Skip to content
Snippets Groups Projects
Select Git revision
0 results
Search by author
  • Any Author
0 authors
  1. Feb 02, 2021
  3. Feb 01, 2021
  5. Jan 27, 2021
  7. Jan 26, 2021
  9. Jan 25, 2021
  11. Jan 22, 2021
  13. Jan 13, 2021
  15. Dec 28, 2020
    • Petr Vorel's avatar
      travis: Use Ubuntu 20.10 groovy · a7d510e4
      Petr Vorel authored 
      Eoan is failing:

E: The repository 'http://security.ubuntu.com/ubuntu

 eoan-security Release' does not have a Release file.
N: Updating from such a repository can't be done securely, and is therefore disabled by default.
N: See apt-secure(8) manpage for repository creation and user configuration details.

And 20.04 LTS focal in Travis is still fails on debconf issue
("debconf: unable to initialize frontend: Dialog")

Old 16.04 LTS xenial is still supported and working in Travis,
thus move to new groovy gives us good coverage both old and new releases.

Signed-off-by: default avatarPetr Vorel <petr.vorel@gmail.com>
      a7d510e4
  17. Nov 29, 2020
  19. Nov 17, 2020
    • Artem Leshchev's avatar
      ping: process interrupts in ping*_receive_error_msg · 2583fb77
      Artem Leshchev authored 
      
If recvmsg(..., MSG_ERRQUEUE) returned error, ping can misinterpret
EHOSTUNREACH from recvmsg in the main loop as error that happened
from his own packets and print the following message:

    ping: recvmsg: No route to host

However, this also could be an error from other ping, that was received
by this process because of RAW socket. So it's better to skip this error
and process it properly in the next cycle.

Signed-off-by: default avatarArtem Leshchev <matshch@yandex-team.ru>
      2583fb77
  21. Nov 01, 2020
    • lac-0073's avatar
      arpping: make update neighbours work again · 86ed0893
      lac-0073 authored 
      The arping is using inconsistent sender_ip_addr and target_ip_addr in
messages.  This causes the client receiving the arp message not to update
the arp table entries.

The specific performance is as follows:

There is a machine 2 with IP 10.20.30.3 configured on eth0:0 that is in the
same IP subnet as eth0.  This IP was originally used on another machine 1,
and th IP needs to be changed back to the machine 1.  When using the arping
command to announce what ethernet address has IP 10.20.30.3, the arp table
on machine 3 is not updated.

Machine 3 original arp table:

    10.20.30.3  machine 2 eth0:0    00:00:00:00:00:02
    10.20.30.2  machine 2 eth0      00:00:00:00:00:02
    10.20.30.1  machine 1 eth0      00:00:00:00:00:01

Create interface eth0:0 on machine 1, and use the arping command to send arp
packets.  Expected outcome on machine 3:

    10.20.30.3  machine 1 eth0:0    00:00:00:00:00:01
    10.20.30.2  machine 2 eth0      00:00:00:00:00:02
    10.20.30.1  machine 1 eth0      00:00:00:00:00:01

Actual results on machine 3:

    10.20.30.3  machine 2 eth0:0    00:00:00:00:00:02
    10.20.30.2  machine 2 eth0      00:00:00:00:00:02
    10.20.30.1  machine 1 eth0      00:00:00:00:00:01

Fixes: https://github.com/iputils/iputils/issues/298


Fixes: 68f12fc4
Signed-off-by: default avatarAichun Li <liaichun@huawei.com>
      86ed0893
  23. Aug 29, 2020
  25. Aug 21, 2020
  27. Aug 18, 2020
    • Petr Vorel's avatar
      doc/ping: Document decimal separator changes · e3467f62
      Petr Vorel authored 
      
Handling -i and -W parameter changed over the time, recently introducing
several *incompatible* changes:

* -i interval option
Until s20121114 ping used '.' (dot) as a decimal separator.

