+1

On Nov 12, 2013, at 10:10 PM, William A. Rowe Jr. <[email protected]> wrote:

> On Wed, 13 Nov 2013 00:07:08 +0200
> Graham Leggett <[email protected]> wrote:
> 
>> On 13 Nov 2013, at 12:00 AM, "William A. Rowe Jr."
>> <[email protected]> wrote:
>> 
>>> Follow-up question; is reuse recommended?  In this small bit of
>>> trunk (comments removed for simplicity);
>>> 
>>> -            else if (!lenp) {
>>> +            else if (f->r->proxyreq == PROXYREQ_RESPONSE) {
>>>                ap_log_rerror(
>>> -                        APLOG_MARK, APLOG_INFO, 0, f->r,
>>> APLOGNO(01585) "Unknown Transfer-Encoding: %s", tenc);
>>> -                return APR_ENOTIMPL;
>>> +                        APLOG_MARK, APLOG_INFO, 0, f->r,
>>> APLOGNO(01586) "Unknown Transfer-Encoding: %s; using
>>> read-until-close", tenc);
>>> +                tenc = NULL;
>>>            }
>>>            else {
>>>                ap_log_rerror(
>>> -                        APLOG_MARK, APLOG_WARNING, 0, f->r,
>>> APLOGNO(01586) "Unknown Transfer-Encoding: %s; using
>>> Content-Length", tenc);
>>> -                tenc = NULL;
>>> +                        APLOG_MARK, APLOG_INFO, 0, f->r,
>>> APLOGNO(01585) "Unknown Transfer-Encoding: %s", tenc);
>>> +                return APR_EGENERAL;
>>> 
>>> Indeed the second case becomes effectively the initial case, a
>>> T-E/C-L combination we throw away and refuse to handle.  Although
>>> this is very subtly different - originally we gave up without a
>>> C-L, now we cough without or without a C-L if this is not a
>>> proxyreq response body.
>>> 
>>> But the revised first case, 1586 'Reading to end', is indeed quite
>>> different than the original second case, 1586 'Using C-L'.  Should
>>> this have been assigned a fresh number?  Otherwise, we end up with
>>> divergent meanings, and the user who stumbles across 1586 will come
>>> up with two very different cases searching the web for their error.
>>> 
>>> So should the second have been assigned a new number?  Or both of
>>> cases assigned new numbers?  Or is reuse acceptable?
>> 
>> I'd say probably not (ie assign a new number) - if the message
>> description has changed materially and means something different from
>> before, the number should probably change to reflect that. Otherwise
>> google searches might return confusing results mixing "before change"
>> and "after change" messages with the same number.
> 
> On your feedback, I've picked up a new APERRNO for the second issue above.
> I'm still thinking the first is an edge case, but since it is the same
> general sort of error with identical error text, it doesn't seem all that
> harmful to persist the old APERRNO.

Reply via email to