#foswiki 2016-07-12,Tue

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
gac410Adding code to check HTTP_X_FORWARDED_HOST fixes bootstrap. Also probably need to set ForceDefaultUrlHost to prevent unsafe redirect errors.
Nope... Login always redirects to localhost. :(
[00:33]
............ (idle for 56mn)
Okay, the bug is that the LoginManger "packs" the original URL without regard to the forceDefaultUrl setting [01:31]
Looks like the bug is in master & Release02x01 as well. We just don't use proxies much. [01:42]
..................................... (idle for 3h4mn)
***gac410 has left [04:46]
................... (idle for 1h32mn)
ChanServ sets mode: +o MichaelDaum [06:18]
...... (idle for 27mn)
ChanServ sets mode: +o CDot [06:45]
......................................................................... (idle for 6h4mn)
ChanServ sets mode: +o Lynnwood
ChanServ sets mode: +o gac410
[12:49]
...... (idle for 29mn)
gac410Hi MichaelDaum, vrurg, Foswiki 3 under Plack/PSGI seems to be working okay. Apache proxy was a bit of a pain, and bootstrap is messed up, but it's working for me.
I think I have a fix for apache bootstrap, no idea if it would also apply to nginx. Have to check HTTP_X_FORWARDED_HOST instead of HTTP_HOST. when generating DefaultUrlHost
[13:20]
vrurgHi gac410, that sounds great! jomo was right on this.
nginx is highly configurable in this respect.
It is possible to define what headers would be passed on and what will they contain.
[13:22]
gac410And there is a bug in LoginManager/TemplateLogin sub _packRequest It packs the URI which means foswiki after login redirects to the proxied backend localhost:5000 not the original url [13:24]
MichaelDaumHi there [13:24]
gac410Can't use the URI there, we have to reconstruct the url and substitute the DefaultUrlHost [13:24]
MichaelDaumbeen trying to figure out the best way to configure a production environment for psgi
painful
Perlbal is abandoned
fastest and most flexible one seems to be uWSGI
[13:24]
gac410ugh... only one backend possible? The apache proxy can do balancing. No idea if that applies? [13:26]
MichaelDaumuwsgi has been integrated nicely into ubuntu
alas
uwsgi_psgi doesn't work together with a local perl (plenv or perlbrew)
so you have to compile it yourself
bummers
[13:26]
gac410yech [13:27]
MichaelDaumdistros are far far behind current perl and uwsgi [13:28]
vrurgMichaelDaum: may I ask you? If there is any significant information to share it in a topic? I'm moving today, have no normal connection till Friday. It's a big luck I happened to be here at this moment. :( [13:28]
MichaelDaumso uwsgi - while being the most advanced architecture - is a disappointment [13:28]
gac410y, the perl distro situation is really poor. But IMO perl made it difficult themselves by deprecating code and making backwards compatibility a disaster. [13:29]
MichaelDaumvrurg, I am still in research mode
best would be to read the irc logs or use slack
anyway. let me report what I've found out so far...
uwsgi is faster than starman, mostly due to the fact that it does not use http to speak to the backend.
[13:29]
gac410I would like to create a page somewhere with the Apache (and others contributed...) configurations for Foswiki 3.0 [13:30]
MichaelDauminstead of mod_proxy you'd use mod_uwsgi ... or whatever it is called in apache land
this is the high performance protocol for a whole lot of backend types, one of which is psgi for perl
[13:31]
gac410Centos will have a usable version in the year 2026 :D [13:32]
MichaelDaumthere even is a so called emperor mode where uwsgi watches a directory of ini files, one for each app, and then spawn workers for it. [13:32]
gac410nice [13:32]
MichaelDaum"emperor" controlling "vassals"
http://uwsgi-docs.readthedocs.io/en/latest/Emperor.html
so then you are supposed to create a foswiki.ini file that configures the sockets and workers and env and psgi app and what not to spawn
sounds simple but ... meh ... I've been going up and down the docs ... docu on ini files is lacking, let alone good example.inis
then the problems start
no SCRIPT_NAME, no PATH_INFO ... you have to tell uwsgi to manage them
done so
[13:33]
gac410sounds like it needs some time to stabilize / mature [13:35]
MichaelDaumnow I get lots of errors that I do _not_ get when firing up foswiki.psgi under starman
related to parts of $Foswiki::cfg not being defined
Use of uninitialized value in list assignment at .../core/lib/Foswiki/Users/BaseUserMapping.pm line 114.
this is $Foswiki::cfg{SuperAdminGroup} not being defined
however LSC has indeed been loaded before successfully ... as it seems
... which renders uwsgi a dead end finally, unfortunately.
[13:36]
.... (idle for 16mn)
gac410I really think the focus now for 3.0 needs to be on completing the work, stabilizing it, documenting it and enlisting some more developers. No more major restructuring unless really critical.
