#foswiki 2012-08-01,Wed

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

WhoWhatWhen
gac410SvenDowideit: regarding IPC::Run - I was a bit concerned about using Sandbox in configure, With my IPC::Run testing, it worked really well, and was very easy to apply. Sandbox it more complex, creates temp files, etc. [00:29]
SvenDowideiti'm more concerned about adding new stuff in 1.1
and then in 1.2, to avoid duplication
if IPC::Run is a more modern, more reliable thing, then I'd prefer to replace Sandbox's run stuff with it
ie, either we believe that our sandbox code is good
or we don't
either way, we should only use one impl
(i'd rather IPC::Run)
but i sure don't like the idea of 1.1.6 using new libs
[00:29]
gac410Agreed. on that. Foswiki::Net calls sendmail directly, no sandbox. IPC::Run is really easy: require IPC::Run; IPC::Run::run( \@cmd, \$text, \$out, \$err ), priceless. [00:31]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Net [00:31]
SvenDowideitneeds testing on bad windows too
cygwin&so on
[00:31]
gac410Yeah agree. I can test WinXP - hm. don't have a cygwin env. any more. :( IPC::Run claims good windows compat. [00:32]
SvenDowideitwe know that Sandbox was tested and has worked for years on a really massive number of platforms
windows isn't enough - i needs to be known good on all the variations
(for it to be considered (and hopefully rejected) on 1.1
[00:32]
gac410It also is really horrendous. It has code that you can never execute because of tests that can never branch. [00:33]
SvenDowideitfor 1.2.0 we can risk it more [00:33]
gac410Well ... :P
:)
[00:34]
SvenDowideitit took over a year to stabelise Sandbox
with lots of screaming
because it was wrtten by linux people that hoped they knew enough windows
[00:34]
gac410I can understand that. It has some horrendously ugly code that from what I could tell, can't ever work in some conditions. [00:34]
SvenDowideitSvenDowideit hasn't ever felt omnipotent enough about that code [00:35]
gac410Maybe it is just cruft that has crept in over time. But there was one glaring conditional code that could never be reached because the variable being tested was just set to a constant before the test.
and I didn't know enough or have platforms to test or consider trying to change it.
[00:36]
SvenDowideitor its code that wasn't ready, but ... its a hairy ball
basically, i would reject the proposal to use IPC::Run in 1.1.6, but would be thrilled to replace the Sandbox mess in 1.2
and the sooner the better
so we have a little time to test
[00:36]
gac410Well I didn't intend to try to replace Sandbox. Only capture the output from sendmail. [00:38]
SvenDowideitwe don't want 2 times the cruft [00:38]
gac410I tried and tried, all sorts of captures, locals, Babar tweaks, nothing would capture stderr, and IPC::Run "just worked" [00:38]
SvenDowideitso proposing to add IPC::Run === replace the Sanbox run code [00:38]
gac410My current code (in stash) says. If IPC::Run available, use it, otherwise use the open =| pipe to sendmail
which if you read the sandbox comments, I suspect is not really cross-platform either.
I was originally going to insert the optional IPC::Run code, and if a user had sendmail dieing, suggest they install IPC::Run from cpan to get better diagnostics.
It was really annoying that I could not get STDERR to capture. Spent a day or two on it with help, but never got anywhere.
I'll update the proposal to defer to 1.2. Though that says we still have no way to debug sendmail. :(
[00:40]
SvenDowideityou say that as though this was a new and terrible change from the last 10 years of fantastic debugability
which i suspect is overdramatising the loss and underplaying the risks of adding dulpicate code cruft
which given who often we add something promising to refactor it in the next release (and then don't) is not great
[00:47]
gac410It just jumps out as one of our really weak points. I don't care about sandbox, or refactoring. I wanted to address the users like Alexis, and several others who spent hours trying to debug broken email. [00:49]
SvenDowideiton the other hand, do you care about adding more code we can't debug? [00:49]
gac410It really doesn't matter that much. I'll just defer to 1.2, not worth the energy. Adding code we can't debug? I don't understand, not worth the effort to get me to understand.