Then in 04d3f798 ("ping,ping6: Do not assume radix point is denoted by
'.' (-i option).") released in s20121121 decided to use decimal point
from locales (depended on $LC_NUMERIC). But even then it depend on
USE_IDN=yes which does not make much sense.

Recently d865d4c4 ("ping: Use C locale for parsing -i/-W without IDN")
addressed that and called setlocale() unconditionally. Maybe condition
based on ENABLE_NLS definition should have been discussed, but that
would probably lead just to LC_ALL=C setup in the scripts. People also
get confused by old behavior when decimal separator depends on locale
as some distros e.g. compile without IDN but other don't (see issue #273)
But that introduced *incompatible* behavior for next release, thus
documenting it in this commit.

562e0d57 fixed problems for -W, but also introduced yet another
*incompatible* behavior as invalid number is no longer fatal error (see
below).

d1961773 ("ping: Add missing preprocesses checks around setlocale()
calls") unintentionally restored the old behavior, which was fixed by
previous commit 7a90bd5 ("ping: Fix handling decimal separator for -i/-W opt").

* -W timeout option
Until s20180629 there was undocumented behavior:

1) timeout silently rounded down to the next lower integral number
2) -W N for N < 1 was rounded to 0 => infinite timeout (obviously a bug)
3) -W Ns worked as expected (sleep for N sec, as 's' was dropped,
this could be considered as a feature)
4) -W Nm -W Nh also slept for N sec instead of N min or hour (obviously
a bug)

918e824d ("ping: add support for sub-second timeouts") released in
s20190324 fixed 1) and 2) by allowing real number with locale based
decimal separator when USE_IDN=yes configured. But that introduced
*incompatible* behavior as broke 3 and 4). As it was not documented in
release notes, people complained (issue #236).

562e0d57 ("ping: allow -i and -W option arguments contain garbage input")
restored 3) and 4) support (everything including and after first
non-numeric character is dropped, warning printed).

See also: #290

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      e3467f62
    • Petr Vorel's avatar
      ping: Fix handling decimal separator for -i/-W opt · 1530bc97
      Petr Vorel authored 
      
This reverts commit d1961773.

Restore the behavior which was introduced in 918e824d. We always want to
use "." as a decimal separator for parsing -i/-W option regardless used
locale and whether IDN is used. Therefore we always need to call
strtod() with C locale. Thus we always need to set C locale for parsing
-i/-W option as was introduced in 918e824d (which brought regression:
missing include <locale.h>).

Also include <locale.h> (d1961773 included it only conditionally via
iputils_common.h and 918e824d didn't include it at all).

This effectively reverts d1961773 ("ping: Add missing preprocesses checks
around setlocale() calls") which unintentionally restored the old
behavior.

Fixes: 918e824d ("ping: add support for sub-second timeouts")
See also: #290

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      1530bc97
  29. Aug 16, 2020
  31. Aug 10, 2020
  33. Aug 05, 2020
    • Petr Vorel's avatar
      travis: Add build with disabled gettext · 272806ee
      Petr Vorel authored 
      
To cover #284 issue (fixed in previous build).

NOTE: not all travis jobs fail when built with -DUSE_GETTEXT=false,
only these are affected:
* Alpine
* CentOS 7
* Debian oldstable
* Ubuntu Xenial

Choose Ubuntu Xenial.

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      272806ee
    • Noah Meyerhans's avatar
      ping: Add missing preprocesses checks around setlocale() calls · d1961773
      Noah Meyerhans authored 
      
This fixes build with -Wimplicit-function-declaration on old distros
(it looks like to be issue on Debian oldstable, CentOS 7,
Ubuntu Xenial and on musl (Alpine), but not on recent ones):