I wonder if a email to the announce list discussing plans for Foswiki 3.0 and asking for assistance would be of any value. Maybe banners on github, and f.o as well.
[13:55]
MichaelDaumy
and we should rename the branch to foswiki3
[14:04]
gac410y. I'd still like to do a merge from master -> foswiki3 or whaterver we call it. I ran one a few times. The conflicts are a pain, but most are pretty simple to resolve.
That would catch up the back level jquery plugin, etc.
I hesitated to actually commit the merge because I didn't have a stable test platform, and vrurg had so much "in process"
now that I have psgi working, and vrurg is otherwise occupied, I should be able to merge. Except now I'm pretty busy too :(
[14:05]
MichaelDaumsame same [14:07]
gac410I'd be a lot more comfortable if we had a few more devs participating in the release meetings. [14:09]
MichaelDaumUndefined subroutine &Foswiki::Store::encode called at /home/www-data/foswiki_distro/core/lib/Foswiki/Store/Rcs/Handler.pm line 69. [14:23]
FoswikiBothttps://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Store [14:23]
MichaelDaumfoswiki3 [14:23]
LynnwoodI'm in the market for VPS hosting once again for a client. Anyone have a VPS host they are very happy with.
?
[14:26]
MichaelDaumLynnwood, hi, try https://www.netcup.eu/ [14:27]
Lynnwoodgood ol' DH continues to raise in the ranks of top-reviewed large providers.
@MichaelDaum - thanks
[14:27]
gac410MichaelDaum: vrurg had asked me to post errors needing resolution to https://foswiki.org/Tasks/Item13897#CRITICAL_BUGS_47PROBLEMS [14:28]
MichaelDaumit is the one we use for foswiki.org as well [14:28]
LynnwoodI'm also meaning to take a closer look at ngix as alternative to apache. You seem to prefer it... [14:28]
gac410That's probably easier than creating tasks for the issues you are finding. [14:28]
MichaelDaumI have to stop here playing with psgi. [14:29]
gac410, I can't log in to foswiki with an RcsWrap store impl
no clue why
[14:34]
gac410y, I guess that most testing has been done with default plain file. I've not tried it either. [14:35]
MichaelDaumafter login it still redirects th http://localhost even though I've enabled {ForceDefaultUrlHost} and am accessing the site via https://localhost ... result "unsafe redirect" [14:36]
gac410Right. I understand that bug. [14:37]
MichaelDaumthis all feels pre-alpha [14:37]
gac410That bug is in master / r21 as well.
But yes, definitely pre alpha. it needs a lot of polishing.
[14:37]
MichaelDaumthe url host doesn't make it thru to the backend down the proxy_pass [14:37]
gac410TemplateLogin::_packRequest packs the request->uri() which does not pick up the forced url [14:38]
MichaelDaumoic [14:38]
gac410That would fail with a proxy on R21 as well. Broken
I'm surprised nobody has opened a task on it. Though probably not a lot of proxy use.
or proxy combined with template login
[14:38]
....... (idle for 33mn)
Lynnwood@gac410 - I had an interesting exchange with MichaelDaum_ this morning regarding the issue I had yesterday with strikeone.
It was simply that (somehow) I was running afoul of strikeone. He provided code that fixed the simpliest case but it remains a mystery to me how to do intervene in form submittal without tripping up strikeone.
For example, I have a simple form that I would like to do two things: 1) automatically submit it when selecting a select option and 2) submitting the form via ajax.
I can make both of those individual things work without tripping up strikeone but if I combine them, I do fail strikeone. :-(
very frustrating.
I would love to _really_ understand what one has to do to not run afoul of strikeone so I can code accordingly.
I'd like to write this up in support as I'm sure others run into it.
I wonder if I could simply include the strikeone call in my submit function...
probably not...
[15:12]
.................... (idle for 1h35mn)
jastin principle that should be possible
I don't think I've done it myself, though
[16:54]
Lynnwoodgrrrr.... it's driving me crazy. I can't get it work any way I've tried.
maybe i'll look at the strikeone.js code to see if I can glimmer any ideas there.
I just have to say it: strikeone does not play well with other js calls.
[17:05]
unfortunately, i can't make much sense of how it processes the Strikeone.submit(this) call. [17:13]
...... (idle for 29mn)
gac410Take a look also at CommentPlugiin and SubscribePlugin ... they play pretty well with strikeone [17:42]
Lynnwoodok
i just figured a way to hack it... but i'm not at all happy with it.
[17:43]
gac410never argue with success :D [17:44]
LynnwoodLike I say, I could get the ajax submit to work ok but when I tried to add an "onchange" to the select element (either directly in the html or via jquery), it would fail.
I finally included a hidden "submit" button in the form and use the onchange to trigger click on button.
uuugggly
[17:44]
well... at least it works... and as you say don't argue with success. [17:58]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)