On Sunday 02 May 2004 04:13 pm, Stas Bekman wrote:
> Really I was looking at some place where I can define my own map of
> code/string pairs and still reuse apr_strerror. I suppose the right way to
> go is to define a wrapper function:
>
> char * my_strerror(apr_status_t statcode, char *buf, apr_size_t bufsize)
> {
> if (statcode < APR_OS_START_USERERR) {
> return apr_strerror(statcode, buf, bufsize);
> }
> else {
> /* handle custom errors here */
> }
> }
>
> Is that the right way to go?
I can't speak for the architects of APR but this is what I ended up doing. It
doesn't feel like an optimal solution, really. What about people building on
your library, what do they do? But it is very similar to other libraries I
have used over the years in this respect.
My code ended up looking like:
/* In my include file... */
#define MY_OS_START_ERROR APR_OS_START_USERERR
#define MY_OS_START_USERERR (APR_OS_START_USERERR + 500)
/* In the related source file:
char *my_strerror (/* ... */)
{
if (statcode < MY_OS_START_ERROR) {
return apr_strerror(/* ... */);
} else if (statcode < MY_OS_START_USERERR) {
/* custom errors from my library */
} else if (statcode < APR_OS_START_CANONERR) {
/* don't know what this is, throw impossible error code string */
} else {
return apr_strerror(/* ... */);
}
}
It's been awhile since I wrote this code but I think the idea was to provide a
higher-level library than mine with a way to have it's own errors which would
start at MY_OS_START_USERERR just like mine started at APR_OS_START_USERERR.
mma
