On Mon, 25 Jun 2018 22:19:11 +0200, Paul de Weerd wrote:
> It comes from this bit in usr.bin/ssh/dh.c:
>
> linenum = 0;
> which = arc4random_uniform(bestcount);
"bestcount" is always > 1, but "which" is in the range [0, bestcount)
where we really want it to be [1, bestcount]
I th
Actually, it appears that the code accounts for which being one
less. The problem is the additional linenum++ introduced in rev
1.64 along with the getline() changes. We should only be incrementing
linenum for each suitable prime, not every line.
- todd
Index: usr.bin/ssh/dh.c
On Mon, 25 Jun 2018 22:19:11 +0200, Paul de Weerd wrote:
> As far as I can see, these are all caused by 'the internet' trying to
> log in: I've not had any problems with sshd misbehaving and when I log
> in no such warning is logged. It happened for 'Invalid user' (36x), 'Failed
> password' (8x),
On Tue, 26 Jun 2018 08:16:24 +0200, Paul de Weerd wrote:
> I think this is correct. While looking at that piece of code, I
> completely missed the incrementing of linenum. Now that you point it
> out explicitly, it's quite obvious that this was the problem all
> along. I'll put this on my ssh j
You should be able to use the URI syntax for this. E.g.
scp scp://user:skey@localhost/tmp/somefile .
- todd
On Tue, 14 Aug 2018 10:43:30 +0200, Martijn van Duren wrote:
> The diff below fixes this. Note that I took special care to make a
> distinction between in place and normal for the 'q' command.
> When running normally the files are concatenated, so we should quit
> immediately. When running in in p
On Tue, 14 Aug 2018 15:16:47 +0200, Martijn van Duren wrote:
> So gnu is somewhat contradictory here. On the one hand it treats
> every file as a new script, based on the hold-space, but the 'q'
> command exits the editor all together.
The 'q' command should quit the editor, not just the current
The fix is to look at mode, not oflags since "w+" will set O_RDWR,
not O_WRONLY.
- todd
Index: lib/libc/stdio/fmemopen.c
===
RCS file: /cvs/src/lib/libc/stdio/fmemopen.c,v
retrieving revision 1.3
diff -u -p -u -r1.3 fmemopen.c
--- l
On Thu, 13 Aug 2020 19:31:04 -0600, Todd C. Miller wrote:
> The fix is to look at mode, not oflags since "w+" will set O_RDWR,
> not O_WRONLY.
Actually, checking O_TRUNC is probably better and is consistent
with the rest of the code. Either diff produces the expected result
On Fri, 14 Aug 2020 03:05:38 -, phillb...@cock.li wrote:
> Also, is size for *mode == 'a' handled correctly?
Yes, there is special handling for O_APPEND that sets the size.
- todd
On Fri, 14 Aug 2020 07:04:13 -0600, Todd C. Miller wrote:
> Yes, there is special handling for O_APPEND that sets the size.
However, there are other bugs with respect to append. I've added
an append flag to the state but we could just as easily store the
open flag instead.
I check
On Wed, 30 Sep 2020 15:25:36 -0400, "Demi M. Obenour" wrote:
> If I run
>
> cksum -a sha256x -ph /dev/fd/3 < /home/_sysupgrade/base68.tgz 3>&1 >/mnt/falc
> on
>
> and the filesystem mounted on /mnt is too small to hold base68.tgz,
> cksum(1) should return an error. Instead, the error is silently
For some reason, cmplabel() in disklabel.c doesn't compare the
bounds values when it checks for changes. I'm not sure what the
reason for that is, it seems like we should write the new label
even if the only change is in the bounds.
- todd
On Mon, 19 Oct 2020 13:26:06 -0600, "Theo de Raadt" wrote:
> Todd C. Miller wrote:
>
> > For some reason, cmplabel() in disklabel.c doesn't compare the
> > bounds values when it checks for changes. I'm not sure what the
> > reason for that is,
On Fri, 30 Oct 2020 15:03:41 +0100, Piotr Durlej wrote:
> auth_approval(3) doesn't always set errno(2) on failure:
Committed, thanks.
