Hi Junio,
On Tue, 23 Dec 2014, Junio C Hamano wrote:
> Johannes Schindelin <Johannes.Schindelin@xxxxxx> writes:
>
> > However, if we had to change the lookup such that it uses an array
> > always, we would have to introduce a function to initialize the
> > struct, always, in particular we would have to find a place to call
> > that initialization function in, say, builtin/fsck.c (actually, in
> > every code path that calls into the fsck machinery).
>
> You would need to call a function to "initialize" the table if you
> support customization by reading the configuration files anyway.
Yes, this is the config machinery. But I need employ that only if I want
to let the caller customize the severity levels. However, the fsck
machinery is also called from places where such a customization is not
offered. They would now need to be changed, too.
> Also I suspect that you can tell the compiler to initialize the
> array in place with default values, perhaps like this?
>
> -- >8 --
> #include <stdio.h>
>
> /* sorted by the default severity (lowest impact first) */
> #define EVENT_LIST(F) \
> F(EVENT_A), \
> F(EVENT_B), \
> F(EVENT_C), \
> F(EVENT_D)
>
> #define ID_(event) ID_ ## event
> enum event_id {
> EVENT_LIST(ID_)
> };
>
>
> enum severity_level {
> severity_info, severity_warn, severity_error
> };
>
> /* below this one are INFO */
> #define FIRST_WARN_EVENT_ID ID_EVENT_B
> /* below this one are WARN */
> #define FIRST_ERROR_EVENT_ID ID_EVENT_C
>
> #define STRING_(s) #s
> #define DESC_(event) \
> { \
> ID_ ## event, \
> STRING_(event), \
> (ID_ ## event < FIRST_WARN_EVENT_ID \
> ? severity_info \
> : ID_ ## event < FIRST_ERROR_EVENT_ID \
> ? severity_warn \
> : severity_error) \
> }
This is exactly the ugly, ugly preprocessor construct I thought you would
meet with contempt. I mean, compared to this, my FUNC() hack is outright
pretty ;-)
And *still*, this is *just* a global table with defaults. I would *still*
need to copy-on-write when the first customization of the severity level
takes place because I cannot allow the global defaults to be modified by
one caller (that would defeat the whole purpose of having per-caller
settings bundled in the fsck_options struct).
You see, I still would need to have a lazy initialization, the complexity
in that part would not be reduced at all.
So I am afraid that this approach really adds complexity rather than
replacing it with something simpler than my current code.
Ciao,
Dscho
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html