[asterisk-bugs] [JIRA] (ASTERISK-25179) CDR(billsec, f) and CDR(duration, f) report incorrect values

[Gianluca Merlo (JIRA)]

     [ https://issues.asterisk.org/jira/browse/ASTERISK-25179?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Gianluca Merlo updated ASTERISK-25179:
--------------------------------------

    Status: Waiting for Feedback  (was: Waiting for Feedback)

Hello Rusty,

I apologize, my explanation was probably a bit chaotic, and led you to misunderstand my observations.

I do not expect CDR(billsec,f) to return a value in milliseconds, but it seems it is returning a floating point value that is a thousandth of the correct value. E.g. if cdr_custom.conf includes

{noformat}
[mappings]
durations.csv => ${CSV_QUOTE(${CDR(billsec)})},${CSV_QUOTE(${CDR(duration)})},${CSV_QUOTE(${CDR(billsec,f)})},${CSV_QUOTE(${CDR(duration,f)})}
{noformat}

the produced durations.csv file contains (again, as an example)

{noformat}
"7","30","0.007000","0.030000"
{noformat}

I hope that this clarifies the issue and helps you reproduce my findings. My initial description included an attempt to explain the portion of code that I deem related to the issue.

> CDR(billsec,f) and CDR(duration,f) report incorrect values
> ----------------------------------------------------------
>
>                 Key: ASTERISK-25179
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-25179
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Functions/func_cdr
>    Affects Versions: 13.4.0
>            Reporter: Gianluca Merlo
>            Assignee: Gianluca Merlo
>
> Hello,
> I noticed, while reviewing CSV CDRs created with cdr_custom, that while CDR(billsec) and CDR(duration) report correct values, the corresponding high precision values CDR(billsec,f) and CDR(duration,f) are wrong. More specifically these values seem to be a thousandth of their non-high-precision counterpart (e.g. if CDR(billsec) is 37, CDR(billsec,f) is 0.037000 ).
> It seems to me that the code in funcs/func_cdr.c gets this value via ast_cdr_getvar() (from main/cdr.c), and assumes that the returned value is in milliseconds, this dividing the value by 1000. The problem appears to be that ast_cdr_getvar() gets the value via cdr_object_format_property() and thus via two specialized functions
> * cdr_object_get_duration()
> * cdr_object_get_billsec()
> which both return the value in seconds. General assumptions seem to indicate that internal representation of these values should be considered in milliseconds. I hope you can verify my guesses, and provide a consistent fix.
> As a side note, I can confirm that the issue was not present in the 1.8 series, and that the affected portions of code have been reorganized since then. In that version, values were returned in milliseconds, and the CDR() functions handled them properly.



--
This message was sent by Atlassian JIRA
(v6.2#6252)