Ruslan Ermilov [Thu, 16 May 2002 15:22:58 +0000 (15:22 +0000)]
MD_EXEC_PREFIX doesn't work for the cross-arch compiler.
The change also makes the `cc -print-search-dirs' output
sane (the pre-3.1 way) in the non-cross case.
Ruslan Ermilov [Thu, 16 May 2002 15:18:13 +0000 (15:18 +0000)]
Make it possible to build a cross compiler for alpha,
ia64 and sparc64 on systems that do not have atoll(3).
The "cross" here doesn't necessarily mean cross-arch.
Brian Somers [Thu, 16 May 2002 13:34:20 +0000 (13:34 +0000)]
Handle MS-CHAPv2 authentication correctly via the RADIUS server (if it's
configured).
Handle internal failures in radius_Authenticate() correctly.
Bump the ppp version number.
This doesn't yet work with MPPE. More will follow.
Maxim Sobolev [Thu, 16 May 2002 10:57:10 +0000 (10:57 +0000)]
Rename struct scr_size into struct _scr_size and struct scrmap into
struct _scrmap, so that it doesn't break C++ programs (name of element of
the structure is the same as the name of the scructure itself).
Warner Losh [Thu, 16 May 2002 06:32:37 +0000 (06:32 +0000)]
Use splnet() when we need to block timeouts rather than splimp() since
the former blocks software interrupts, while the latter blocks
hardware interrupts.
Avoid one place where I'm at splnet across a call to copyout. Leave
one in place to give bde something to complain about :-). Actaully,
I'll fix it in a subsequent commit.
Mike Silbersack [Wed, 15 May 2002 17:05:28 +0000 (17:05 +0000)]
Enhance the use of the watchdog timer in this driver so that it will
allow recovery from transmission lockups which occur in the middle
of the descriptor list, rather than just at the beginning.
For some unknown reason, Rhine II chips have a tendency to stop
transmitting while under heavy load, possibly due to collisions.
Whether this behavior is due to a hardware bug or a driver glitch
is unknown as of now.
In either case, this change allows the driver to gracefully recover
from such situations.
Special thanks go to The Anarcat <anarcat@anarcat.dyndns.org>, who
bugged me into looking at this and to
Dominic Marks <dominic_marks@btinternet.com>, who performed a great
deal of testing to help characterize this problem.
Tony Finch [Wed, 15 May 2002 16:30:28 +0000 (16:30 +0000)]
Upgrade unifdef:
* It now knows about the existence of #elif which would have
caused it to produce incorrect results in some situations.
* It can now process #if and #elif lines according to the
values of symbols that are specified on the command line.
The expression parser is only a simple subset of what C
allows but it should be sufficient for most real-world
code (it can cope with everything it finds in xterm).
* It has an option for printing all of the symbols that might
control #if processing. The unifdefall script uses this
option along with cpp -dM to strip all #ifs from a file.
* It has much larger static limits.
* It handles nested #ifs much more completely.
There have also been many style improvements: KNF; ANSI function
definitions; all global stuff moved to the top of the file; use
stdbool instead of h0h0bool; const-correctness; err(3) instead
of fprintf(stderr, ...); enum instead of #define; commentary.
I used NetBSD's unifdef as the basis of this since it has received
the most attention over the years.
Ruslan Ermilov [Wed, 15 May 2002 16:19:54 +0000 (16:19 +0000)]
Rename `includes' to `buildincludes'.
Rename `incsinstall' to `installincludes'.
Make `includes' a -j safe shortcut for `buildincludes' + `installincludes'.
`buildincludes' and `installincludes' are SUBDIR friendly, if run directly.
Juli Mallett [Wed, 15 May 2002 09:49:59 +0000 (09:49 +0000)]
Make 'user' and 'group' const as to not discard the qualifier from the
functions we use to assign them. Doesn't seem to be anything else that
relies on these being non-const.
Tim J. Robbins [Wed, 15 May 2002 03:17:01 +0000 (03:17 +0000)]
Use getopt() to reject options (asa has none), and to the "--" end of options
marker. Exit non-zero if we cannot open one of the input files. Update
standards conformance and exit status statements in manual page.
Kelly Yancey [Wed, 15 May 2002 01:03:32 +0000 (01:03 +0000)]
Reset token-ring source routing control field on receipt of ethernet frame
without source routing information. This restores the behaviour in this
scenario to that of prior to my last commit.
Ian Dowse [Tue, 14 May 2002 23:24:28 +0000 (23:24 +0000)]
Use fgetln to remove the static limit on the length of lines in
/etc/exports. Oversized lines were unlikely due to the large 10k
limit, but any found would cause mountd to exit with an error. Also
fix one or two compiler warnings.
Brian Feldman [Tue, 14 May 2002 23:15:42 +0000 (23:15 +0000)]
o Clean up tmp file usage a little by using mkstemp(3) instead of
mktemp(3). It would be amazingly unlikely, but the former method
could result in a symlink attack. A better solution would use
${TMPDIR}, though.
o Make sed not overwrite old backup files with no warning.
Robert Watson [Tue, 14 May 2002 23:07:15 +0000 (23:07 +0000)]
p_cansignal() returns an errno value; at some point, the check for
inter-process signalling ceased to preserve and return that value,
instead always returning EPERM. This meant that it was possible
to "probe" the pid space for processes that were not otherwise
visible. This change reverts that reversion.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs
Jeff Roberson [Tue, 14 May 2002 21:59:49 +0000 (21:59 +0000)]
Disable the shared locking namei() code for now. It breaks several stacking
filesystems. This is on hold until the rest of VFS Locking is reviewed and
deemed safe. It can be enabled with 'options LOOKUP_SHARED'.
Maxim Sobolev [Tue, 14 May 2002 21:42:37 +0000 (21:42 +0000)]
- Make use of DEPOROGINs (if there are any) when installing package;
- fix few bogosities here and there;
- move some common routines into the library.
Joerg Wunsch [Tue, 14 May 2002 21:28:45 +0000 (21:28 +0000)]
After some comments from bde, rewrite the loops to avoid turning the
previously used "micro-optimization" (count-down loop) into a
pessimization. Now the loops are written in the more natural count-up
form.
Also, while being there, i made the logic in out_fdc() similar to the
logic in in_fdc(). The old implementation was a bit bogus anyway
since it first tested the DIO bit and only afterwards the RQM bit.
However, according to the description of the i82077, the DIO bit is
only guaranteed to be valid once the RQM bit is set. Thus, the old
implementatoin would have had the chance to misbehave on a controller
that is implemented in accordance with the i82077 description (but is
not bug-for-bug compatible).