- todd
On Tue, 17 Nov 2020 20:07:15 -0300, Gleydson Soares wrote:
> Accordingly to
> https://tools.ietf.org/html/draft-ietf-secsh-scp-sftp-ssh-uri-04
> "host" part of the URI referenced at
> https://tools.ietf.org/html/rfc3986#section-3.2.2
> it can be an IPv6/IPv4 address or a domain name itself, so th
On Mon, 18 Jan 2021 14:17:43 +0200, Jyrki Saarela wrote:
> However, now that I actually can access the box again,
> /var/run/dmesg.boot contains
> uid 0 on /: out of inodes
> uid 0 on /: out of inodes
> uid 0 on /: out of ioodes
> uid 0 on /: out of inodes
> uid 0 on /: out of inodes
> uid 0 on /:
On Mon, 04 Nov 2019 09:11:10 +0100, Rudolf Leitgeb wrote:
> This potentially erroneous call to put_entries() happens only, if
> a malloc at a very early stage fails. In my opinion, the call to
> put_entries in pfnote() should be wrapped.
That fix looks correct to me.
- todd
On Wed, 06 Nov 2019 15:48:31 -0700, Allen Smith wrote:
> When booting into the 6.6 kernel for the first time on a EdgeRouter 4, the
> host starts bringing up services and when the pppoe interface starts to
> come up I see the following:
You are not the only one to run into this. There's a workar
On Sun, 17 Nov 2019 20:38:59 +0100, Alexander Bluhm wrote:
> I think the best way to handle it, is to make the kernel strict and
> fix userland. If the kernel would allow the sloppiest userland
> program to succeed, creating security would be hard.
Sorry, I don't agree. We cannot expect userlan
On Mon, 18 Nov 2019 19:02:13 +0100, Alexander Bluhm wrote:
> The get functions like SIOCGIFAFLAG_IN6 did not have a length check.
> There the incoming address is used to find the correct interface
> address for the requested information. This is what dhcpcd uses.
>
> For the inet6 netmask the sin
On Thu, 21 Nov 2019 11:52:53 +0100, Raimo Niskanen wrote:
> On OpenBSD 6.6 amd64 patch 006, i get peculiar results from readlink(1)
> with arguments -f with a symlink to /
>
> $ readlink -f /
> /
>
> $ ln -s / test; readlink -f test; rm test
> readlink: test: Is a directory
That appears to be a b
On Fri, 13 Dec 2019 22:37:31 +0200, Dennis Lindroos wrote:
> If the user argument has the ":style" suffix attached to it then
> getpwnam(user) will return a NULL pointer.
> I tried just using strsep(3) to strip off the auth style and it works for
> me (this is probably not the safest bit of coding
This sounds like the loop in softdep_process_worklist() is never
exiting. It shouldn't run for more than a second, though.
FreeBSD breaks out of the loop if process_worklist_item() can't
make progress. You could try the following (untested) diff to see
if it changes the behavior.
- todd
Index
On Tue, 03 Mar 2020 10:20:04 +0100, Mark Patruck wrote:
> After ~3 days with the system up and running, the crash after doing
> a "reboot" looks different. Now it's in handle_workitem_freeblocks(),
> according to objdump
The problem is that removed files were not actually getting removed
since pr
Here is a minimal fix that only addresses the tight CPU loop in
softdep_process_worklist(). It will exit the loop if we cannot
make progress instead of spinning.
process_worklist_item() now returns 1 if it processed an item or 0
if it could not. The existing semantics of matchcnt have been
prese
On Sat, 07 Mar 2020 19:35:10 -0700, Bob Beck wrote:
> makes sense to me and has my ok. could we see if bluhm@ can be sure this
> still works with his workload?
Thanks, waiting to see if bluhm@ can confirm this doesn't cause
problems makes sense. I'm currently travelling but will be home
this
I just committed my minimal fix.
