[jira] [Resolved] (MYNEWT-124) Log buffer issues

["Marko Kiiskila (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/MYNEWT-124?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Marko Kiiskila resolved MYNEWT-124.
-----------------------------------
    Resolution: Fixed

Made console_printf() stream out when using baselibc.

This same change cannot be made for log_printf(), given the interface to logging APIs. log_append() would need a way to append bytes one at a time to a log entry.

> Log buffer issues
> -----------------
>
>                 Key: MYNEWT-124
>                 URL: https://issues.apache.org/jira/browse/MYNEWT-124
>             Project: Mynewt
>          Issue Type: Bug
>          Components: Misc
>            Reporter: Christopher Collins
>            Assignee: Marko Kiiskila
>             Fix For: v0_8_0_rel
>
>
> When a message is written to a log, some buffers are allocated on the stack in various places in the code.  This results in two issues:
> # Undue stack-size requirements placed on any task which uses sys/log.
> # Hard limit placed on the maximum length of a log line.
> There are two functions which exhibit this behavior:
> * log_printf (LOG_PRINTF_MAX_ENTRY_LEN = 128)
> * console_vprintf (CONS_OUTPUT_MAX_LINE = 128)
> I do not have a proposed fix.  I believe Marko suggested streaming bytes to the destination, rather than writing to an intermediate buffer.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Re: [jira] [Resolved] (MYNEWT-124) Log buffer issues

[Sterling Hughes <st...@apache.org>]

On 3/30/16 3:43 PM, Marko Kiiskila (JIRA) wrote:
>
>       [ https://issues.apache.org/jira/browse/MYNEWT-124?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
>
> Marko Kiiskila resolved MYNEWT-124.
> -----------------------------------
>      Resolution: Fixed
>
> Made console_printf() stream out when using baselibc.
>
> This same change cannot be made for log_printf(), given the interface to logging APIs. log_append() would need a way to append bytes one at a time to a log entry.
>

This could be a buffer per-log that's passed in on init.

It's not great, but beats putting it on the stack.

sterling