Hi Mladen,
I'm fine with rolling back and committing in steps.
Will do later today. Usually I do it like that (see jk_map comits
yesterday), but it got a little late yesterday :(
This was motivated by BZ43229. We partially lost track of the many
return codes of the service() methods of all the different workers and
the helper methods in jk_ajp_common(). We test against unused return
codes, but forgot others. I hope you'll see, that the changes make
sense, after breaking them up into parts, which are easier to track.
After adding the code comments about the used return values, I'm
thinking about the reverse process, i.e.
- discussing after which stages we accept a request as recoverable by
the local worker or/and by an lb (and then check the code against that
definition)
- defining under which conditions an lb should put a worker into error
- defining, which meaning our service return codes have
- checking the bundled is_error
Regards,
Rainer
Mladen Turk wrote:
Huh, this is one huge single commit :(
I must say it's really hard to follow something like that, and now I
see that I'll for sure never do again something like that by myself.
Can we in the future try to make the commits that a single-topic
related? I know it's additional hassle, but lot less then the one
when someone really tries to understand the patch.
Regards,
Mladen.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