- todd
On Thu, 16 Apr 2020 23:39:29 +0200, Ingo Schwarze wrote:
> I believe the patch is correct and i'd like to commit it.
>
> I have tested all code path directly leading from gz_open()
> into check_header().
>
> However, i'm not quite sure how to test the call of check_header()
> from gzread(). Marc,
Is there any advantage to mfs defaulting to ffs2?
- todd
On Tue, 19 May 2020 14:04:37 +0200, Otto Moerbeek wrote:
> In 18 years, yes. But the -O2 case should work whartever the default
> is for mfs.
I agree that -O2 should work for mfs, I'm just wasn't sure that
should be the default for mfs. We don't actually have a way to
specify the ffs version wit
On Fri, 22 May 2020 00:06:28 +0200, Sebastian Benoit wrote:
> Hi Leah,
>
> thanks for your report and researching this!
>
> Anyone else happy with this?
> I'd like to commit it ;)
Yes, OK millert@
- todd
On Mon, 25 May 2020 16:04:25 -0600, Bob Beck wrote:
> getlock()'s behaviour changes in the case of a writeable mail spool. if we
> want to keep supporting this, I we can modify the pledge as follows:
I thought we decided not to adjust the pledge when I brought it up
last time. Here's the diff I
On Thu, 11 Jun 2020 12:36:27 +0100, Stuart Henderson wrote:
> This "fixes" it ...
>
> I think the most sensible approach for now is the backout diff
> in my previous mail. Any OKs for that?
The strlcpy() is wrong now that inputFS is a pointer.
It should be:
strlcpy(inputFS, *FS, len_inputFS
This should be fixed by the commit I just made to awk/lib.c.
The strlcpy() length parameter was incorrect.
- todd
On Tue, 28 Jun 2016 15:24:43 -, Florian Obser wrote:
> There is still a bug:
>
> $ mkdir foo; cd foo; touch bar; rm -rf "" bar; echo $?; ls -la; \
> rm -rf nonexistend bar; echo $?; ls -la
> 0
> total 8
> drwxr-xr-x 2 florian wheel 512 Jun 28 17:20 ./
> drwxrwxrwt 9 root wheel 51
On Tue, 28 Jun 2016 10:03:00 -0700, Philip Guenther wrote:
> On Tue, 28 Jun 2016, Todd C. Miller wrote:
> > I think this needs to be fixed in fts(3) instead. The following diff
> > fixes it for me but has only been lightly tested.
>
> As I noted in icb, first chunk looks
On Tue, 28 Jun 2016 10:20:43 -0700, patrick keshishian wrote:
> Since POSIX was mentioned, "" is not a valid filename[1]. So
> it isn't the case of a file does not exist (ENOENT), it is a case
> of an invalid filename (EINVAL?).
I don't think it really matters. Furthermore, callers of fts(3)
are
On Mon, 08 Aug 2016 12:48:49 +0200, Marc Espie wrote:
> Well, unless someone can repair connect, here's a very ugly hack that
> should prevent SIGWINCH from fucking ftp up.
It's not like SIGWINCH is special, let's just handle EADDRINUSE
preceded by EINTR.
- todd
Index: usr.bin/ftp/ftp.c
==
POSIX says connect(2) does not restart so we need to do the same
kind of dance as async connect(2).
- todd
Index: usr.bin/ftp/ftp.c
===
RCS file: /cvs/src/usr.bin/ftp/ftp.c,v
retrieving revision 1.96
diff -u -p -u -r1.96 ftp.c
--- u
On Mon, 08 Aug 2016 13:11:48 -0700, Philip Guenther wrote:
> Untested, but the obvious port from the other BSDs to have connect() leave
> the connect running asynchronously.
Looks good, OK millert@
- todd
On Mon, 08 Aug 2016 13:11:48 -0700, Philip Guenther wrote:
> Untested, but the obvious port from the other BSDs to have connect() leave
> the connect running asynchronously.
