]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agofix indentation of a comment
luigi [Sun, 23 Jun 2002 09:14:24 +0000 (09:14 +0000)]
fix indentation of a comment

22 years agofix a typo in a comment
luigi [Sun, 23 Jun 2002 09:13:46 +0000 (09:13 +0000)]
fix a typo in a comment

22 years agoRemove ip_fw_fwd_addr (forgotten in previous commit)
luigi [Sun, 23 Jun 2002 09:03:42 +0000 (09:03 +0000)]
Remove ip_fw_fwd_addr (forgotten in previous commit)
remove some extra whitespace.

22 years agoplxcard for OLDCARD isn't going to happen.
imp [Sun, 23 Jun 2002 07:32:33 +0000 (07:32 +0000)]
plxcard for OLDCARD isn't going to happen.

22 years agoplxcard for OLDCARD almost certainly isn't going to happen.
imp [Sun, 23 Jun 2002 07:31:29 +0000 (07:31 +0000)]
plxcard for OLDCARD almost certainly isn't going to happen.

22 years agoAs disclosed to arch@, make more interfaces standard. This allows for
imp [Sun, 23 Jun 2002 07:27:24 +0000 (07:27 +0000)]
As disclosed to arch@, make more interfaces standard.  This allows for
easier loading of modules that might refer to these interfaces.  None
of the code that implements them is standard, just the glue.  This
bloats the kernel a whopping 8k.

Silence on: arch@

22 years agoRename the BALLOC flags from B_* to BA_* to avoid confusion with the
dillon [Sun, 23 Jun 2002 06:12:22 +0000 (06:12 +0000)]
Rename the BALLOC flags from B_* to BA_* to avoid confusion with the
struct buf B_ flags.

Approved by: mckusick

22 years agoWhat:
imp [Sun, 23 Jun 2002 01:55:10 +0000 (01:55 +0000)]
What:
o ToPIC is happy with two cards now, even when the two cards are
  modems.
o Fix (all?) hangs on boot when power is applied to the card.  I
  suspect that this will make the Ricoh bridges happier and also
  make a lot of VAIO owners happy (confirm to me in private email
  please :-).
o All Cardbus bridges should now support 3.3V, X.XV and Y.YV cards,
  to the extent that the underlying hardware supports such cards.
  (X.X and Y.Y haven't been assigned values yet :-).
o Better 3.3V support for Ricoh ISA bridges.

How:
o Don't mess with the power register when scanning the cards.  It
  is unnecessary and causes BADVcc conditions on many chipsets.  These
  in turn can cause an interrupt storm.
o Make pcic_disable reset the slot's voltage.
o Move initializing voltage for the slot until after it has been
  disabled.
o Fix a lot of issues with the pcic_cardbus_power routine.  We
  now properly enable the card and take it out of reset after
  a power change.
o When detecting the card's voltage, if we're in a BadVcc state,
  direct the bridge to rescan the card for what it supports.
  (we might need to in the future set the power register to 0
   before doing this).
o Don't preserve CLKSTOP.  need to revisit this.
o Better support for Ricoh ISA bridges for 3.3V cards.
o Don't write to PCIC_POWER directly as offten, but instead go
  through the pcic_power interface.
o All cardbus bridges now default to use cardbus power control.
o Add misc register definitions.
o remove some (now) bogus comments.

Extra Special Thanks To: Scott Lamber for his kind and generous loan
of a Toshiba laptop with a ToPIC 100 in it for my use.

22 years ago o Remove the unnecessary acquisition and release of Giant around fdrop()
alc [Sun, 23 Jun 2002 01:48:22 +0000 (01:48 +0000)]
 o Remove the unnecessary acquisition and release of Giant around fdrop()
   in mmap(2).

22 years agoCorrect default location of history file.
grog [Sun, 23 Jun 2002 01:32:00 +0000 (01:32 +0000)]
Correct default location of history file.

Observed by: Chris Bolt <chris-lists@bolt.cx>

22 years agoOnly got one of the Perl usages.
obrien [Sun, 23 Jun 2002 00:57:17 +0000 (00:57 +0000)]
Only got one of the Perl usages.

22 years agoReplace a perl usage with an AWK one.
obrien [Sun, 23 Jun 2002 00:55:03 +0000 (00:55 +0000)]
Replace a perl usage with an AWK one.

