Front page | perl.perl5.porters |
Postings from July 2001
On Thu, 12 Jul 2001, Jarkko Hietaniemi wrote: > Well, at the very least they need to be declared, then. Otherwise the > atfork_lock and atfork_unlock are were not visible in miniperlmain.c. they are declared, but my patch put them in unixish.h, probably not the right place, perl.h would probably do. > > can you run 'make perl.i' and see if THX is really needed in the atfork > > functions? if it is, i don't see how this patch will work since the > > It works. yeah, it compiles, but THX will not be passed to Perl_atfork_* when pthread's fork() is called. i think just moving the Perl_atfork_* prototypes from unixish.h to a place where everything can see is the right fix. similar to how perl.h has: Malloc_t Perl_malloc (MEM_SIZE nbytes); Malloc_t Perl_calloc (MEM_SIZE elements, MEM_SIZE size); Malloc_t Perl_realloc (Malloc_t where, MEM_SIZE nbytes); without the THX/embed.pl magic. > > pthread_atfork() handlers will not get passed any args. i think they > > The function pointer typedef was admittedly a bit icky... but without > that part the Compaq C compiler get serious indigestion about mismatching > function signatures at pthread_atfork(). right, which keeping the original 'void Perl_fork_*(void)' prototype should cure with the ickyness.Thread Previous | Thread Next