We should also update the manual accordingly. Alternately, we could move
the useful bits out of EINPROGRESS and into the
On Mon, 08 Aug 2016 18:20:38 -0700, Philip Guenther wrote:
> Testing found a bug: returning EALREADY shouldn't be dependent on the
> socket being nonblocking. Yay testing.
That seems reasonable--the EINTR case should behave like the
non-blocking case.
> Second chunk the the tweak to the libpth
On Mon, 08 Aug 2016 14:35:39 -0600, "Todd C. Miller" wrote:
> We should also update the manual accordingly. Alternately, we could move
> the useful bits out of EINPROGRESS and into the main body of the manual
> and simply refer to them in ERRORS.
This diff moves some of the
On Wed, 17 Aug 2016 20:50:25 -0400, Michael Reed wrote:
> Thanks everyone for looking into this, but I'm still experiencing the
> pkg_add issue in snapshots:
>
> $ sysctl -n kern.version
>
> OpenBSD 6.0-current (GENERIC.MP) #2366
On Wed, 23 Nov 2016 12:38:11 +0100, Marc Espie wrote:
> Thinking some more, I'm pretty sure this should also be fixed at the shell
> level.
>
> I see nothing in POSIX that says the shell ought to provide non default
> handlers for these signals through, as opposed to make(1) which has a full
> pa
Thanks, I've broken these up into smaller diffs and committed them.
I did need to add expected-stdout and expected-stderr lines for
and-list-error-3 to prevent it from failing since the
test harness reports a failure if there is unexpected output.
- todd.
This is probably due to cron caching the spool dir mtime as a time_t
instead of a struct timespec.
- todd
Here's a diff to use st_mtim instead of st_mtime which fixes the
issue.
- todd
Index: usr.sbin/cron/atrun.c
===
RCS file: /cvs/src/usr.sbin/cron/atrun.c,v
retrieving revision 1.42
diff -u -p -u -r1.42 atrun.c
--- usr.sbin/cron/atrun
On Mon, 11 Jan 2016 11:06:11 +0100, viq wrote:
> Indeed it does, same operations now result in logs like this:
> Jan 11 10:03:12 OpenBSD-current-amd64 crontab[6314]: (root) REPLACE
> (root)
> Jan 11 10:03:12 OpenBSD-current-amd64 cron[21738]: (root) RELOAD (root)
> Jan 11 10:03:12 OpenBSD-current-
Here's a simple fix. I've matched the indent style of the existing
code.
- todd
Index: lib/libc/gen/fnmatch.c
===
RCS file: /cvs/src/lib/libc/gen/fnmatch.c,v
retrieving revision 1.19
diff -u -p -u -r1.19 fnmatch.c
--- lib/libc/gen/
Committed, thanks.
- todd
To reproduce this you need to have /etc/malloc.conf set to fill
with junk. Below is a simple workaround, there may be a better
solution though.
- todd
Index: lib.c
===
RCS file: /cvs/src/usr.bin/awk/lib.c,v
retrieving revision 1.21
I think it is best to simply initialize "record" to be the empty
string. This is sufficient to fix the crash.
- todd
Index: usr.bin/awk/lib.c
===
RCS file: /cvs/src/usr.bin/awk/lib.c,v
retrieving revision 1.21
diff -u -p -u -r1.21
On Tue, 17 May 2016 16:21:47 +0200, Theo Buehler wrote:
> I agree with your diagnosis. skeyinit tries to fchown the file to the
> target user and gets EPERM since it is running with pledge.
>
> Here's a patch that disables pledge for skeyinit if it is run as root
> and there is a target user spec
On Tue, 17 May 2016 17:03:06 +0200, Theo Buehler wrote:
> > @@ -151,6 +157,11 @@ main(int argc, char **argv)
> > } else if (strcmp(pp->pw_name, me) != 0 && getuid() != 0) {
> > /* Only root can change other's S/Keys. */
> > errx(1, "Permission de
Seems like a good candidate for ratecheck(9).
