[Kamailio-Devel] [Fwd: Re: [Serdev] logging framework changes]

[Daniel-Constantin Mierla]

It is interesting indeed and could be helpful for debugging.

Is siplog (I see #include <siplog.h>) a library ?

Cheers,
Daniel
 

On 10/14/08 12:44, Klaus Darilion wrote:
> might be interesting for kamailio too (as option)
>
> regards
> klaus
>
> -------- Original-Nachricht --------
> Betreff: Re: [Serdev] logging framework changes
> Datum: Mon, 13 Oct 2008 14:53:51 -0700
> Von: Maxim Sobolev <sobomax at sippysoft.com>
> Organisation: Sippy Software, Inc.
> An: Ondrej Martinek <ondra at iptel.org>
> CC: serdev at lists.iptel.org
> Referenzen: <48F3A5C8.5070306 at iptel.org>
>
> Good stuff, Ondrej. What we have found very useful over the years is
> adding some reference to the SIP messages in question (such as Call-Id)
> to the log output. Then you can easily correlate particular call flow
> with the errors/warnings that it has caused. Perhaps it's something you
> should consider adding at least as an opt-in:
>
> That's the code we use in our private version to do it:
>
> --- parser/msg_parser.c.orig
> +++ parser/msg_parser.c
> @@ -43,6 +43,8 @@
>    #include <string.h>
>    #include <stdlib.h>
>
> +#include <siplog.h>
> +
>    #include "../comp_defs.h"
>    #include "msg_parser.h"
>    #include "parser_f.h"
> @@ -626,6 +628,16 @@
>           DBG("exiting parse_msg\n");
>    #endif
>
> +       parse_headers(msg, HDR_CALLID_F, 0);
> +       if (msg->callid != NULL && msg->callid->body.len > 0) {
> +               tmp = alloca(msg->callid->body.len + 1);
> +               memcpy(tmp, msg->callid->body.s, msg->callid->body.len);
> +               tmp[msg->callid->body.len] = '\0';
> +               msg->log = siplog_open("ser", tmp, 0);
> +       } else {
> +               msg->log = NULL;
> +       }
> +
>           return 0;
>
>    error:
> @@ -652,6 +664,7 @@
>    /*only the content*/
>    void free_sip_msg(struct sip_msg* msg)
>    {
> +       if (msg->log) siplog_close(msg->log);
>           if (msg->new_uri.s) { pkg_free(msg->new_uri.s);
> msg->new_uri.len=0; }
>           if (msg->dst_uri.s) { pkg_free(msg->dst_uri.s);
> msg->dst_uri.len=0; }
>           if (msg->headers)     free_hdr_field_lst(msg->headers);
>
>
> Ondrej Martinek wrote:
>   
>> Hi guys,
>>
>> I've prepared a patch to SER logging framework that _automatically_ 
>> prefix log messages with the log level and module name strings.  In 
>> short, you no longer need to explicitly include the log level or module 
>> name string in the message.
>>
>> Eg.:
>>    Instead of:
>>       LOG(L_ERR, "tm: ERROR: something failed\n");
>>    you will write:
>>       ERR("something failed\n");
>>
>> The new stderr/syslog format is:
>>
>>    LEVEL: MODULE [FILE:LINE]: MESSAGE
>>    LEVEL: MODULE: MESSAGE  ... if NO_DEBUG macro is defined
>>
>> MODULE is "<core>" for the core and "<script>" for the ser.cfg generated 
>> log messages.
>>
>> Another feature is that LOG(L_...) macro is superseded by the 
>> corresponding ERR/DBG/...() shorter equivalents (of course, it's still 
>> used when the first argument isn't constant).
>>
>> The main reason was to unify the log messages and enable their further 
>> processing by text tools (filtering, colorification, you-name-it,...).
>>
>> Since many log messages need to be updated I created a perl script that 
>> fixes most of them in one shot:
>>   - any log level and module name prefix is removed
>>   - LOG(L_ERR/DBG/etc., ...) is replaced by shorter ERR/DBG/etc.() macros
>>   - fix the messages where the log level and the prefix mismatch:
>>
>>   Eg.: ERR vs. BUG
>>        LOG(L_ERR, "BUG: tm: foo\n")
>>     becomes:
>>        BUG("foo\n");
>>
>>
>> Note that because of this script the patch will touch a lot of files in 
>> many modules.  Please feel free to ask about details, suggest or object 
>> to the changes.  I plan to commit it soon if there will be no serious 
>> objections.
>>     
>
>
>
>   

-- 
Daniel-Constantin Mierla
http://www.asipto.com