Submitted by: Bob Willcox <bob@immure.com>

22 years agoKTR_CT* had one too many trailing zeroes, making KTR_CT5-8 too large for
jake [Sun, 23 Jun 2002 00:38:04 +0000 (00:38 +0000)]
KTR_CT* had one too many trailing zeroes, making KTR_CT5-8 too large for
ktr_mask.

22 years agoFix a bug related to marking pages virtually uncacheable due to illegal
jake [Sat, 22 Jun 2002 23:55:15 +0000 (23:55 +0000)]
Fix a bug related to marking pages virtually uncacheable due to illegal
dcache aliasing.  A page that already had more than 1 mapping of the
same virtual colour would not be correctly uncached.

Noticed by: Artur Grabowski <art@openbsd.org>

22 years agoAdd additional cpuid feature flags and put into a canonical format.
mp [Sat, 22 Jun 2002 23:00:33 +0000 (23:00 +0000)]
Add additional cpuid feature flags and put into a canonical format.

MFC after: 1 week

22 years agoGet rid of paranoia that zeros the boot block area as this has
mckusick [Sat, 22 Jun 2002 22:44:09 +0000 (22:44 +0000)]
Get rid of paranoia that zeros the boot block area as this has
bad effect on existing bootstraps.

Submitted by: Jake Burkholder <jake@locore.ca>
Sponsored by: DARPA & NAI Labs.

22 years agoFix several bugs in the i386 asm statements used to speed up Internet
jdp [Sat, 22 Jun 2002 22:35:53 +0000 (22:35 +0000)]
Fix several bugs in the i386 asm statements used to speed up Internet
checksumming.  These bugs could possibly cause bad code to be
generated at elevated optimization levels.

First, eliminate the use of preprocessor magic to form the address
fields of asm instructions.  It hid the actual addresses being
referenced from the compiler.  Without knowledge of all the data
dependencies, the compiler might possibly use optimizations which
would result in incorrect code.

Use "__asm __volatile" rather than "__asm" for instruction sequences
that pass information through the condition codes (the carry bit, in
this case).  Without __volatile, the compiler might add unrelated
code between consecutive __asm instructions, modifying the condition
codes.  I have seen GCC insert stack pointer adjustments in this
way, for example.  Unfortunately, GCC doesn't provide a way to
specify dependencies on the condition codes.  You can specify that
they are clobbered, but not that you are going to use them as input.

Finally, simplify the LOAD macro.  This macro is used as a poor
man's prefetch.  The simpler version gives the compiler more leeway
about just how it performs the prefetch.

MFC after: 1 week

22 years agoAdd a reference to the "make world" section of the Handbook.
bmah [Sat, 22 Jun 2002 22:23:00 +0000 (22:23 +0000)]
Add a reference to the "make world" section of the Handbook.

PR: 39460
Submitted by: nik

22 years agoThis patch fixes a size problem with the stat structure for
mckusick [Sat, 22 Jun 2002 22:01:13 +0000 (22:01 +0000)]
This patch fixes a size problem with the stat structure for
64-bit architectures that was introduced in the UFS2 code
merge two days ago. The stat structure change that caused
the problem was the addition of the file create time.

Submitted by: Bruce Evans <bde@zeta.org.au>
Sponsored by: DARPA & NAI Labs.

22 years agoNew release notes: MCA support on ia64, rp(4) update (+MFC), tx(4)
bmah [Sat, 22 Jun 2002 21:52:42 +0000 (21:52 +0000)]
New release notes:  MCA support on ia64, rp(4) update (+MFC), tx(4)
multicast filtering, ahd driver, calendar(1) -W/-F, nice(1) -n,
pam_echo(8), pam_exec(8), pselect(3), wc(1) -m, USD/PSD documents
restored.

22 years agoWe don't need to check the return value of malloc() against
mux [Sat, 22 Jun 2002 21:44:11 +0000 (21:44 +0000)]
We don't need to check the return value of malloc() against
NULL when the M_WAITOK flag is specified.

22 years agoThis patch fixes a problem whereby filesystems that ran
mckusick [Sat, 22 Jun 2002 21:24:58 +0000 (21:24 +0000)]
This patch fixes a problem whereby filesystems that ran
out of inodes in a cylinder group would fail to check for
free inodes in other cylinder groups. This bug was introduced
in the UFS2 code merge two days ago.