- todd
On Thu, 26 May 2016 23:24:16 +0200, Tim Kuijsten wrote:
> rename s to dst, just like in strncat, strcpy and strncpy.
Committed, thanks.
- todd
This should be fixed by rev 1.177 of i386 pmap.c that mlarkin@
committed today. An updated i386 snapshot with the fix will be
available soon.
- todd
Solaris and Linux also produce "x\ny\nz\n" so that is likely the
expected behavior.
- todd
On Sat, 09 May 2015 06:47:05 +0200, =?utf-8?Q?S=C3=A9bastien?= Marie wrote:
> Here a small patch to sed to make 'i' and 'a' command to always append
> "\n" after 'text'.
>
> While here, remove 'len' field from 'struct s_appends'. It was just used
> for AP_STRING (used for 'a' command), and the sw
On Thu, 09 Jul 2015 19:00:30 +1200, Peter Kane wrote:
> Just a note to mention that man release in -current still refers to
> sudo in a number of places.
Yes, they are being kept as placeholders until there is a sudo
replacement.
- todd
On Thu, 30 Jul 2015 17:50:29 +0300, Vadim Zhukov wrote:
> Too bad that I don't know how to make a test for this issue in
> regress/usr.bin/ksh, though. It could tests stdin, stdout, stderr and
> files, but I dunno how to test the shell prompt. Any ideas?
How about this?
- todd
Index: regress.t
On Thu, 20 Aug 2015 02:09:52 -1000, Antonio Querubin wrote:
> The HTTP daemon on ftp5.usa.openbsd.org is refusing connections on it's
> IPv6 address.
Fixed.
- todd
On Sat, 24 Oct 2015 02:38:27 +0200, cedric.kr...@b2ck.com wrote:
> >Synopsis:sort ignore reverse option if ignore-leading-blanks is set
> >Category:system
> >Environment:
> System : OpenBSD 5.8
> Details : OpenBSD 5.8 (GENERIC.MP) #1: Wed Oct 14 19:38:08 CEST 201
> 5
>
On Thu, 29 Oct 2015 21:38:07 -0400, "Ted Unangst" wrote:
> fc -e - | -s [-g] [old=new] [prefix]
>
> 1. What does fc stand for? The manual doesn't tell me.
According to bash it is "fix command" though I thought it was "find
command". Who knows?
> 2. What's prefix?
I believe that lets you edit
On Thu, 29 Oct 2015 20:01:28 -0700, Philip Guenther wrote:
> We should also fix the built-in alias 'r' to use fc -s instead of fc -e -
> Diff below.
OK millert@
- todd
FWIW, "fc" has been renamed "hist" in ksh93. I'm not sure we want
to do that as it is likely to cause confusion with the csh-like
history.
- todd
Is this an improvement?
- todd
Index: ksh.1
===
RCS file: /cvs/src/bin/ksh/ksh.1,v
retrieving revision 1.163
diff -u -p -u -r1.163 ksh.1
--- ksh.1 30 Oct 2015 03:13:52 - 1.163
+++ ksh.1 30 Oct 2015 03:36:30 -000
On Sun, 08 Nov 2015 12:30:47 +0100, Theo Buehler wrote:
> * We can make a pledge("id") at the start. Drop this after setrlimit(2)
> * Try to find the kmem group early on and use setegid(2) instead of
> initgroups(2). Pass kmem's gid as an argument to kvm_mkdb().
> * If the kmem group wasn't fo
On Sun, 08 Nov 2015 12:30:47 +0100, Theo Buehler wrote:
> * We can make a pledge("id") at the start. Drop this after setrlimit(2)
> * Try to find the kmem group early on and use setegid(2) instead of
> initgroups(2). Pass kmem's gid as an argument to kvm_mkdb().
> * If the kmem group wasn't fo
There is a missing rip() before the strtonum(). The rest of the
diff allows the -n option to work with -s.
