Menu
▾
▴
middle-man-developers
|
From: Joshua N P. <jpr...@po...> - 2004-07-20 13:34:56
|
I am trying to cache all files downloaded from ftp.au.debian.org to a disk cache. Here is the relevant part of my middleman config: <profile> <enabled>true</enabled> <comment>debian sources</comment> <profiles>debian</profiles> <protocol></protocol> <host>(ftp.au.debian.org|debian.planetmirror.com)</host> <file></file> <mime></mime> <portrange></portrange> <addprofiles></addprofiles> <removeprofiles></removeprofiles> <urlcommand></urlcommand> </profile> <store> <enabled>true</enabled> <comment></comment> <profiles>debian</profiles> <path>/var/cache/middleman/debian</path> <maxdisksize>500M</maxdisksize> <diskextra>50M</diskextra> </store> This isn't working, however. For example: wget http://debian.planetmirror.com:80/debian/dists/unstable/Release When I look at mman/cache, I find this file stored in the memory cache. This might be another bug since I set the maximum size of the memory cache to zero. The other funny thing is that the mman/cache page reports that the memory cache is 2147483648% used (looks like div by zero). Anyway, what I want is to immediately cache these files on disk whether I enable the memory cache or not. Perhaps there can be a flag in the store indicating whether to use the memory cache? --=20 A new cognitive theory of emotion, http://openheartlogic.org |
|
From: Joshua N P. <jpr...@po...> - 2004-07-20 16:41:43
|
On Tue, Jul 20, 2004 at 07:04:29PM +0530, Joshua N Pritikin wrote: > Anyway, what I want is to immediately cache these files on disk > whether I enable the memory cache or not. Perhaps there can be a flag > in the store indicating whether to use the memory cache? Umm ... reading CacheSection::select_store ... I'm not sure where middleman decides whether to use the mem or disk cache but ... I suggest adding a greediness parameter for stores. If the store is greedy then the file is stored in the most greedy store. The sbalancemethod can be used to resolve ties. How does this sound? --=20 A new cognitive theory of emotion, http://openheartlogic.org |
|
From: Joshua N P. <jpr...@po...> - 2004-07-21 17:09:06
|
On Tue, Jul 20, 2004 at 10:10:35PM +0530, Joshua N Pritikin wrote: > I suggest adding a greediness parameter for stores. If the store is > greedy then the file is stored in the most greedy store. The > sbalancemethod can be used to resolve ties. Anyone home? Shall I try to create a patch for this? --=20 A new cognitive theory of emotion, http://openheartlogic.org |
|
From: Joshua N P. <jpr...@po...> - 2004-07-25 11:52:28
|
On Wed, Jul 21, 2004 at 10:37:11PM +0530, Joshua N Pritikin wrote: > Shall I try to create a patch for this? So I submitted a few patches in the SF patch tracker. I don't understand some of the code in ProfileSection::profiles_update. If connection->rheader is not NULL then you must set the MIME type to the regexp .* to get a match. This seems contrary to the documentation. I guess I'll file a bug for this too. --=20 A new cognitive theory of emotion, http://openheartlogic.org |
|
From: Jason M. <ja...@sy...> - 2004-07-25 13:12:48
|
profiles_checked is called twice, the first call connection->rheader will always be NULL, on the second it won't and only entries with a MIME pattern will be checked (i.e. the ones that weren't checked the first call) On Sunday 25 July 2004 07:52 am, Joshua N Pritikin wrote: > On Wed, Jul 21, 2004 at 10:37:11PM +0530, Joshua N Pritikin wrote: > > Shall I try to create a patch for this? > > So I submitted a few patches in the SF patch tracker. > > I don't understand some of the code in > ProfileSection::profiles_update. If connection->rheader is not NULL > then you must set the MIME type to the regexp .* to get a match. This > seems contrary to the documentation. I guess I'll file a bug for this > too. > -- Oh drat these computers, they're so naughty and so complex - I could pinch them! - Marvin the Martian |
|
From: Joshua N P. <jpr...@po...> - 2004-07-26 11:19:16
|
On Sun, Jul 25, 2004 at 09:12:43AM -0400, Jason McLaughlin wrote: > profiles_checked is called twice, the first call connection->rheader will= always be NULL, on the second > it won't and only entries with a MIME pattern will be checked (i.e. the o= nes that weren't checked the first call) Oh! I see. Hrm. What is actually happening is that connection->header->host is NULL the first time that profiles_update is called. The second time profiles_update is called, the host is set but there is no MIME pattern so the profile is skipped. This is for a regular HTTP_REQUEST via transparent proxy. I'm not sure what is the correct fix. There is a big chunk of code =66rom main.c:931 to 970. It seems like some of this code needs to happen the first time the header is parsed. If you can give me a clue then I can probably put together a patch. --=20 A new cognitive theory of emotion, http://openheartlogic.org |
Thanks for helping keep SourceForge clean.
X