We've got too many bigger issues to solve.
[00:51]
SvenDowideiti really can't understand
the big issue is no-one grokes Sandbox
and you have pointed out a good replacement candidate
[00:51]
gac410I agree with that. and Refactoring sandbox is a really good 1.2 thing.
I thought dipping toes in water for sendmail might be a good start.
[00:52]
SvenDowideitadding IPC:Run in 1.1 will give everyone the excuse to just leave it there
as we all have done many times before
[00:52]
gac410That sure sounds like a great reason to do nothing then :P [00:53]
SvenDowideitbest eg == the __THREE__ diff engines in core
each time someone added a plugin, they said they'd refactor it next big release
and nope, they never did
[00:53]
gac410IIRC, I actually tried to use Sandbox btw, in my sendmail fighting, and even that didn't capture stderr. It's been too many months since I gave up on sandbox. Ah... also sandbox will not accept a full command
You have to use the template tokens to build the command.
[00:55]
SvenDowideitwhich means Sandbox needs fixing
mind you, there's a reason why it uses TOKENS
[00:56]
gac410To use sendmail, you have to pipe in the complete email message. [00:56]
SvenDowideitsomething to do with param injection [00:56]
gac410Right. Sendmail just didn't fit what sandbox does. Not sure if trying to replace Sandbox with IPC::Run is the right solution.
Since we probably still want the token parsing.
[00:57]
SvenDowideithuh?
sandbox would wrap IPC::Run
[00:58]
gac410okay. In which case, Net would stlll have to use IPC::Run directly, since sandbox can't pipe in the email. [00:58]
SvenDowideitunless Sandbox was appropriated added to
Sandbox is intended to be our one and only interface to the world
its not as though its hard to write a new method
[00:59]
gac410Okay, so next part. It also proposed adding File;:Which so that the bin/configure checker could discover where rcs is installed. Improves odds of successful installation on BSD variants.
That doesn't touch the core operational code, Is that a defer to 1.2 as well?
[01:04]
SvenDowideiti seriously thought you wanted to make 1.1.6 a proper, real patch release
not a 'well i didnt' touch the code, so i can add XYZ
[01:05]
gac410It IS a patch release. It don't install on OSX or FreeBSD. It's busted. Unless you go poking at the config. [01:05]
SvenDowideitSvenDowideit backs away to let you RM, as you seem to have very different ideas about patch [01:06]
gac410It's a users bug report. configure botches rcs configuration on osx and bsd [01:06]
SvenDowideitso what did it previously do?
as it used to work on osx - i've fixed that before
[01:07]
gac410Nothing. It doesn't work out of the box.
Right. You have to enable expert settings and change the path on all the rcs commands.
[01:07]
SvenDowideitit was definitly working around 2005-6
as i made it work
[01:07]
gac410You can make it work. It doesn't work out of the box. [01:07]
SvenDowideitand it was automatic
no, it worked out of the box
[01:07]
gac410Ah. Probably just selected rcsLite? [01:08]
SvenDowideitno
back then rcslite was too risky to sell
and i did this for a pretty big installer job
back when i built installers for osx, rpm, solaris, windows and a could of others
i'm only saying that it is possible to fix the bug without adding new modules
[01:08]
gac410on FreeBSD it's egrep and fgrep, /usr/bin/grep vs. /bin/grep on linux. it's not rcs. :( Brain fade here. Anyway. Yes I can write my own parse, scan, and find the location of grep on the path. [01:14]
SvenDowideitwhy?
ie, why write your own, it must already be in svn
tbh, i think it was just some simple conditional -e stuff in the TWiki.cfg
which would probly be enough for a patch release
given that this magic was used before we had configure at all
[01:14]
gac410I thought that the philosophy would be to use cpan rather than invent our own stuff. I would much rather put a well known tested CPAN module in than try to write my own stuff. I got my A** handed to me when i wrote my own email address recognizer. [01:17]
SvenDowideiti would agree, in a non=patch release
SvenDowideit gives up and lets RM RM
[01:17]
gac410gac410 gives up too. I'll rm and not fix bugs. [01:18]
................ (idle for 1h18mn)
pharveywiki! drowning! [02:36]
gac410huh? [02:37]
pharveydon't mind me, our wiki is just getting hammered
in other news... greetings!
