Re: [RFC] Cleaning up die() error messages

"H. Peter Anvin" [off-list ref] writes:

Not unless any of those linked structures can be reached from outside 
the git library.

OK.  How about things like open file descriptors and mmap'ed
regions then?  That is:

    fd = open();
    xmalloc(); /* may die now */
    close(fd);

I think in the long run we would be better off if we properly
dealt with the case where (x)malloc returns NULL.  Of course,
that would make libification a bigger task, so we do not have to
do it from day one.