An inode is allocated by calling ffs_valloc which calls
ffs_hashalloc to do the filesystem scan. Ffs_hashalloc
walks around the cylinder groups calling its passed allocator
(ffs_nodealloccg in this case) until the allocator returns a
non-zero result. The bug is that ffs_hashalloc expects the
passed allocator function to return a 64-bit ufs2_daddr_t.
When allocating inodes, it calls ffs_nodealloccg which was
returning a 32-bit ino_t. The ffs_hashalloc code checked
a 64-bit return value and usually found random non-zero bits in
the high 32-bits so decided that the allocation had succeeded
(in this case in the only cylinder group that it checked).
When the result was passed back to ffs_valloc it looked at
only the bottom 32-bits, saw zero and declared the system
out of inodes. But ffs_hashalloc had really only checked
one cylinder group.

The fix is to change ffs_nodealloccg to return 64-bit results.

Sponsored by: DARPA & NAI Labs.
Submitted by: Poul-Henning Kamp <phk@critter.freebsd.dk>
Reviewed by: Maxime Henrion <mux@freebsd.org>

22 years agoDon't expect NUL terminated data in all netgraph messages received.
brian [Sat, 22 Jun 2002 21:01:47 +0000 (21:01 +0000)]
Don't expect NUL terminated data in all netgraph messages received.
Only display message hook values we understand.

22 years agoNUL terminate the ACNAME passed to userland.
brian [Sat, 22 Jun 2002 21:00:53 +0000 (21:00 +0000)]
NUL terminate the ACNAME passed to userland.

22 years agoWarning fix.
mux [Sat, 22 Jun 2002 20:46:59 +0000 (20:46 +0000)]
Warning fix.

Reviewed by: peter

22 years agoCreate sendmail required directories if sendmail is enabled.
bsd [Sat, 22 Jun 2002 19:44:25 +0000 (19:44 +0000)]
Create sendmail required directories if sendmail is enabled.

22 years agoOLDCARD version of GENERIC.
imp [Sat, 22 Jun 2002 19:23:57 +0000 (19:23 +0000)]
OLDCARD version of GENERIC.

22 years ago o Reduce the scope of Giant in vm_mmap() to just the code that manipulates
alc [Sat, 22 Jun 2002 19:13:56 +0000 (19:13 +0000)]
 o Reduce the scope of Giant in vm_mmap() to just the code that manipulates
   a vnode.  (Thus, MAP_ANON and MAP_STACK never acquire Giant.)

22 years agoFix a bug in vfs_bio_clrbuf(). The single-page-clrbuf optimization was
dillon [Sat, 22 Jun 2002 19:09:35 +0000 (19:09 +0000)]
Fix a bug in vfs_bio_clrbuf().  The single-page-clrbuf optimization was
improperly clearing more then just the invalid portions of the page.  (This
bug is not known to have been triggered by anything).

Submitted by: tegge
MFC after: 7 days

22 years ago o Replace mtx_assert(&Giant, MA_OWNED) in dev_pager_alloc()
alc [Sat, 22 Jun 2002 18:36:51 +0000 (18:36 +0000)]
 o Replace mtx_assert(&Giant, MA_OWNED) in dev_pager_alloc()
   with the acquisition and release of Giant.  (Annotate as MPSAFE.)
 o Reorder the sanity checks in dev_pager_alloc() to reduce
   the time that Giant is held.

22 years agoThe last bits of the alloca -> mmap fix. IA64 and SPARC64 (current only).
dillon [Sat, 22 Jun 2002 18:36:21 +0000 (18:36 +0000)]
The last bits of the alloca -> mmap fix.  IA64 and SPARC64 (current only).
Untested (testing request went unanswered), but sparc64 is not expected to
cause problems.  IA64 is not expected to cause problems but the patch was
slightly more complex so the possibility exists.

Approved by:    jdp

22 years agoUse rm -f in the clean target, as seems to be common practice, and also avoids
jmallett [Sat, 22 Jun 2002 18:16:24 +0000 (18:16 +0000)]
Use rm -f in the clean target, as seems to be common practice, and also avoids
errors if no LINT exists.

