Re: error_log binary unsafe

On 10/26/2013 11:54 PM, Yasuo Ohgaki wrote:

Hi Joe, On Sun, Oct 27, 2013 at 7:48 AM, Yasuo Ohgaki <yohgaki@ohgaki.net <mailto:yohgaki@ohgaki.net>> wrote:

    On Sat, Oct 26, 2013 at 2:38 PM, Joe Watkins
    <pthreads@pthreads.org <mailto:pthreads@pthreads.org>> wrote:

        Mail is not yet handled, TCP/IP is not supported any more,
        streams are binary safe.
        The SAPI and default error logging mechanism are all that
        require attention.

        The patch is not final and doesn't include a fix for every
        implementation of SAPI.

        I don't see the need for confusion ??

    Generally speaking, I'm not against making functions/features
    binary safe.

    There are many implementations of syslog/SAPI and not sure if it
    is good for all.
    It could cause BC issue also. For example, application like OSSEC
    HIDS detects
    possible intrusion by analyzing logs. Patching SAPI may break
    these applications.

    I'm not against applying your patch to master, but it's not for
    released versions.
    We needs UPGRADE note if the patch is applied.

I think it's good for master. Do you have commit karma? If not, I'm willing to merge your patch unless there are not objections. Regards, -- Yasuo Ohgaki yohgaki@ohgaki.net <mailto:yohgaki@ohgaki.net>

I don't have karma ... There are lots of SAPI's, but this patch wasn't meant to implement them all, only to provide a route whereby they can implement binary safe log_message in the shape of log_message_ex. The patch implements binsafe log for cli and cgi, do we need to implement any more ?? Cheers Joe