[10/22] cc -Iping/ping.p -Iping -I../ping -I. -I.. -fdiagnostics-color=always -pipe -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Wpedantic -std=c99 -g -include config.h -include git-version.h -MD -MQ ping/ping.p/ping.c.o -MF ping/ping.p/ping.c.o.d -o ping/ping.p/ping.c.o -c ../ping/ping.c
FAILED: ping/ping.p/ping.c.o
cc -Iping/ping.p -Iping -I../ping -I. -I.. -fdiagnostics-color=always -pipe -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Wpedantic -std=c99 -g -include config.h -include git-version.h -MD -MQ ping/ping.p/ping.c.o -MF ping/ping.p/ping.c.o.d -o ping/ping.p/ping.c.o -c ../ping/ping.c
../ping/ping.c: In function ‘ping_strtod’:
../ping/ping.c:181:2: warning: implicit declaration of function ‘setlocale’ [-Wimplicit-function-declaration]
181 |  setlocale(LC_ALL, "C");
|  ^~~~~~~~~
../ping/ping.c:181:12: error: ‘LC_ALL’ undeclared (first use in this function); did you mean ‘P_ALL’?
181 |  setlocale(LC_ALL, "C");
|            ^~~~~~
|            P_ALL
../ping/ping.c:181:12: note: each undeclared identifier is reported only once for each function it appears in

Fixes: 918e824d ("ping: add support for sub-second timeouts")
Fixes: #284

Reviewed-by: default avatarPetr Vorel <pvorel@suse.cz>
Signed-off-by: default avatarNoah Meyerhans <noahm@debian.org>
[ pvorel: add missing include to have <locale.h>, extend commit message ]
Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      d1961773
    • Petr Vorel's avatar
      build.sh: Add security related CFLAGS · 61ccc0ee
      Petr Vorel authored 
      
-Wformat -Werror=format-security -Werror=implicit-function-declaration
-Werror=return-type -fno-common

Used for travis CI build, can be overridden for local builds.

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      61ccc0ee
    • Petr Vorel's avatar
      travis: Add openSUSE Leap · a041bddc
      Petr Vorel authored 
      
to have yet another stable distro.

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      a041bddc
    • Petr Vorel's avatar
      travis: Add s390x cross compile build · a87c2f0f
      Petr Vorel authored 
      
on Debian testing

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      a87c2f0f
    • Petr Vorel's avatar
      travis: Merge debian.cross-compile.*.sh · 49ade851
      Petr Vorel authored 
      
Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      49ade851
    • Petr Vorel's avatar
      travis: Temporarily use Debian unstable for aarch64 · aee49ee6
      Petr Vorel authored 
      
to fix Travis before release.

Debian testing has package dependency problems:
The following packages have unmet dependencies:
 libc6:arm64 : Depends: libgcc-s1:arm64 but it is not installable
E: Unable to correct problems, you have held broken packages.

Debian stable cannot be used because 0.49.2 fails to cross compile:

Cross C compiler: aarch64-linux-gnu-gcc (gcc 8.3.0)
Traceback (most recent call last):
  File "/usr/lib/python3/dist-packages/mesonbuild/mesonmain.py", line 112, in run
    return options.run_func(options)
...
  File "/usr/lib/python3/dist-packages/mesonbuild/environment.py", line 971, in detect_static_linker
    linker = self.cross_info.config['binaries']['ar']
KeyError: 'ar'

(Debian testing uses 0.54.3, unstable: 0.55)

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      aee49ee6
  35. Aug 04, 2020
  37. Aug 03, 2020
  39. Jul 05, 2020
  41. Jun 21, 2020
  43. Jun 01, 2020
    • Petr Vorel's avatar
      travis: Use Debian testing instead of unstable · 4bf9a5cb
      Petr Vorel authored 
      
aarch64 builds are failing due package problems on Debian:
The following packages have unmet dependencies:

libidn2-0:arm64 : Depends: libunistring2:arm64 (>= 0.9.7) but it is not going to be installed
E: Error, pkgProblemResolver::Resolve generated breaks, this may be caused by held packages.

Although these problems will be probably fixed soon, Debian testing is
more stable and still new enough.

Signed-off-by: default avatarPetr Vorel <pvorel@suse.cz>
      4bf9a5cb
  45. May 26, 2020