Submitted by: dwcjr

22 years agoo Remove the initialization of unused fields in the struct
mux [Sat, 22 Jun 2002 18:07:05 +0000 (18:07 +0000)]
o Remove the initialization of unused fields in the struct
  uio now that we don't use uiomove() anymore.
o Enforce stricter checks on the length of the iov's in
  nmount(2) since we now malloc() them individually and
  corrupted iov's could make the kernel crash in malloc()
  with "kmem_map too small".

Reviewed by: phk

22 years ago o In vm_map_insert(), replace GIANT_REQUIRED by the acquisition and
alc [Sat, 22 Jun 2002 17:47:12 +0000 (17:47 +0000)]
 o In vm_map_insert(), replace GIANT_REQUIRED by the acquisition and
   release of Giant around the direct manipulation of the vm_object and
   the optional call to pmap_object_init_pt().
 o In vm_map_findspace(), remove GIANT_REQUIRED.  Instead, acquire and
   release Giant around the occasional call to pmap_growkernel().
 o In vm_map_find(), remove GIANT_REQUIRED.

22 years agoFor the benefit of those Emacs users amongst us, only cut out gdb.info
obrien [Sat, 22 Jun 2002 17:10:53 +0000 (17:10 +0000)]
For the benefit of those Emacs users amongst us, only cut out gdb.info
rather than *.info.

22 years agoWhen linking staticaly libtermcap is a postrequisite of libreadline.
obrien [Sat, 22 Jun 2002 17:07:45 +0000 (17:07 +0000)]
When linking staticaly libtermcap is a postrequisite of libreadline.
Otherwise `tgoto' (only used by libreadline) isn't resolved.

Submitted by: bde

22 years agoTake the guaranteed fix -- turn off .info docs for now until we get the
obrien [Sat, 22 Jun 2002 16:52:34 +0000 (16:52 +0000)]
Take the guaranteed fix -- turn off .info docs for now until we get the
gdb docs sorted out.

22 years agoMake this compile on Alpha.
n_hibma [Sat, 22 Jun 2002 16:33:29 +0000 (16:33 +0000)]
Make this compile on Alpha.

22 years agoClock frequencies reported by sysctl should be unsigned values. Discovered
mp [Sat, 22 Jun 2002 16:30:18 +0000 (16:30 +0000)]
Clock frequencies reported by sysctl should be unsigned values. Discovered
when machdep.tsc_freq returned a negative number on a 2.2GHz Xeon.

Submitted by: Brian Harrison <bharrison@ironport.com>
Reviewed by: phk
MFC after: 1 week

22 years agoStylistic nit:
jmallett [Sat, 22 Jun 2002 12:59:40 +0000 (12:59 +0000)]
Stylistic nit:

main()'s argv argument is char*[], and functions that may inherit that arg
should use char*[] as well.

22 years agoCheck for results of repeated calls to strnsubst(), as well as for the
jmallett [Sat, 22 Jun 2002 12:58:42 +0000 (12:58 +0000)]
Check for results of repeated calls to strnsubst(), as well as for the
behaviour with NULL match string, as that has changed over time.

22 years agoImprove the handling of Encode and Decode operations in MD5.
phk [Sat, 22 Jun 2002 12:54:11 +0000 (12:54 +0000)]
Improve the handling of Encode and Decode operations in MD5.

Use memcpy for all little-endian architectures, sys/kern/md5c.c indicates
this should be safe for all currently supported LE archs.

Change the Encode and Decode functions for other archs to use le32toh()
and htole32() functions instead of explicit byte shuffling.

On sparc64 this gives md5(1) about 8% speed increase.

22 years agoWhen giving an example that relies on shell expansion/globbing, don't use a
jmallett [Sat, 22 Jun 2002 12:47:56 +0000 (12:47 +0000)]
When giving an example that relies on shell expansion/globbing, don't use a
replstr for -J that will be interpereted by the shell.

MFC after: 1 day

22 years agoRemove (almost all) global variables that were used to hold
luigi [Sat, 22 Jun 2002 11:51:02 +0000 (11:51 +0000)]
Remove (almost all) global variables that were used to hold
packet forwarding state ("annotations") during ip processing.
The code is considerably cleaner now.

The variables removed by this change are:

        ip_divert_cookie        used by divert sockets
        ip_fw_fwd_addr          used for transparent ip redirection
        last_pkt                used by dynamic pipes in dummynet

Removal of the first two has been done by carrying the annotations
into volatile structs prepended to the mbuf chains, and adding
appropriate code to add/remove annotations in the routines which
make use of them, i.e. ip_input(), ip_output(), tcp_input(),
bdg_forward(), ether_demux(), ether_output_frame(), div_output().

