On 2/12/2013 2:41 PM, Ben Johnson wrote: > > > On 2/11/2013 3:00 PM, Ben Johnson wrote: >> >> >> On 2/11/2013 2:38 PM, Nick Kew wrote: >>> >>> On 11 Feb 2013, at 18:43, Ben Johnson wrote: >>> >>>> The documentation provides a sample configuration snippet ( >>>> http://httpd.apache.org/docs/current/mod/mod_authn_socache.html ): >>> >>> Whoops! You found a documentation bug. >>> >>> AuthnCacheSOCache can't go in a <Directory>, because >>> it determines things that happen at server startup (initialising >>> the cache), when <Directory> or equivalent is meaningless. >>> >> >> Thanks, Nick. I submitted a note on the relevant documentation page. >> >> Might you have insight regarding the second half of my post (the module >> faults and thread termination)? It seems like an unusual problem. >> >> -Ben > > After digging-around in the source code, I'm left wondering if this > FIXME has any relevance: > >>From http://fossies.org/dox/httpd-2.4.3/mod__authn__socache_8c_source.html > > 371 /* error: give up and pass the buck */ > 372 /* FIXME: getting this for NOTFOUND - prolly a bug in mod_socache */ > 373 ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r, APLOGNO(01686) > 374 "Error accessing authentication cache"); > 375 return AUTH_USER_NOT_FOUND; > > Does the FIXME note mean that the message "Error accessing > authentication cache" is returned (erroneously) when the authentication > cache file doesn't exist at all? If so, then I suppose that doesn't > apply here, because the file definitely exists. > > When I start Apache with the following configuration line present > > AuthnCacheSOCache dbm > > the process creates two files in the "log" directory: > socache-dbm-authn-socache.dir and socache-dbm-authn-socache.pag. Both > file are zero-bytes in length until I make a request that requires > authentication, at which time the .pag file is populated with what looks > to be correct/valid information: > > [a bunch of null-bytes and > such]{SHA}H4tRn3FmQ32/zfGNp52eeApuyyA=web_user:ben > > (hash changed for privacy reasons) > > So, it's not as though the files can't be created or don't exist. > Further, I'm starting Apache with Administrator credentials, to > eliminate obscure permission issues. > > It bears mention that this is not the first time I've run into > concurrency/locking problems with Apache on Windows. In fact, to this > day, a similar issue plagues my stack: my log is full of message like this: > > [core:error] [pid 9140:tid 1152] (OS 32)The process cannot access the > file because it is being used by another process. : [client > fe80::1ddc:6806:70b6:8546:59306] AH00127: Cannot map GET > /project/images/banner_products.jpg HTTP/1.1 to file, referer: > https://ben-pc/project/products/ > > I can't help but wonder if the same problem is at the root of both symptoms. > > In AH00127's case, I have posted about this in numerous places and have > not received a single response, e.g.: > > http://www.apachelounge.com/viewtopic.php?p=23176 > > https://bugs.php.net/bug.php?id=63413 > > If nobody is able to shed light on this issue, I'll open a bug report. > > Thanks, > > -Ben
A full week has elapsed and the note that I submitted regarding this issue has yet to appear at http://httpd.apache.org/docs/current/mod/mod_authn_socache.html . Am I being impatient, or am I taking time out of my day to submit feedback to a black hole? In any event, while I still intend to file a bug report regarding this issue, I did find a simple workaround (thanks to user glsmith at Apache Lounge): use shmcb instead of dbm when on Windows. -Ben --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