- todd
Index: usr.bin/skeyinit/skeyinit.c
===
RCS file: /cvs/src/usr.bin/skeyinit/skeyinit.c,v
retrieving revision 1.62
diff
On Sat, 28 Nov 2015 20:19:22 -0800, Serguey Parkhomovsky wrote:
> Looks like you beat me to it! skeyinit still needs pledge getpw, though:
It's not needed for the plain passwd db but maybe it is for YP.
I've added getpw in the pledge call and then revoked getpw, proc,
and exec once authentication
On Thu, 10 Aug 2017 12:27:10 -0400, Martin Pieuchot wrote:
> On 10/08/17(Thu) 18:21, Mark Kettenis wrote:
> > > Date: Thu, 10 Aug 2017 12:10:27 -0400
> > > From: Martin Pieuchot
> > >
> > > Building a profiled binary, using -pg with clang doesn't work as
> > > expected. A gmon.out is properly g
On Sat, 12 Aug 2017 14:51:59 +0200, =?UTF-8?Q?Ren=c3=a9_Scharfe?= wrote:
> getdelim(3) and getline(3) fail if they are unable to allocate
> enough memory to hold a line. recallocarray(3) will set errno to
> ENOMEM in that case. This fact is currently not mentioned in the
> ERRORS section of lib/
On Thu, 24 Aug 2017 16:15:48 -0300, "R0me0 ***" wrote:
> I have noticed the following behavior:
>
>
> urtwn0: device timeout
>
> The system just hangs and hard reboot is needed.
That sounds similar to this bug:
https://marc.info/?l=openbsd-bugs&m=150157788716680&w=2
Which was fixed in revisio
This is not a high quality man page but the following diff corrects
the error.
- todd
Index: lib/libc/gen/isgreater.3
===
RCS file: /cvs/src/lib/libc/gen/isgreater.3,v
retrieving revision 1.3
diff -u -p -u -r1.3 isgreater.3
--- lib/
This isn't really a bug but we can probably make the manual less
confusing. Does this help?
- todd
Index: lib/libc/gen/syslog.3
===
RCS file: /cvs/src/lib/libc/gen/syslog.3,v
retrieving revision 1.33
diff -u -p -u -r1.33 syslog.3
-
On Mon, 18 Sep 2017 16:50:41 +0200, Ingo Schwarze wrote:
> Indeed, the above code is nonsensical.
> It will write the same bytes repeatedly in case of partial writes.
OK millert@
That does look much better, though I'm not convinced that write(2)
will ever return 0 unless nbytes is also 0. POSIX
On Wed, 01 Nov 2017 14:03:13 +0100, Jeremie Courreges-Anglas wrote:
> Obviously x_init_emacs does more than than just set x_tty. For example,
> it touches kblist and AEDIT. While bypassing x_tty may run fine now
> because those variables are initialized to zero, I find this
> non-obvious and fra
This adds some missing length checks and fixes the crash.
It may just be hiding the source of the actual bug, however.
- todd
Index: usr.bin/locate/locate/fastfind.c
===
RCS file: /cvs/src/usr.bin/locate/locate/fastfind.c,v
retrievi
On Fri, 08 Dec 2017 08:58:17 -0700, "Todd C. Miller" wrote:
> This adds some missing length checks and fixes the crash.
> It may just be hiding the source of the actual bug, however.
Updated diff that adds another missing length check.
- todd
Index: usr.bin/locate/l
It turns out that locate will *always* go past the end of the buffer
due to the missing length checks. Usually this is not a problem
as mmap returns page-sized buffers. But if the length of the buffer
is an even multiple of the page size it will dereference an address
outside the buffer and crash
On Sun, 17 Dec 2017 18:11:46 +0200, Artturi Alm wrote:
> something doesn't feel right here, as manpage suggests path to be
> optional? and that first try with ~/ does report success after failing?
> didn't know any use-case for /~/ before trying this :)
Yes, the path is supposed to be optional.
