The fix is waiting for integration on the h/parse-number-32bit-bigendian-fix branch. The first patch actually fixes the issue, the 2nd adds a testcase to avoid future regressions. It should easy be easy to apply it to 3.5. Thanks for the report. On Fri, Mar 7, 2014 at 5:16 PM, Balazs Scheidler <bazsi77@gmail.com> wrote:
I didn't seem to be right, rather it's a more important bug. All LL_NUMBER tokens are parsed into a gint64 value through a glong * pointer. So all 32 bit big endian platforms seem to be affected.
... a patch is coming soonish.
On Fri, Mar 7, 2014 at 5:13 PM, Balazs Scheidler <bazsi77@gmail.com>wrote:
hmm... I think I may have a potential issue at hand, the function parse_number_with_suffix() returns its result using a long pointer, and time_reopen is just a int. Can it happen that you might use a big-endian platform where the size of a long and a int is different?
Do you have compilation warnings while compiling syslog-ng?
On Wed, Mar 5, 2014 at 11:24 PM, Tamayo, Andres <Andres.Tamayo@viasat.com
wrote:
Hello,
I have found a solution to the problem (syslog-ng OSE 3.5.1 is not accepting global options) but I still do not understand what is the cause of the problem.
First some background information. I have built syslog-ng into a buildroot toolchain and flashed into an embedded development board. When I was prototyping my syslog-ng configuration file in the host computer, I was able to use configuration like this:
time_reopen(10);
However, when I started to use syslog-ng on the embedded target, the configuration previously mentioned did not work. If I included this line in my syslog-ng.conf file; syslog-ng would automatically assume time_reopen was equal to zero and show debug messages like this one:
syslog connection failed; fd='8', server='AF_INET(192.168.1.10:1901)', error='Connection refused (146)', time_reopen='0'
If I deleted that configuration line from the configuration file, syslog-ng would use its default value. This is the message I would see if the configuration line was removed:
syslog connection failed; fd='8', server='AF_INET(192.168.1.10:1901)', error='Connection refused (146)', time_reopen='60'
But what if I needed to use time_reopen=10 instead.
Anyway, as I said in the beginning I do now know what is causing this behavior but I found a solution.....hex notation. If I use the following configuration line instead:
time_reopen(0xa);
Then, I would see the following syslog-ng debug message:
syslog connection failed; fd='8', server='AF_INET(192.168.1.10:1901)', error='Connection refused (146)', time_reopen='10'
The same is true for the following configuration items:
flush_timeout
log_fetch_limit
log_iw_size
log_fifo_size
max-connections
There may be other configuration items that exhibit the same behavior.
If anybody can explain why this is happening I would appreciate an explanation, thank you.
I hope this helps.
*From:* Tamayo, Andres [mailto:Andres.Tamayo@viasat.com<Andres.Tamayo@viasat.com>]
*Sent:* Friday, February 28, 2014 3:49 PM *To:* 'syslog-ng@lists.balabit.hu' *Subject:* [syslog-ng] syslog-ng OSE 3.5.1 is not accepting global options
Hello,
I am experiencing problems with syslog-ng OSE 3.5.1 reading the configuration file and accepting its global options. I had problems with 2 global options specified in my syslog-ng.conf file:
@version: 3.5
@include "scl.conf"
options
{
flush_lines(0);
time_reopen(10);
log_fifo_size(1000);
use_dns(no);
use_fqdn(yes);
keep_hostname(no);
keep_timestamp(yes);
chain_hostnames(no);
perm(0640);
owner("root");
group("root");
create_dirs(yes);
dir_perm(0750);
dir_owner("root");
dir_group("root");
};
source s_cache
{
file("/path/to/some/file.txt", flags(no-parse));
};
destination d_network
{
syslog("192.168.1.10" port(1901) transport("tcp"));
};
log
{
source(s_cache);
destination(d_network);
};
The time_reopen and log_fifo_size seem to be forced to zero if I try to specify alternate values for them as shown above. If I do not include them in the configuration file, syslog-ng uses their default values time_reopen=60 and log_fifo_size=10000.
Does anybody know what may be causing this problem? The log_fifo_size default value is OK but I need a faster time_reopen window and I cannot find a way to specify it. If I do, syslog-ng forces it to time_reopen=0. I can tell because I get message like this when it is trying to reconnect:
Syslog connection failed; fd='8', server='AF_INET(192.168.1.10:1901)', error='Connection refused (146)', time_reopen='0'
If I remove the time_reopen(10) global option from the syslog-ng.conf file, then I get the following message when syslog-ng it is trying to reconnect:
Syslog connection failed; fd='8', server='AF_INET(192.168.1.10:1901)', error='Connection refused (146)', time_reopen='60'
Please let me know if you have any questions for me.
Thank you for your help.
______________________________________________________________________________ Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng Documentation: http://www.balabit.com/support/documentation/?product=syslog-ng FAQ: http://www.balabit.com/wiki/syslog-ng-faq
-- Bazsi
-- Bazsi
-- Bazsi