]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoThis commit was generated by cvs2svn to compensate for changes in r98005,
Darren Reed [Fri, 7 Jun 2002 08:58:22 +0000 (08:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r98005,
which included commits to RCS files with non-trunk default branches.

22 years agoImport IPFilter 3.4.28
Darren Reed [Fri, 7 Jun 2002 08:58:22 +0000 (08:58 +0000)]
Import IPFilter 3.4.28

22 years agoCommit changes that happened in IPFilter versions 3.4.27 - 3.4.28
Darren Reed [Fri, 7 Jun 2002 08:56:30 +0000 (08:56 +0000)]
Commit changes that happened in IPFilter versions 3.4.27 - 3.4.28

22 years agoContrary to what the comments said, rs -H did not recycle storage, it just
Tim J. Robbins [Fri, 7 Jun 2002 07:08:06 +0000 (07:08 +0000)]
Contrary to what the comments said, rs -H did not recycle storage, it just
kept writing past the end of the buffer. Correct the code so that it
actually does recycle storage.

22 years agoDisabling automatic --same-owner option when running as uid 0 along with
Maxim Sobolev [Fri, 7 Jun 2002 06:02:35 +0000 (06:02 +0000)]
Disabling automatic --same-owner option when running as uid 0 along with
the --same-permissions was an overkill, so put it back. This is consistent
with what our old tar did.

Suggested by: dillon

22 years ago- Fixup / remove obsolete comments.
John Baldwin [Fri, 7 Jun 2002 05:47:35 +0000 (05:47 +0000)]
- Fixup / remove obsolete comments.
- ktrace no longer requires Giant so do ktrace syscall events before and
  after acquiring and releasing Giant, respectively.
- For i386, ia32 syscalls on ia64, powerpc, and sparc64, get rid of the
  goto bad hack and instead use the model on ia64 and alpha were we
  skip the actual syscall invocation if error != 0.  This fixes a bug
  where if we the copyin() of the arguments failed for a syscall that
  was not marked MP safe, we would try to release Giant when we had
  not acquired it.

22 years agoWe no longer need to acqure Giant in ast() for ktrpsig() in postsig() now
John Baldwin [Fri, 7 Jun 2002 05:43:40 +0000 (05:43 +0000)]
We no longer need to acqure Giant in ast() for ktrpsig() in postsig() now
that ktrace no longer needs Giant.

22 years ago- trapsignal() no longer needs to acquire Giant for ktrpsig().
John Baldwin [Fri, 7 Jun 2002 05:43:02 +0000 (05:43 +0000)]
- trapsignal() no longer needs to acquire Giant for ktrpsig().
- Catch up to new ktrace API.

22 years ago- Proper locking for p_tracep and p_traceflag.
John Baldwin [Fri, 7 Jun 2002 05:42:25 +0000 (05:42 +0000)]
- Proper locking for p_tracep and p_traceflag.
- Catch up to new ktrace API.

22 years agoProperly lock accesses to p_tracep and p_traceflag. Also make a few
John Baldwin [Fri, 7 Jun 2002 05:41:27 +0000 (05:41 +0000)]
Properly lock accesses to p_tracep and p_traceflag.  Also make a few
ktrace-only things #ifdef KTRACE that were not before.

22 years agoDescribe ia64_unaligned and geom directories. I hope I got these right.
Juli Mallett [Fri, 7 Jun 2002 05:41:15 +0000 (05:41 +0000)]
Describe ia64_unaligned and geom directories.  I hope I got these right.

22 years ago- Catch up to new ktrace API.
John Baldwin [Fri, 7 Jun 2002 05:39:16 +0000 (05:39 +0000)]
- Catch up to new ktrace API.
- ktrace trace points in msleep() and cv_wait() no longer need Giant.

22 years agoCatch up to changes in ktrace API.
John Baldwin [Fri, 7 Jun 2002 05:37:18 +0000 (05:37 +0000)]
Catch up to changes in ktrace API.