On passing, remove a bug in divert handling of fragmented packet.
Now it is the fragment at offset 0 which sets the divert status of
the whole packet, whereas formerly it was the last incoming fragment
to decide.

Removal of last_pkt required a change in the interface of ip_fw_chk()
and dummynet_io(). On passing, use the same mechanism for dummynet
annotations and for divert/forward annotations.

option IPFIREWALL_FORWARD is effectively useless, the code to
implement it is very small and is now in by default to avoid the
obfuscation of conditionally compiled code.

NOTES:
 * there is at least one global variable left, sro_fwd, in ip_output().
   I am not sure if/how this can be removed.

 * I have deliberately avoided gratuitous style changes in this commit
   to avoid cluttering the diffs. Minor stule cleanup will likely be
   necessary

 * this commit only focused on the IP layer. I am sure there is a
   number of global variables used in the TCP and maybe UDP stack.

 * despite the number of files touched, there are absolutely no API's
   or data structures changed by this commit (except the interfaces of
   ip_fw_chk() and dummynet_io(), which are internal anyways), so
   an MFC is quite safe and unintrusive (and desirable, given the
   improved readability of the code).

MFC after: 10 days

22 years agoDefine an mbuf type, MT_TAG, used for volatile annotations
luigi [Sat, 22 Jun 2002 11:29:08 +0000 (11:29 +0000)]
Define an mbuf type, MT_TAG, used for volatile annotations
prepended to mbuf chains in the network stack.
Reuse a previoulsy unused value to avoid changes in other
data structures.

22 years agoReplace the SWAP(var0,var1) macro with SWAP(type,var0,var1) and use it as
jmallett [Sat, 22 Jun 2002 10:44:47 +0000 (10:44 +0000)]
Replace the SWAP(var0,var1) macro with SWAP(type,var0,var1) and use it as
is appropriate to avoid using typeof/__typeof__.  It is worth noting that
SWAP() is only ever used to swap pointer values so 'void *' assumptions would
have been acceptable, but I'd gladly pay you tuesday for a cheeseburger^W
cleaner interface today.

Poked into submission by: bde

22 years agoUnused macro.
jmallett [Sat, 22 Jun 2002 10:34:08 +0000 (10:34 +0000)]
Unused macro.

22 years agoAlways drop the p_args reference we held for copyout, even if we're about
mini [Sat, 22 Jun 2002 10:05:50 +0000 (10:05 +0000)]
Always drop the p_args reference we held for copyout, even if we're about
to change it. This fixes a leak triggered by setproctitle(3).

Approved by: alfred
Noticed by: Peter Jeremy <peter.jeremy@alcatel.com.au>

22 years agoGrre-^#$(*+#@$!@% I hate .info files, I really do (thus I have NO_INFO set...)
obrien [Sat, 22 Jun 2002 08:08:51 +0000 (08:08 +0000)]
Grre-^#$(*+#@$!@% I hate .info files, I really do (thus I have NO_INFO set...)
Don't try to read files from a non-existant gdb.291/.

22 years ago o Replace GIANT_REQUIRED in swap_pager_alloc() by the acquisition and
alc [Sat, 22 Jun 2002 08:03:21 +0000 (08:03 +0000)]
 o Replace GIANT_REQUIRED in swap_pager_alloc() by the acquisition and
   release of Giant.  (Annotate as MPSAFE.)

22 years ago o Remove GIANT_REQUIRED from phys_pager_alloc(). If handle isn't NULL,
alc [Sat, 22 Jun 2002 07:54:42 +0000 (07:54 +0000)]
 o Remove GIANT_REQUIRED from phys_pager_alloc().  If handle isn't NULL,
   acquire and release Giant.  If handle is NULL, Giant isn't needed.
 o Annotate phys_pager_alloc() and phys_pager_dealloc() as MPSAFE.

22 years ago o Replace GIANT_REQUIRED in vnode_pager_alloc() by the acquisition and
alc [Sat, 22 Jun 2002 07:28:06 +0000 (07:28 +0000)]
 o Replace GIANT_REQUIRED in vnode_pager_alloc() by the acquisition and
   release of Giant.  (Annotate as MPSAFE.)
 o Also, in vnode_pager_alloc(), remove an unnecessary re-initialization
   of struct vm_object::flags and move a statement that is duplicated
   in both branches of an if-else.

