Re: pmServiceDiscoveryInterrupt() commit a8b87e2 et al.

[fche@xxxxxxxxxx (Frank Ch. Eigler)]

Dave Brolley <brolley@xxxxxxxxxx> writes:

> [...]
>> typedef struct {
>>      int         version;
>>      int         interrupted;      /* could become a flags bitfield? */
>>      const char  *globalOptions;   /* and this could be more explicit, */
>>                                /* instead of every possibly option being */
>>                                /* coerced into one comma-separated string? */
>> } pmDiscoveryOptions;

> [...] I ended up going with this approach. [...]

Oh well.  We should leave a note in the docs & code with warning signs
about the complete loss of compiler type checking inherent in this
approach, and (in case it can help) make a habit of valgrind-running
every app that uses this API.

There is no point using bitfields for the resolve flag, and reserving
space for future bits, as the top version number will have to serve as
sufficient discriminator between all future pmDiscoveryOption struct
variants.

All places in libpcp that now accept a discoveryOptions struct must at
minimum assert that ->version == 1, lest an app built against a future
incompatible version is run against this one, leading to a crash.

The man/man3/pmdiscoverservices.3 needs an update & documentation for
this new struct (including initialization requirements);
man3/pmservicediscoveryinterrupt.3 should be removed.

- FChE