22 years agoOverhaul the ktrace subsystem a bit. For the most part, the actual vnode
John Baldwin [Fri, 7 Jun 2002 05:32:59 +0000 (05:32 +0000)]
Overhaul the ktrace subsystem a bit.  For the most part, the actual vnode
operations to dump a ktrace event out to an output file are now handled
asychronously by a ktrace worker thread.  This enables most ktrace events
to not need Giant once p_tracep and p_traceflag are suitably protected by
the new ktrace_lock.

There is a single todo list of pending ktrace requests.  The various
ktrace tracepoints allocate a ktrace request object and tack it onto the
end of the queue.  The ktrace kernel thread grabs requests off the head of
the queue and processes them using the trace vnode and credentials of the
thread triggering the event.

Since we cannot assume that the user memory referenced when doing a
ktrgenio() will be valid and since we can't access it from the ktrace
worker thread without a bit of hassle anyways, ktrgenio() requests are
still handled synchronously.  However, in order to ensure that the requests
from a given thread still maintain relative order to one another, when a
synchronous ktrace event (such as a genio event) is triggered, we still put
the request object on the todo list to synchronize with the worker thread.
The original thread blocks atomically with putting the item on the queue.
When the worker thread comes across an asynchronous request, it wakes up
the original thread and then blocks to ensure it doesn't manage to write a
later event before the original thread has a chance to write out the
synchronous event.  When the original thread wakes up, it writes out the
synchronous using its own context and then finally wakes the worker thread
back up.  Yuck.  The sychronous events aren't pretty but they do work.

Since ktrace events can be triggered in fairly low-level areas (msleep()
and cv_wait() for example) the ktrace code is designed to use very few
locks when posting an event (currently just the ktrace_mtx lock and the
vnode interlock to bump the refcoun on the trace vnode).  This also means
that we can't allocate a ktrace request object when an event is triggered.
Instead, ktrace request objects are allocated from a pre-allocated pool
and returned to the pool after a request is serviced.

The size of this pool defaults to 100 objects, which is about 13k on an
i386 kernel.  The size of the pool can be adjusted at compile time via the
KTRACE_REQUEST_POOL kernel option, at boot time via the
kern.ktrace_request_pool loader tunable, or at runtime via the
kern.ktrace_request_pool sysctl.

If the pool of request objects is exhausted, then a warning message is
printed to the console.  The message is rate-limited in that it is only
printed once until the size of the pool is adjusted via the sysctl.

I have tested all kernel traces but have not tested user traces submitted
by utrace(2), though they should work fine in theory.

Since a ktrace request has several properties (content of event, trace
vnode, details of originating process, credentials for I/O, etc.), I chose
to drop the first argument to the various ktrfoo() functions.  Currently
the functions just assume the event is posted from curthread.  If there is
a great desire to do so, I suppose I could instead put back the first
argument but this time make it a thread pointer instead of a vnode pointer.

Also, KTRPOINT() now takes a thread as its first argument instead of a
process.  This is because the check for a recursive ktrace event is now
per-thread instead of process-wide.

Tested on: i386
Compiles on: sparc64, alpha

22 years agoAdd the m4(1) regression test.
Juli Mallett [Fri, 7 Jun 2002 05:32:00 +0000 (05:32 +0000)]
Add the m4(1) regression test.

22 years agoAdd a regression test for m4(1).
Juli Mallett [Fri, 7 Jun 2002 05:31:32 +0000 (05:31 +0000)]
Add a regression test for m4(1).

This checks the behaviour of changecom() in both GNU and BSD modes of our m4.

22 years agoUse a larger data type to prevent counters wrapping so quickly.
Matthew N. Dodd [Fri, 7 Jun 2002 05:29:13 +0000 (05:29 +0000)]
Use a larger data type to prevent counters wrapping so quickly.
Silence a warning.

22 years agoSync this up with changes made in v1.8 of sys/netatm/atm_if.h
Matthew N. Dodd [Fri, 7 Jun 2002 05:25:10 +0000 (05:25 +0000)]
Sync this up with changes made in v1.8 of sys/netatm/atm_if.h

22 years agoRecognize Adaptec ANA-5910/30/40[A] boards.
Matthew N. Dodd [Fri, 7 Jun 2002 05:23:01 +0000 (05:23 +0000)]
Recognize Adaptec ANA-5910/30/40[A] boards.
Read the MAC address from Adaptec boards correctly.