22 years agoNow that the pattern space contains no trailing newline, modify the `l'
tjr [Sat, 22 Jun 2002 03:00:52 +0000 (03:00 +0000)]
Now that the pattern space contains no trailing newline, modify the `l'
command's output so it's the same as what SUSv3 specifies.

22 years agoMake it clear that this applies only to UFS1 file systems, as UFS2 will
rwatson [Sat, 22 Jun 2002 01:49:03 +0000 (01:49 +0000)]
Make it clear that this applies only to UFS1 file systems, as UFS2 will
have native extended attributes rather than stacked extended attributes.
While I'm at it, make sure UFS_EXTATTR is not spelt FFS_EXTATTR.

Sponsored by: DARPA, NAI Labs
Obtained from: TrustedBSD Project

22 years agoDon't store newlines at the end of each line in the hold/pattern spaces,
tjr [Sat, 22 Jun 2002 01:42:26 +0000 (01:42 +0000)]
Don't store newlines at the end of each line in the hold/pattern spaces,
instead add the newline when the pattern space is printed. Make the `G' and
`H' commands add a newline to the space before the data, remove bogus
addition of newline from `x' command.

PR: 29790, 38195

22 years ago o Remove GIANT_REQUIRED from vslock().
alc [Sat, 22 Jun 2002 01:26:02 +0000 (01:26 +0000)]
 o Remove GIANT_REQUIRED from vslock().
 o Annotate kernacc(), useracc(), and vslock() as MPSAFE.

Motivated by: alfred

22 years agoTweaks to make these files live beside our hacked config.
obrien [Fri, 21 Jun 2002 22:54:24 +0000 (22:54 +0000)]
Tweaks to make these files live beside our hacked config.

22 years agoFix logic which resulted in missing a call to INP_UNLOCK().
hsu [Fri, 21 Jun 2002 22:54:16 +0000 (22:54 +0000)]
Fix logic which resulted in missing a call to INP_UNLOCK().

Submitted by: jlemon, mux

22 years agoTweaks for the 64-bit compiler.
obrien [Fri, 21 Jun 2002 22:53:03 +0000 (22:53 +0000)]
Tweaks for the 64-bit compiler.

22 years agoTCP notify functions can change the pcb list.
hsu [Fri, 21 Jun 2002 22:52:48 +0000 (22:52 +0000)]
TCP notify functions can change the pcb list.

22 years agoEnable cd9660 support by default.
jake [Fri, 21 Jun 2002 22:34:51 +0000 (22:34 +0000)]
Enable cd9660 support by default.

22 years agoEnable UFS1_AND_UFS2 support for sparc64 by default. Booting from ufs1 or
jake [Fri, 21 Jun 2002 22:33:56 +0000 (22:33 +0000)]
Enable UFS1_AND_UFS2 support for sparc64 by default.  Booting from ufs1 or
ufs2 filesystems seems to work fine.

22 years agoRemove CAPABILITIES from NOTES
rwatson [Fri, 21 Jun 2002 19:53:04 +0000 (19:53 +0000)]
Remove CAPABILITIES from NOTES

22 years agoHone the rpcbind dependency checking.
dougb [Fri, 21 Jun 2002 19:50:01 +0000 (19:50 +0000)]
Hone the rpcbind dependency checking.

Submitted by: Mike Makonnen <makonnen@pacbell.net>

22 years agoClarify revision 1.14:
obrien [Fri, 21 Jun 2002 19:24:12 +0000 (19:24 +0000)]
Clarify revision 1.14:
Gcc 3.1's -O0 and -O1 actually optimized alignment for space, but we feel
it should optimize alignment for time like Gcc 2.95 used to.  Optimization
for space should give 1-byte alignment on i386's, but doesn't quite.

22 years agoForced commit to keep cvs happy.
peter [Fri, 21 Jun 2002 19:21:41 +0000 (19:21 +0000)]
Forced commit to keep cvs happy.

22 years agoForced commit to keep cvs happy.
peter [Fri, 21 Jun 2002 19:21:32 +0000 (19:21 +0000)]
Forced commit to keep cvs happy.

