Re: monit&dietlibc

[Jan-Henrik Haukeland]

Christian Hopp <address@hidden> writes:

>> Still todo... (might be useful for other "normal" platforms)...

Yes

>> 1) replacement of gethostbyname against gethostbyname_r because...
>>    "warning: gethostbyname() leaks memory.  Use gethostbyname_r instead!"

It may leak memory if it is used in a multi-threaded environment, but
it does not leak memory per se! 

> Let forget about it... it's gonna get really ugly... for Linux!  The
> gethostbyname_r interface of Linux differs from all other system and
> it is not posix compatible.

Ehh? is the signature for gethostbyname_r in GLIBC different in
e.g. Linux glibc and Solaris glibc? 

BTW, the problem with the _r functions is that there is not really a
standard for them and I'm pretty sure that they are not even mentioned
in POSIX.

> ; /* this glibc "invention" is so ugly, I'm going to throw up any minute
> ;  * now */
> ; extern int gethostbyname_r(const char* NAME, struct hostent* RESULT_BUF,
> ;                      char* BUF, size_t BUFLEN, struct hostent** RESULT,
> ;                      int* H_ERRNOP) __THROW;

The implementation is not *that* different from Solaris
gethostbyname_r (and the one mentioned by Stevens in UNP):

     struct hostent *gethostbyname_r(const char *name,
          struct hostent *result, char *buffer, int buflen,
          int *h_errnop);

and it looks easy to implement a wrapper that support both
versions. For instance (nb! untested):

 struct hostent *gethostbyname_RRRR(const char *name,
        struct hostent *result, char *buffer, int buflen,
        int *h_errnop) {

           #ifdef glibc
             struct hostent *r;
             gethostbyname(name, result, buffer, buflen, &r, h_errnop);
             return r;
           #else 
             /* solaris */
             return gethostbyname_r(name, result, buffer, buflen, h_errnop);
           #endif
 }




-- 
Jan-Henrik Haukeland