Bits borrowed from sys/pci/if_en_pci.c.

22 years agoAdd a new SYSINIT subsystem for KTRACE.
John Baldwin [Fri, 7 Jun 2002 05:11:39 +0000 (05:11 +0000)]
Add a new SYSINIT subsystem for KTRACE.

22 years ago- Add a per-thread member 'td_inktrace' to be used by ktrace to detect
John Baldwin [Fri, 7 Jun 2002 05:11:08 +0000 (05:11 +0000)]
- Add a per-thread member 'td_inktrace' to be used by ktrace to detect
  when a thread is in the ktrace subsystem to avoid ktrace'ing internal
  ktrace events.
- Update the locking notes for p_traceflag and p_tracep taking into account
  the new ktrace_lock mutex.

22 years agoNo need to check for "-" explicitly in getopt() loop.
Tim J. Robbins [Fri, 7 Jun 2002 02:35:21 +0000 (02:35 +0000)]
No need to check for "-" explicitly in getopt() loop.

22 years agoExit >1 when an error occurs, regardless of whether -s is specified (SUSv3)
Tim J. Robbins [Fri, 7 Jun 2002 02:30:56 +0000 (02:30 +0000)]
Exit >1 when an error occurs, regardless of whether -s is specified (SUSv3)

22 years ago'device hea' is no longer broken.
Matthew N. Dodd [Fri, 7 Jun 2002 02:04:09 +0000 (02:04 +0000)]
'device hea' is no longer broken.
Add 'nowerror' to a few 'hea' files to ignore warnings on volatiles.

22 years agoMove some code around.
Matthew N. Dodd [Fri, 7 Jun 2002 01:55:42 +0000 (01:55 +0000)]
Move some code around.
Minor whitespace changes.

22 years agoANSI style function declarations.
Juli Mallett [Fri, 7 Jun 2002 01:41:54 +0000 (01:41 +0000)]
ANSI style function declarations.

22 years agoExpand description of environment variables, refer to environ(7) where
Tim J. Robbins [Fri, 7 Jun 2002 01:29:49 +0000 (01:29 +0000)]
Expand description of environment variables, refer to environ(7) where
appropriate.

22 years agoStyle nit: group together optional flags under the one Fl macro.
Tim J. Robbins [Fri, 7 Jun 2002 01:17:54 +0000 (01:17 +0000)]
Style nit: group together optional flags under the one Fl macro.

22 years agoBack out previous back out of previously correct code.
Juli Mallett [Fri, 7 Jun 2002 01:14:05 +0000 (01:14 +0000)]
Back out previous back out of previously correct code.

Double-plus-pointy-hat to: jmallett
Submitted by: bde

22 years agoRespect LC_CTYPE and LC_COLLATE environment variables, mainly for their
Tim J. Robbins [Fri, 7 Jun 2002 01:04:24 +0000 (01:04 +0000)]
Respect LC_CTYPE and LC_COLLATE environment variables, mainly for their
effects on regular expressions. The libc regex code does not handle
equivalence classes well (for example), but we do our best.

22 years agoUse size_t or ino_t instead of int, and remove a few warnings.
Giorgos Keramidas [Fri, 7 Jun 2002 01:01:08 +0000 (01:01 +0000)]
Use size_t or ino_t instead of int, and remove a few warnings.
The makewhatis utility is now ok with WARNS=3.

Reviewed by: bde

22 years agoRefer to environ(7) for description of COLUMNS. Note the effects of LANG,
Tim J. Robbins [Fri, 7 Jun 2002 00:50:22 +0000 (00:50 +0000)]
Refer to environ(7) for description of COLUMNS. Note the effects of LANG,
LC_ALL and LC_TIME upon the utility.

22 years agoDocument the standard COLUMNS environment variable here to avoid redundant
Tim J. Robbins [Fri, 7 Jun 2002 00:45:06 +0000 (00:45 +0000)]
Document the standard COLUMNS environment variable here to avoid redundant
descriptions of its effects in utilities' manual pages.

