On 03/10/2018 02.17, Matthias Andree wrote:
> Am 03.10.18 um 00:42 schrieb Carlos E. R.:
>> On 02/10/2018 21.23, Kamil Jońca wrote:
>>> "Carlos E. R." <> writes:
>>>
>>>> On 02/10/2018 20.03, Kamil Jońca wrote:
>>>>> "Carlos E. R." <> writes:
>>>>>
>>>>>> On 01/10/2018 20.09, Kamil Jońca wrote:
>>>>>>> Matthias Andree <mat...@gm...> writes:
>>>>>>>
>>>>>>> [..]
>>>>>>>> The consideration is whether this could go into 6.4, or would have to
>>>>>>>> wait for 7.0. The log file format is not really specified, so changing
>>>>>>>> it isn't breaking any documented interface, just breaking past practice,
>>>>>>>> so it might cause some astonishment. Opinions?
>>>>>>> Regardless of these questions I would as NOT to change logging via
>>>>>>> syslog.
>>>>>> Timestamping via syslog is done by the syslog daemon, not by the
>>>>>> application that sends the messages.
>>>> Don't you understand? Fetchmail can not change logging via syslog.
>>> Please correct me if I am wrong.
>>> 1. So far fetchmail produces "log messages"
>>> 2. these messages are identical, in case when logging to file/stderr or
>>> syslog.
>>> 3. syslog adds timestamp to messages received from fetchmail
>>> 4. now we want to extend tetchmail log message with timestamps.
>>> 5. so syslog will get messages with timestamps, and adds its own
>>> timestamp.
>>> 6. so, after adding timestamp to fetchmail messages and logging to
>>> syslog we will have TWO timestamps in logs. I think this is ugly, and I
>>> kindly ask Mathhias, not to change messages when someone pick "set
>>> syslog" option.
>> Well, I have never seen a programmer make that error in a released
>> version. I'm sure they know. The message is formed, sent to syslog, or
>> some extra entries are added, like a timestapmp, then printed (or
>> written to file).
>>
>>
> To calm your concerns,
>
> 1. I thank Kamil for the hint to make sure we don't bluntly apply
> timestamping all over the map, to avoid double logging in syslog;
>
> 2. I thank Carlos for the trust. Carlos, rest assured that oversights
> and mistakes do happen, and if the code that formats the message is
> shared between syslog and fetchmail code, that concern of Kamil's is
> indeed rather substantiated. And the formatting itself _is_ shared code.
> O:-) See report.c for the function report(). Hans's patch, however, only
> changed the file output, not the syslog output, and generated an
> additional system call, and would not change the log buffer, so syslog
> output would remain unadorned with the new logfile time stamp.
Welcome :-)
Without looking at the code, I was doing educated guesses. For instance,
from my own small programs I know that if I goof at that, I will notice
the error fast enough on testing it - and yes, errors do happen, but we
all do tests for that reason ;-)
--
Cheers / Saludos,
Carlos E. R.
(from openSUSE 15.0 (Legolas))
|