On Sun, 17 Dec 2017 14:20:10 -0700, "Todd C. Miller" wrote:
> Yes, the path is supposed to be optional. A missing path should
> be treated as ".". The following diff fixes the crash and the exit
> value on error. I'll cook up a better diff with warnings later
On Thu, 21 Dec 2017 20:15:00 -0500, "Ted Unangst" wrote:
> Mark Karpilovskij wrote:
> > If only a single call to SipHash_Update is performed or if the size of
> > processed data is a multiple of sizeof(ctx->buf), this bug does nothing.
> > However when we performed several updates of various lengt
I had a similar panic today that ocurred shortly after reaching the
mclpools limit.
WARNING: mclpools limit reached; increase kern.maxclusters
em0: em_get_buf: slot 6 already has an mbuf
panic: em_rxeof: NULL mbuf in slot 7 (nrx 12, filled 128)
Starting stack trace...
panic() at panic+0x10b
em_rxe
I can no longer reproduce the panic with that diff.
- todd
On Tue, 25 Nov 2014 00:36:49 +0200, Kaspars Bankovskis wrote:
> Well, another approach..
I think it makes more sense to make it line buffered.
- todd
On Tue, 25 Nov 2014 00:53:10 +0200, Kaspars Bankovskis wrote:
> Something like this?
Please don't use setlinebuf() as it is not portable. Try:
setvbuf(ttyout, NULL, _IOLBF, 0);
instead.
- todd
On Sun, 30 Nov 2014 09:06:45 +0100, Otto Moerbeek wrote:
> Nice. I'm not 100% convinced yet this is the best fix. The actual
> problem is that the scanning for the new lowzero goes out of bounds.
> The diff below also fixes the crash for me. The question is do we need
> a -1 entry as a guard at th
Thanks, committed.
- todd
On Mon, 01 Dec 2014 08:38:26 +0100, Otto Moerbeek wrote:
> - I could not convince myself that j + from[k] in the k loop
> is properly bounded.
Isn't this exactly what the previous loop to set maxtable is for?
The only unbounded access I could find was the ++lowzero loop.
> - The change is (pa
On OpenBSD, POLLPRI and POLLRDBAND are the equivalent of exceptfds
in select(2). They are used to indicate out of band socket data.
On systems with STREAMS they may be used for other things.
- todd
I think this is more accurate.
- todd
Index: lib/libc/sys/poll.2
===
RCS file: /cvs/src/lib/libc/sys/poll.2,v
retrieving revision 1.29
diff -u -r1.29 poll.2
--- lib/libc/sys/poll.2 5 Feb 2015 02:33:09 - 1.29
+++ lib/libc/s
On Tue, 17 Feb 2015 19:34:35 -0500, Raul Miller wrote:
> So... let's say I'm writing code that uses poll.
>
> What would I need to read to discover whether POLLRDBAND is relevant
> or not, to my coding effort?
You would need to read the manual pages for the device types you
are polling. Like th
The problem is that the sdmmc driver is not present in the RAMDISK
kernel, ony in GENERIC. We have sdmmc in RAMDISK for arm and octeon.
It should probably be added to amd64 (and i386?) RAMDISK as well.
Without the device in the RAMDISK kernel you won't be able to
install.
- todd
Ah, right. I was looking at the floppy RAMDISK kernel, not the CD
one. Terry, if you start a shell instead of installing do you see
an sdmmc device in the output of the dmesg command?
- todd
Fixed, thanks. This is actually a case of the code being changed
to match the comment when the comment should have been changed to
match the code :-)
- todd
On Thu, 11 May 2017 08:32:05 +0300, appchec...@npo-echelon.ru wrote:
> Hi!� Please look [this code
> fragment](https://github.com/openbsd/src/blob/14a309e3a4feb469e2b128b77f1f214
> 4a55b1bbb/usr.bin/tic/dump_entry.c#L296):```
>
> #define FNKEY(i)(((i)<= 65 && (i)>= 75) || ((i)<= 216 && (i)>=
1 - 100 of 212 matches
Mail list logo