22 years agoIMO it was a quite ugly idea that if we are running as uid 0 then we can
Maxim Sobolev [Fri, 7 Jun 2002 00:03:23 +0000 (00:03 +0000)]
IMO it was a quite ugly idea that if we are running as uid 0 then we can
safely ignore current umask(2) and assume that permissions should be set
right like in the archive. Not only it violates POLA, but introduces
huge potential security vulnerability, particularly for ports, where
many popular archives come with 777 files and dirs.

22 years agoRemove tunables to hide warnings that no longer exist.
Juli Mallett [Fri, 7 Jun 2002 00:01:00 +0000 (00:01 +0000)]
Remove tunables to hide warnings that no longer exist.

22 years agoCast to long to match format. Hidden by revision 1.18 of Makefile.
Juli Mallett [Fri, 7 Jun 2002 00:00:29 +0000 (00:00 +0000)]
Cast to long to match format.  Hidden by revision 1.18 of Makefile.

22 years agodebug.{numvnodes,freevnodes} moved to vfs.
Dag-Erling Smørgrav [Thu, 6 Jun 2002 23:01:50 +0000 (23:01 +0000)]
debug.{numvnodes,freevnodes} moved to vfs.

22 years agoWork around a bug in the Linux version of ski, that's specific to
Marcel Moolenaar [Thu, 6 Jun 2002 22:49:23 +0000 (22:49 +0000)]
Work around a bug in the Linux version of ski, that's specific to
SSC_GET_RTC. This fixes the panic seen shortly after mounting the
root file system.

Thanks to: "K.Sumitani" <ksumitani@mui.biglobe.ne.jp>

22 years agoClarify the bit about realloc() and its `ptr' argument a bit.
Giorgos Keramidas [Thu, 6 Jun 2002 22:11:19 +0000 (22:11 +0000)]
Clarify the bit about realloc() and its `ptr' argument a bit.
Hopefully, now it is more clear that the memory referenced by the
ptr argument of realloc(ptr,size) is freed and only the return value
of realloc() points to a valid memory area upon successful completion.

Submitted by: Martin Faxer <gmh003532@brfmasthugget.se>

22 years agoUse a global `now' variable for the current time, and initialise it at
Juli Mallett [Thu, 6 Jun 2002 21:21:25 +0000 (21:21 +0000)]
Use a global `now' variable for the current time, and initialise it at
startup, right after calling setlocale(3).

22 years agoImplement a SUSv3-ignorant but "time"-similar format for "etime", elapsed
Juli Mallett [Thu, 6 Jun 2002 21:18:37 +0000 (21:18 +0000)]
Implement a SUSv3-ignorant but "time"-similar format for "etime", elapsed
run time (NOT cpu time).  cputime() and elapsed() both need to honour SUSv3
now.

22 years agosysctlbyname(3) returns -1 on failure, not any other value < 0.
Juli Mallett [Thu, 6 Jun 2002 20:59:15 +0000 (20:59 +0000)]
sysctlbyname(3) returns -1 on failure, not any other value < 0.

22 years agoChange the all locks list from a STAILQ to a TAILQ. This bloats struct
John Baldwin [Thu, 6 Jun 2002 20:51:04 +0000 (20:51 +0000)]
Change the all locks list from a STAILQ to a TAILQ.  This bloats struct
lock_object by another pointer (though all of lock_object should be
conditional on LOCK_DEBUG anyways) in exchange for an O(1) TAILQ_REMOVE()
in witness_destroy() (called for every mtx_destroy() and sx_destroy())
instead of an O(n) STAILQ_REMOVE.  Since WITNESS is so dog slow as it is,
the speed-up is worth the space cost.

Suggested by: iedowse

22 years agoFix a typo in my recently added comment: s/beleived/believed/
Semen Ustimenko [Thu, 6 Jun 2002 20:43:03 +0000 (20:43 +0000)]
Fix a typo in my recently added comment: s/beleived/believed/

Submitted by: keramida

22 years agoSupport the SUSv3 `rgroup' format.
Juli Mallett [Thu, 6 Jun 2002 20:42:16 +0000 (20:42 +0000)]
Support the SUSv3 `rgroup' format.

Clean up some local style bogons.

