Pádraig Brady <P <at> draigBrady.com> writes: > __xstat64(3, "symlink", 0xbfee893c) = 0 > linkat(-100, 0xbfee8d3e, -100, 0xbfee8d46, 1024) = -1
> stat64("symlink", {st_mode=S_IFREG|0664, st_size=0, ...}) = 0 > connect(-100, {sa_family=0x7973 /* AF_??? */, sa_data="mlink\0hardlink"}, 4294967196) = -1 EINVAL Cute - those two lines match between the two straces, but the second trace doesn't know about the linkat syscall, and tries to interpret it as a connect syscall instead. But yep, definite proof that linkat(,AT_SYMLINK_FOLLOW) needs a rpl_linkat wrapper. -- Eric Blake