Path: utzoo!utgpu!water!watmath!clyde!rutgers!rochester!cornell!uw-beaver!uw-june!uw-entropy!dataio!pilchuck!ssc!kent
From: kent@ssc.UUCP (Kent)
Newsgroups: comp.sources.bugs
Subject: shar
Keywords: printf word size null pointer
Message-ID: <965@ssc.UUCP>
Date: 14 Jan 88 19:59:27 GMT
Organization: SSC, Inc., Seattle, WA
Lines: 21

This may have been brought up before...

I have found a native int size dependency in Gary Perlman's shar, 
which was posted a while ago. It prints FSIZE, which is of type 
off_t, which is usually a long.  In the printf()'s, the %d's 
corresponding to the FSIZEs must be changed to %ld or it won't work
on a 16 bit machine. 

Somebody elsewhere (in comp.lang.c, maybe) was annoyed about types 
like time_t, which pretend to improve portability via data hiding, but 
actually make a mess, since we need to know the underlying type 
anyway. 

This is an excellent example of the kind of problem these things
have in c.  The only safe way to handle these things (that I can think
of) is to use a flock of conversion macros, in stdjunk.h or whatever.

#define off_ttol(x) ((long)x)
#define ltooff_t(x) ((off_t)x)

etcetera...