22 years ago* Use the new stat(1) to get mode information from the files we
Doug Barton [Thu, 6 Jun 2002 20:38:22 +0000 (20:38 +0000)]
* Use the new stat(1) to get mode information from the files we
are installing.
* Since this means that for now we can't accomodate non-standard
umask's, warn the user accordingly.
* Convert the "press enter to continue" prompt into a function.

22 years agoCast arg_max to size_t when comparing it (times 4, plus 1) against SIZE_MAX. I
Juli Mallett [Thu, 6 Jun 2002 20:29:39 +0000 (20:29 +0000)]
Cast arg_max to size_t when comparing it (times 4, plus 1) against SIZE_MAX.  I
was worried about truncation of arg_max by this cast, but if it gets truncated,
we know it'll obviously be greater than SIZE_MAX anyway.

Big pointy hat to: jmallett
Submitted by: keramida

22 years agoSUSv3 conform on the "comm" and "args" formats, and make correct the "command"
Juli Mallett [Thu, 6 Jun 2002 20:21:47 +0000 (20:21 +0000)]
SUSv3 conform on the "comm" and "args" formats, and make correct the "command"
format, since it's BSDlike, and "comm" is actually different.

22 years agoRemove the old nextboot from Whistler. AFAIK this hasn't been useable in
Gordon Tetlow [Thu, 6 Jun 2002 20:20:58 +0000 (20:20 +0000)]
Remove the old nextboot from Whistler. AFAIK this hasn't been useable in
quite a while and only works on i386.

22 years agos/daddr_t/ufs_daddr_t/g
Poul-Henning Kamp [Thu, 6 Jun 2002 19:59:46 +0000 (19:59 +0000)]
s/daddr_t/ufs_daddr_t/g

This should fix the issues which cropped up after daddr_t grew up.

Sponsored by: DARPA & NAI Labs.

22 years agoHook stat into the build
Doug Barton [Thu, 6 Jun 2002 19:36:51 +0000 (19:36 +0000)]
Hook stat into the build
Fix a trailing ws nit while I'm here

22 years agoThis commit was generated by cvs2svn to compensate for changes in r97952,
Doug Barton [Thu, 6 Jun 2002 19:27:17 +0000 (19:27 +0000)]
This commit was generated by cvs2svn to compensate for changes in r97952,
which included commits to RCS files with non-trunk default branches.

22 years agoVirgin import of NetBSD's stat(1)
Doug Barton [Thu, 6 Jun 2002 19:27:17 +0000 (19:27 +0000)]
Virgin import of NetBSD's stat(1)

22 years agoUnhook the old nextboot(8) from the build. If no one yells too loudly, this
Gordon Tetlow [Thu, 6 Jun 2002 19:26:15 +0000 (19:26 +0000)]
Unhook the old nextboot(8) from the build. If no one yells too loudly, this
will be accompanied by a cvs rm of the actual nextboot sources.

Approved by: jake (mentor)

22 years agos/!SIGNOTEMPY/SIGISEMPTY/
Chad David [Thu, 6 Jun 2002 19:12:41 +0000 (19:12 +0000)]
s/!SIGNOTEMPY/SIGISEMPTY/

Reviewed by: marcel, jhb, alfred

22 years agoDocument MTX_DUPOK
Chad David [Thu, 6 Jun 2002 19:08:11 +0000 (19:08 +0000)]
Document MTX_DUPOK

Requested by: jeff (long ago)

22 years agoHandle "dead" witnesses better in the situation of several short term locks
John Baldwin [Thu, 6 Jun 2002 19:04:38 +0000 (19:04 +0000)]
Handle "dead" witnesses better in the situation of several short term locks
being created and destroyed without a single long-term one around to ensure
the witness associated with that group of locks stays alive.  The pipe
mutexes are an example of this group.  For a dead witness we no longer
clear the witness name.  Instead, when looking up the witness for a lock,
if a dead witness' (a witness with a refcount of 0) w_name pointer is
identical to the witness name of the lock then we revive that witness
instead of using a new witness for the lock.  This results in far fewer
dead witness objects and also better preserves locking orders over the long
term resulting in more correct lock order checking.  Note that we can't
ever derefence w_name of a dead witness since we don't know if the string
it is pointing to has been free()'d or kldunload()'d out from under us.

