I identified the culprit (did a bi-sect using git), it is:
commit 125fdf82c3df848ac8065cb04592badb62262b6d
Author: Henrik Grubbstr<C3><B6>m (Grubba) <<grubba[at]grubba.org>>
Date: Tue Jun 24 20:14:10 2008 +0000
Now uses Mapping.ShadowedMapping to implement all_constants().
Any idea where the fix goes?
Stephen R. van den Berg wrote:
>Pike versions 7.7 up till around June 23rd 2008 seem to work fine, but later
>revisions (I'm not entirely certain as to from when exactly) give me the
>following result after Roxen has already started up all the virtual servers,
>probably when it tries to write the first logfile entry (this was with
>the 7.7 snapshot of 3 days ago):
>ug)
> 2:47:24 :
> 0m 9.8s : Enabling all modules for www.softwaregids.nl...
> : All modules for www.softwaregids.nl enabled in 0.9 seconds
> :
> : Notice: Dumping failed for
>server/base_server/highlight_pike.pike (n
>ot a bug)
> : Notice: Dumping failed for server/protocols/http.pike (not a
>bug)
>Cannot encode objects without _encode.
>etc/roxen_master.pike:592: master()->MyCodec()->encode_object(roxen)
>base_server/roxen.pike:5879: roxen->compile_log_format("$ip_number
>$request-time
>/$cache-status $user [$cern_date] \"$method $full_resource $protocol\"
>$response
> $length \"$referer\" \"$user_agent_raw\"")
>base_server/roxen.pike:5694: roxen->run_log_format("$ip_number
>$request-time/$ca
>che-status $user [$cern_date] \"$method $full_resource $protocol\"
>$response $le
>ngth \"$referer\" \"$user_agent_raw\"",roxen->LogFile()->write,,,1)
>etc.
>Where should I look to fix this? In Pike, or in Roxen?
>Any hints where?
--
Sincerely,
Stephen R. van den Berg.
Real programmers don't produce results, they return exit codes.
|