This patch fixes many warnings from the beginning of the build up to
and including the depend stage. Nearly all warnings should be gone
even with -Wall.
In init() there was code iterating over all of the possible file
descriptors in a svc_fdset. fdsets are limited to FD_SETSIZE. This
caused coredumps on FreeBSD 10, and possibly other hidden issues.
Moving to poll(), rather than select() would be better, but is a bigger
job. For now, just limit to the FD_SETSIZE that select() requires.
Currently, mp_rpc_server.C tries 538 million ports to acquire an
available transient rpcbind port number. This is bad when rpcbind is
running in secure mode (and you are not using tirpc) - Xsession will
'hang' at the dthello (blue) screen filling up your error logs with
RPC errors.
Now, just try +- 50 (for a total of 100 ports) before bailing. The
dthello 'blue screen of death' is the most common problem in starting
CDE when rpcbind isn't set up properly. This should at least not
cause the appearance of a 'hang'.
This reverts commit 44e384aedb.
This code is actually needed. If svcfd_create() is not available, it
should be fixed only for those systems that it affects.
__rpc_xdr is no longer available on FreeBSD 10.
(XDR is typedef'd as "struct XDR" and not "struct __rpc_xdr").
By the way, why did we ever need this? Probably
it should be removed. Leaving for __OpenBSD__ for now.
This code tried to automatically generate the X DISPLAY
from the combination of the hostname and display number;
however 127.0.0.1:0 is normally rejected by X11, so this
technique is no good. Fixes dticon hang on startup, caused by
XOpenDisplay failure leading to this message from tttrace:
tt_default_session_set(0x0x875190=="X 127.0.0.1 0") = 1032 (TT_ERR_ACCESS)
- Const strings referenced by non-const variables.
- Incorrect format specifers for printing addresses
- Unused variables
- Signed comparison to unsigned
Also fix an incorrect enumeration value in a switch statement.
This is a non-POSIX/ISO-C header. It is ok to include this on Linux, but it
is obsolete on BSD; FreeBSD even throws an error if you include it with
__STDC__ defined. Every system should nowadays have malloc() defined in
stdlib.h.
Diff is largely mechanical, replacing malloc.h with stdlib.h where it is not
yet included anyway.
In part of the tooltalk rpc code (mp_message.c), it was assumed that on
the majority of platforms, sizeof(uid_t)=sizeof(gid_t)=sizeof(long). On
Linux-x64, uid_t is an unsigned int, which makes the code fail: all
tooltalk messages fail to send with an RPC_CANTENCODEARGS at the
rpc-level, and TT_INTERNAL_ERR for the actual program. We instead
change the code to explicitly examine sizeof(uid_t) to see whether it is
int or long sized. This allows tooltalk-dependent functinoality
like logout and multiple calls to dtfile to work.
Enums may be represented with a smaller type than int; however, they are
automatically promoted to int when passed in va_arg lists, just as
short, char, etc. are. GCC thus "knows" that you never want to call
va_arg with an enum type, and instead inserts an abort.
Peter Howkins