22 years agofix typo in _SYS_SYSPROTO_H_ case: s/mlockall_args/munlockall_args
Alfred Perlstein [Thu, 6 Jun 2002 18:51:14 +0000 (18:51 +0000)]
fix typo in _SYS_SYSPROTO_H_ case: s/mlockall_args/munlockall_args

Submitted by: Mark Santcroos <marks@ripe.net>

22 years agoFactor out some code in preparation for un-kmeming fstat(1).
Dag-Erling Smørgrav [Thu, 6 Jun 2002 18:39:53 +0000 (18:39 +0000)]
Factor out some code in preparation for un-kmeming fstat(1).

Sponsored by: DARPA, NAI Labs

22 years agoRefer to the command format by its SUSv3 name (comm), and list command as an
Juli Mallett [Thu, 6 Jun 2002 18:10:12 +0000 (18:10 +0000)]
Refer to the command format by its SUSv3 name (comm), and list command as an
alias.

22 years agoAs per behaviour on SVR4 systems, to allow any desirable type of header in the
Juli Mallett [Thu, 6 Jun 2002 18:04:46 +0000 (18:04 +0000)]
As per behaviour on SVR4 systems, to allow any desirable type of header in the
override, seperate by comma (',') only, rather than any type of whitespace
(the literal space character (' ') had already been removed from this list).

This allows things like:
miamivice# ps -opid='Process
> Identifier'
Process
Identifier
              1350
              1445
              1450

To work.

22 years agoMatch the DBX_OUTPUT_MAIN_SOURCE_FILE_END definition in dbxelf.h.
David E. O'Brien [Thu, 6 Jun 2002 18:00:58 +0000 (18:00 +0000)]
Match the DBX_OUTPUT_MAIN_SOURCE_FILE_END definition in dbxelf.h.

Bug found by: bde

22 years agoGratuitous whitespace cleanup.
Dag-Erling Smørgrav [Thu, 6 Jun 2002 16:59:24 +0000 (16:59 +0000)]
Gratuitous whitespace cleanup.

22 years agoHook up the ahd driver.
Justin T. Gibbs [Thu, 6 Jun 2002 16:35:58 +0000 (16:35 +0000)]
Hook up the ahd driver.

22 years agoRemove one more multi-line string literal.
Justin T. Gibbs [Thu, 6 Jun 2002 16:07:18 +0000 (16:07 +0000)]
Remove one more multi-line string literal.

22 years agoMove some sysctls from the debug tree to the vfs tree.
Dag-Erling Smørgrav [Thu, 6 Jun 2002 15:50:22 +0000 (15:50 +0000)]
Move some sysctls from the debug tree to the vfs tree.

22 years agoGratuitous whitespace cleanup.
Dag-Erling Smørgrav [Thu, 6 Jun 2002 15:46:38 +0000 (15:46 +0000)]
Gratuitous whitespace cleanup.

22 years agoFixed typos.
Ruslan Ermilov [Thu, 6 Jun 2002 14:08:04 +0000 (14:08 +0000)]
Fixed typos.

22 years agoTidy up.
Ruslan Ermilov [Thu, 6 Jun 2002 14:07:45 +0000 (14:07 +0000)]
Tidy up.

22 years agoTidy up.
Ruslan Ermilov [Thu, 6 Jun 2002 13:55:01 +0000 (13:55 +0000)]
Tidy up.

22 years agolibfetch now depends on libcrypto and libssl.
Ruslan Ermilov [Thu, 6 Jun 2002 13:45:46 +0000 (13:45 +0000)]
libfetch now depends on libcrypto and libssl.

22 years agoBack out rev 1.19 because
Andrey A. Chernov [Thu, 6 Jun 2002 13:44:14 +0000 (13:44 +0000)]
Back out rev 1.19 because

1) It breaks uniq for real life languages when "substitute" directive used in
the collating table.
2) It breaks uniq usage in tool chain with other localized utilities which
use collate.
3) To follow LC_COLLATE it is directly allowed for uniq
by POSIX P1003.1 Draft7 (7.3.2). It means that rev 1.19 gains no additional
POSIX conformance.

