Store only up to MAXPRCON constituents, not MAXCHRNV; code doesn't
fully support more than three anyway. Remove member p_nv, use item
type I_NONE for unused slots. Rename members p_vtype, p_vamt to
p_ctype, p_camt to avoid confusion with variable-style storage.
(pchr): Initializers adapted.
(nullify_objects, materials_cost, materials_charge, prod,
show_sect_capab, grind): Adapt, simplify where possible, protect
against bad item types in pchr[].
(MAXCHRNV): Unused, remove.
them. From Marc Olzheim.
Type modifier 'l' was missing in many places, probably rendering the
server useless on 64-bit hosts.
(ef_flush, ef_write, ef_extend, lwpCreate, lwpDestroy): Use conversion
specifier 'p' to print pointers.
(check_market): Fix display of loan amount.
(doland): Fix unescaped '%' (undefined behavior).
(ldump, ndump, pdump, sdump): Don't use flag '0' with conversion
specifier 's' (undefined behavior).
(dump, ldump, lost, ndump, pdump, sdump, empth_create, update_sched):
Cast time_t and pthread_t to long for printing.
(lwpStackCheck, lwpStackCheckUsed, finish_sects): Insert cast to fix
argument type on all platforms.
(prod): Remove extra argument.
(perform_mission, airdamage, retreat_land1, lwpReschedule): Format
string missed arguments.
thread entrypoints:
(lwpSelect, shutdown_sequence): Parameters didn't match thread entry
point prototype.
(lwpEntryPoint): Arguments didn't match thread entry point prototype.
Change linkage of functions without prototype declaration to static
where possible.
Remove some superflous declarations, replace others by suitable
includes.
versions of GNU libc, which can lead to nasty heap smashes (observed
with 2.2.93). Don't define RESOLVE_IPADDRESS for now, until we
readjust thread stack sizes.
effect. Replace calls by struct assignment where possible. Replace
clear buffer, copy string to buffer by strncpy(). Use assignment to
clear when that's clearer. Replace overlapping copy through bounce
buffer by memmove(). Replace rest by standard memset() and memcpy().
Also use sizeof() instead of literal array sizes for robustness, and
instead of symbolic array sizes for clarity.
windows. For obvious reasons, such warnings ought to be fixed.
Making compilation abort is the sledgehammer method to force that.
It's a stupid method for us, as we develop on various platforms, and
don't test every single commit under Windows. It gratuitously breaks
programs that work fine. Leave them as warnings, and fix them on
sight.
