On Wed, Mar 17, 1999 at 10:29:01AM +0000, Nicholas Clark wrote: > I now have a functional 2.2.2 ELF system, which successfully compiled glibc2 > yesterday without any intervention. > > I've just compiled the latest test release of perl and it fails 4 > regression tests for various reasons. op/groups.t caught my eye. > > op/groups.t compares the output from the groups command with getgroups(). > It is failing because perl expects the returned array of GIDs to be > 32 bit integers, whereas they are actually 16 bit integers. > glibc's bits/types.h has __gid_t as __u_int, whereas the kernel header > asm/posix_types.h has "typedef unsigned short __kernel_gid_t" > > Is glibc2.1 at fault here? The philosophy behind glibc is to pull the interface into the library and away from the kernel, so the kernel can be changed without affecting applications. For this reason, you're not supposed to mix kernel and libc headers. I happen to know that gid_t is supposed to be expanding at some stage, so we really should be seeing 32 bit gid_t in userland. If we're not, then that's a bug in glibc, IMO. -- Matthew Wilcox <[EMAIL PROTECTED]> "I decry the current tendency to seek patents on algorithms. There are better ways to earn a living than to prevent other people from making use of one's contributions to computer science." -- Donald E. Knuth, TAoCP vol 3 unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