22 years agoForced commit to keep cvs happy.
peter [Fri, 21 Jun 2002 19:19:22 +0000 (19:19 +0000)]
Forced commit to keep cvs happy.

22 years agoForced commit to keep cvs happy.
peter [Fri, 21 Jun 2002 19:19:06 +0000 (19:19 +0000)]
Forced commit to keep cvs happy.

22 years agoForced commit to note we are pretending contrib/gdb.291 never existed.
obrien [Fri, 21 Jun 2002 19:09:45 +0000 (19:09 +0000)]
Forced commit to note we are pretending contrib/gdb.291 never existed.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r98576,
obrien [Fri, 21 Jun 2002 19:07:21 +0000 (19:07 +0000)]
This commit was generated by cvs2svn to compensate for changes in r98576,
which included commits to RCS files with non-trunk default branches.

22 years agoSync with NetBSD's mainline.
obrien [Fri, 21 Jun 2002 19:07:21 +0000 (19:07 +0000)]
Sync with NetBSD's mainline.

22 years agoWe need to link with libiberty.a before libreadline.so. Both supply
obrien [Fri, 21 Jun 2002 19:05:20 +0000 (19:05 +0000)]
We need to link with libiberty.a before libreadline.so.  Both supply
xmalloc() and xrealloc() and the mixed usage of xmalloc in some .c's from
libiberty.a and other .c's from libreadline.so produces an unusable binary
on the Alpha.

While I am here, preventatively move other libs in the link order.

Submitted by: gallatin

22 years agoForced commit after RCS file surgery.
peter [Fri, 21 Jun 2002 18:51:53 +0000 (18:51 +0000)]
Forced commit after RCS file surgery.

22 years agocc -O0 and -O1 didn't do the easy optimization of alignment for space.
obrien [Fri, 21 Jun 2002 18:21:30 +0000 (18:21 +0000)]
cc -O0 and -O1 didn't do the easy optimization of alignment for space.
Instead it attempted to do the easy optimization of alignment for time,
which should be to 1-byte alignment on i386's.

Submitted by: bde

22 years agoBDE prefers this organization.
obrien [Fri, 21 Jun 2002 18:11:22 +0000 (18:11 +0000)]
BDE prefers this organization.

Submitted by: bde

22 years agoSync with NetBSD.
obrien [Fri, 21 Jun 2002 15:56:16 +0000 (15:56 +0000)]
Sync with NetBSD.
* don't whine about non-regular files. It is perfectly normal to keep a
  CVS or RCS directory in /etc/rc.d.
* manpage tweak

22 years agoThis commit was generated by cvs2svn to compensate for changes in r98567,
obrien [Fri, 21 Jun 2002 15:56:16 +0000 (15:56 +0000)]
This commit was generated by cvs2svn to compensate for changes in r98567,
which included commits to RCS files with non-trunk default branches.

22 years agoReduce the -I's as we don't need them all.
obrien [Fri, 21 Jun 2002 15:52:05 +0000 (15:52 +0000)]
Reduce the -I's as we don't need them all.

22 years agoNew release notes: ng_device(4), UFS2, newfs(8) -O, newgrp(1),
bmah [Fri, 21 Jun 2002 15:36:24 +0000 (15:36 +0000)]
New release notes:  ng_device(4), UFS2, newfs(8) -O, newgrp(1),
pam_ksu(8), uuidgen(1).

Updated release notes:  libpcap-0.7.1, tcpdump-3.7.1.

Also do some markup fixes to recently-added items so that they don't
try to use entities for manpages that have been removed, in the spirit
of rev. 1.359.

22 years ago__FBSDID() strategic insertion.
jmallett [Fri, 21 Jun 2002 11:52:59 +0000 (11:52 +0000)]
__FBSDID() strategic insertion.

22 years agoKill __P, yuck.
jmallett [Fri, 21 Jun 2002 11:42:37 +0000 (11:42 +0000)]
Kill __P, yuck.

22 years agoMark unused variables __unused.
jmallett [Fri, 21 Jun 2002 11:40:03 +0000 (11:40 +0000)]
Mark unused variables __unused.

Built standalone, inetd(8) is WARNS=5 clean, WARNS=6 if you ignore %m fits.

22 years agoUse __typeof__ instead of typeof.
jmallett [Fri, 21 Jun 2002 11:25:11 +0000 (11:25 +0000)]
Use __typeof__ instead of typeof.

