't confirm that the server-default DSM implementation is
> +     * actually workable. Choose a fine one for probing then it is used on the
> +     * new database.
> +     */
> +    printf(_("selecting dynamic shared memory implementation ... "));
> +    fflush(stdout);
> +    dynamic_shared_memory_type = choose_dsm_implementation();
> +    printf("%s\n", dynamic_shared_memory_type);
> +
>      printf(_("selecting default max_connections ... "));
>      fflush(stdout);
>  

I don't understand that comment.  initdb does test whether dsm=posix
works.  What more were you hoping for?

> @@ -996,10 +1006,11 @@ test_config_settings(void)
>                   "\"%s\" --boot -x0 %s "
>                   "-c max_connections=%d "
>                   "-c shared_buffers=%d "
> -                 "-c dynamic_shared_memory_type=none "
> +                 "-c dynamic_shared_memory_type=%s "
>                   "< \"%s\" > \"%s\" 2>&1",
>                   backend_exec, boot_options,
>                   test_conns, test_buffs,
> +                 dynamic_shared_memory_type,
>                   DEVNULL, DEVNULL);
>          status = system(cmd);
>          if (status == 0)

We could perhaps avoid some variability here by running the
bootstrapping runs in initdb using hardcoded dsm settings of
"sysv"/"windows".

> --- a/src/include/storage/dsm_impl.h
> +++ b/src/include/storage/dsm_impl.h
> @@ -14,11 +14,10 @@
>  #define DSM_IMPL_H
>  
>  /* Dynamic shared memory implementations. */
> -#define DSM_IMPL_NONE            0
> -#define DSM_IMPL_POSIX            1
> -#define DSM_IMPL_SYSV            2
> -#define DSM_IMPL_WINDOWS        3
> -#define DSM_IMPL_MMAP            4
> +#define DSM_IMPL_POSIX            0
> +#define DSM_IMPL_SYSV            1
> +#define DSM_IMPL_WINDOWS        2
> +#define DSM_IMPL_MMAP            3

I would avoid renumbering here.  It was kind of sensible to have NONE =
0, so I'd keep the non-NONE ones as non-zero.

-- 
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services