22 years agoSupport suidperl as well.
Dag-Erling Smørgrav [Thu, 6 Jun 2002 12:12:23 +0000 (12:12 +0000)]
Support suidperl as well.

22 years agoAdded missing DPADD detected by ``make checkdpadd''.
Ruslan Ermilov [Thu, 6 Jun 2002 11:30:57 +0000 (11:30 +0000)]
Added missing DPADD detected by ``make checkdpadd''.

22 years agoCorrect FreeBSD release of first appearance in the HISTORY section
Sheldon Hearn [Thu, 6 Jun 2002 10:51:25 +0000 (10:51 +0000)]
Correct FreeBSD release of first appearance in the HISTORY section
(5.0 -> 4.6).

22 years agoUse "bwrbg" as description when we sleep for background writing,
Poul-Henning Kamp [Thu, 6 Jun 2002 08:56:10 +0000 (08:56 +0000)]
Use "bwrbg" as description when we sleep for background writing,
"biord" was misleading in every possible way.

22 years agoMake sc_saver_keyb_only (sceen saver interrupted by keyboard input only)
Alfred Perlstein [Thu, 6 Jun 2002 06:02:17 +0000 (06:02 +0000)]
Make sc_saver_keyb_only (sceen saver interrupted by keyboard input only)
the default.

22 years agoDocument the -f and -v options of the unset builtin.
Tim J. Robbins [Thu, 6 Jun 2002 04:12:50 +0000 (04:12 +0000)]
Document the -f and -v options of the unset builtin.

22 years agoQuote the output of the no-argument form of the `set' builtin for re-input
Tim J. Robbins [Thu, 6 Jun 2002 04:02:50 +0000 (04:02 +0000)]
Quote the output of the no-argument form of the `set' builtin for re-input
to the shell.

22 years agoAdd the SUSv3 -p ("portable") option to both the export and readonly
Tim J. Robbins [Thu, 6 Jun 2002 03:57:22 +0000 (03:57 +0000)]
Add the SUSv3 -p ("portable") option to both the export and readonly
builtins. This makes export/readonly print lines in the form
"export name=value".

22 years agoRead specs from /usr/libdata/gcc/specs if it exists.
David E. O'Brien [Thu, 6 Jun 2002 03:47:02 +0000 (03:47 +0000)]
Read specs from /usr/libdata/gcc/specs if it exists.

22 years agoAllow the specification of LONG_TYPE_SIZE on the command line.
David E. O'Brien [Thu, 6 Jun 2002 03:39:03 +0000 (03:39 +0000)]
Allow the specification of LONG_TYPE_SIZE on the command line.

Submitted by: bde

22 years agoImplement "-mno-align-long-strings" which prevents pessimization of strings
David E. O'Brien [Thu, 6 Jun 2002 03:36:32 +0000 (03:36 +0000)]
Implement "-mno-align-long-strings" which prevents pessimization of strings
for space.  -Os could do this, but it was easy to hack an MD version.
This saves a whole 32 bytes in boot2, so I think it is worth using it.
(keep how much worse gcc 3.2 will compile boot2...)

Submitted by: bde (minus gcc 3.2 commentary)

22 years agoDidn't get them all in revision 1.50.
David E. O'Brien [Thu, 6 Jun 2002 03:30:40 +0000 (03:30 +0000)]
Didn't get them all in revision 1.50.

22 years agoDon't modify output that is to be quoted if it contains no IFS characters
Tim J. Robbins [Thu, 6 Jun 2002 03:29:23 +0000 (03:29 +0000)]
Don't modify output that is to be quoted if it contains no IFS characters
or shell metacharacters.

22 years agoFix excessive alignment for the aout case, and unbreak genassym for that case.
David E. O'Brien [Thu, 6 Jun 2002 03:28:49 +0000 (03:28 +0000)]
Fix excessive alignment for the aout case, and unbreak genassym for that case.
Our aout linker has alignment hacks and doesn't need gcc to do anything special.