22 years agoKill bad whitespace and do some style cleanups as a result of the protoize.
jmallett [Fri, 21 Jun 2002 11:24:21 +0000 (11:24 +0000)]
Kill bad whitespace and do some style cleanups as a result of the protoize.

22 years agoANSI prototypes via protoize(1).
jmallett [Fri, 21 Jun 2002 11:18:42 +0000 (11:18 +0000)]
ANSI prototypes via protoize(1).

22 years agoBetter fix for style.
markm [Fri, 21 Jun 2002 11:06:11 +0000 (11:06 +0000)]
Better fix for style.

Suggested by: ru (but modified a bit by markm)

22 years agoRevert the part of Kirks UFS2 commit which added divdi3.c and moddi3.c
phk [Fri, 21 Jun 2002 11:00:00 +0000 (11:00 +0000)]
Revert the part of Kirks UFS2 commit which added divdi3.c and moddi3.c
to libi386, this issue was resolved already in a cleaner way.

22 years agoClean up hex() and octal() to return and work with unsigned integers since
jmallett [Fri, 21 Jun 2002 10:56:44 +0000 (10:56 +0000)]
Clean up hex() and octal() to return and work with unsigned integers since
they scan values of unsigned types, and since they do not need otherwise,
have them take const char * arguments.

22 years agoFix warnings generated elsewhere.
markm [Fri, 21 Jun 2002 10:22:39 +0000 (10:22 +0000)]
Fix warnings generated elsewhere.

22 years agoStyle tidy-up.
markm [Fri, 21 Jun 2002 10:21:21 +0000 (10:21 +0000)]
Style tidy-up.

22 years agoModernise this code by rounding up the usual suspects: register keyword,
markm [Fri, 21 Jun 2002 10:14:57 +0000 (10:14 +0000)]
Modernise this code by rounding up the usual suspects: register keyword,
consts, ANSIfication of functions, proper use of __FBSDID() and so on.
This commit is non-functional from a code-execution perspective, but
pr(1) now compiles WARNS=6 clean.

22 years agoDon't say that an umounted filesystem is required for -n. An unmounted
bde [Fri, 21 Jun 2002 09:57:34 +0000 (09:57 +0000)]
Don't say that an umounted filesystem is required for -n.  An unmounted
filesystem is no more or less required for -n than for any other option.
The previous commit clarified the actual requirement.

22 years agoFix incorrect library ordering. I thought I'd committed this already...
des [Fri, 21 Jun 2002 09:56:38 +0000 (09:56 +0000)]
Fix incorrect library ordering.  I thought I'd committed this already...

22 years ago#include <sys/time.h> before <sys/stat.h> to get timespec.
phk [Fri, 21 Jun 2002 09:23:12 +0000 (09:23 +0000)]
#include <sys/time.h> before <sys/stat.h> to get timespec.

22 years agoMake NO_OPENSSL actually imply NO_OPENSSH, as documented in make.conf(5).
ru [Fri, 21 Jun 2002 08:54:03 +0000 (08:54 +0000)]
Make NO_OPENSSL actually imply NO_OPENSSH, as documented in make.conf(5).

22 years agoSkip fields in the manner required by POSIX, and the way V7 did it.
tjr [Fri, 21 Jun 2002 07:59:31 +0000 (07:59 +0000)]
Skip fields in the manner required by POSIX, and the way V7 did it.

MFC after: 1 week

22 years agoQuiet ``make objlink'' when NOOBJ is defined.
ru [Fri, 21 Jun 2002 07:30:51 +0000 (07:30 +0000)]
Quiet ``make objlink'' when NOOBJ is defined.

PR: bin/21142
Submitted by: Craig Leres <leres@ee.lbl.gov>

22 years agoNewline characters should not participate in line comparisons. Only apparent
tjr [Fri, 21 Jun 2002 07:08:34 +0000 (07:08 +0000)]
Newline characters should not participate in line comparisons. Only apparent
when -s is used or the last line of the file is missing a newline.
Noticed by the textutils test suite.

MFC after: 1 week

22 years agoA gross patch to tidy up the formatting.
ru [Fri, 21 Jun 2002 06:50:16 +0000 (06:50 +0000)]
A gross patch to tidy up the formatting.

Not objected to by: grog