Tested for a year by: bde :-)

22 years agoTidy do-while.
David E. O'Brien [Thu, 6 Jun 2002 03:26:36 +0000 (03:26 +0000)]
Tidy do-while.

Requested by: bde

22 years agoCompare lines with strcmp(), not strcoll(). We are interested only in
Tim J. Robbins [Thu, 6 Jun 2002 03:13:08 +0000 (03:13 +0000)]
Compare lines with strcmp(), not strcoll(). We are interested only in
equality, not ordering.

22 years agoSync usage() with manual page synopsis.
Tim J. Robbins [Thu, 6 Jun 2002 03:05:30 +0000 (03:05 +0000)]
Sync usage() with manual page synopsis.

22 years agoCast pid_t to long for printf()ing
Brian Somers [Thu, 6 Jun 2002 01:39:46 +0000 (01:39 +0000)]
Cast pid_t to long for printf()ing

Obtained from: OpenBSD

22 years agoRead specs from /usr/libdata/gcc/specs if it exists.
David E. O'Brien [Thu, 6 Jun 2002 00:49:58 +0000 (00:49 +0000)]
Read specs from /usr/libdata/gcc/specs if it exists.

22 years agoFixed overflow in the bounds checking in dscheck(). It assumed that
Bruce Evans [Thu, 6 Jun 2002 00:35:07 +0000 (00:35 +0000)]
Fixed overflow in the bounds checking in dscheck().  It assumed that
daadr_t is no larger than a long, and some other relatively harmless
things (*blush*).  Overflow for subtracting a daddr_t from a u_long
caused "truncation" of the i/o for attempts to access blocks beyond
the end of the actually cause expansion of the i/o to a preposterous
size.

22 years agoA script to generate all of the packages/*/* symlinks described in the INDEX
Brian Somers [Thu, 6 Jun 2002 00:18:49 +0000 (00:18 +0000)]
A script to generate all of the packages/*/* symlinks described in the INDEX

22 years agoA script to validate that the packages and dependencies listed in
Brian Somers [Thu, 6 Jun 2002 00:17:11 +0000 (00:17 +0000)]
A script to validate that the packages and dependencies listed in
the INDEX file exist.

22 years agoAdd 'utility' in a place that it was missing.
Giorgos Keramidas [Wed, 5 Jun 2002 23:39:56 +0000 (23:39 +0000)]
Add 'utility' in a place that it was missing.

PR: docs/38814
Submitted by: Chris Pepper <pepper@rockefeller.edu>
MFC after: 10 days

22 years agoConst'ify variables to make it clear we're not writing to the mbuf data.
Archie Cobbs [Wed, 5 Jun 2002 23:35:31 +0000 (23:35 +0000)]
Const'ify variables to make it clear we're not writing to the mbuf data.

Reviewed by: julian, brian
MFC after: 1 week

22 years agoFix bug where an mbuf was being written to without checking M_WRITABLE().
Archie Cobbs [Wed, 5 Jun 2002 23:32:56 +0000 (23:32 +0000)]
Fix bug where an mbuf was being written to without checking M_WRITABLE().
Eliminate some of the unnecessary complexity of ng_ether_glueback_header().
Simplify two functions a bit by doing the NG_FREE_META(meta) earlier.

Reviewed by: julian, brian
MFC after: 1 week

22 years agoFix bugs where mbuf data was being accessed without m_pullup().
Archie Cobbs [Wed, 5 Jun 2002 23:29:29 +0000 (23:29 +0000)]
Fix bugs where mbuf data was being accessed without m_pullup().

Reviewed by: julian, brian
MFC after: 1 week

22 years agoSilence GCC warnings about multi-line strings.
Justin T. Gibbs [Wed, 5 Jun 2002 22:51:54 +0000 (22:51 +0000)]
Silence GCC warnings about multi-line strings.

Sync Perforce IDs.

22 years agoAdd some PCI-X register definitions.
Justin T. Gibbs [Wed, 5 Jun 2002 22:25:51 +0000 (22:25 +0000)]
Add some PCI-X register definitions.

PCIM_CMD_SERREN -> PCIM_CMD_SERRESPEN to be consistent with
the PERR definition.