[02:37]
SvenDowideitgo look at my blog post
basically, turn off SESSION cookies for guest users
and then use the lack of that session cookie to show simpler html :)
where simpler html == the html files that the page cache makes
[02:38]
pharveypage cache is ruinous, or is the new one better? [02:40]
SvenDowideitthe new one is better for ~36 hours [02:40]
pharveyincidentally, our server has crashed last month and this month at exactly midnight GMT - Foswiki is unable to rotate our 80MB+ events.log inside of the 30 second fcgid timeout [02:40]
SvenDowideitwhich is enough time to fix stuff
merge tim's task stuff
then use it to rotate logs
SvenDowideit is currently fighting simple html&css
[02:40]
pharveyI think I want an option to stop automatic log rotation, or do we already have that option [02:41]
SvenDowideitpass [02:41]
gac410yeah. CDot's replacement plainfile logger is a disaster with big logs. [02:41]
SvenDowideiti think y
pharvey, would you like a mongodb logger :p
[02:41]
pharveythe fix is to login and run ./view manually from CLI, then it's happy again. [02:42]
SvenDowideitor a DBI logger [02:42]
pharveypharvey wonders why his Foswiki logs can't be rotated from lograte.d like everything else [02:42]
SvenDowideitSvenDowideit is prodding Log::Dispatch [02:42]
pharveylogrotate* [02:42]
SvenDowideitthey can
but we don't rely on its existance
[02:42]
gac410dbi would be nice, but being able to tail the logs is a big plus [02:42]
pharveyyeah, I prefer text in this case.. poor-man-with-leprosy stats = grep foo |wc, haha. [02:43]
SvenDowideiti like syslog :) [02:43]
gac410The old method of writing the current logfile with yyyymm in the filename was much faster. [02:43]
SvenDowideitfaster, but buggier? [02:43]
pharveyhow does foswiki.org installation handle it? [02:43]
SvenDowideitSvenDowideit has no idea why cdot changed it :) [02:43]
gac410Crash [02:43]
pharveyso foswiki.org crashes every month? [02:43]
gac410Well this time it made it. It doesn't usualy crash for us though. The one transaction crashes, but the events.log file just keeps getting appended. Once it has a current timestamp it doesn't try to rotate again.
To force a rotate at f.o I had to touch the *.log file with a --date from the prior month
[02:44]
pharveysomehow trin.org.au ends up with 6 view requests apparently all trying to rotate the same logfile (or maybe they're just hanging waiting for a lock) [02:45]
gac410No locking. :( [02:46]
pharvey(I have max_procs 6) [02:46]
gac410Basically the rotate is a free-for-all disaster. [02:46]
pharveyI assume Foswiki is digging through the events.log trying to find the end-of-month boundary so it can be truncated? Except no further requests are logged, because they're all busy trying to rotate when that happens...
argh
raising an urgent task, then.
Does this apply to Release01x01 too?
[02:46]
gac410Yes.
I think there might already be a task. though maybe not.
[02:46]
pharveynothing under http://foswiki.org/Tasks/FoswikiLogger [02:48]
gac410Nope, no task, right
For 1.1.4, it was even worse because the rotate would crash if the logfile contained an embedded newline in a record.
I fixed that for 1.1.5. It reads the whole logfile into memory, and then writes it out to the yyyymm file(s). If one is found existing, it dies.
So multiple tasks writing, first one who touches a yymmmm file, it will break all the others when they try to write the file.
(I fixed the embedded newline issue. I had nothing to do with the read it all in to memory to split the file idea.
[02:48]
pharveytried to describe the problem at Item12022 [02:54]
FoswikiBothttp://foswiki.org/Tasks/Item12022 [ Item12022: Silently rotating foswiki logs inside a view request breaks busy sites ] [02:54]
gac410Agree it needs an urgent task. CDot's rationale iirc, was that plainfile was just a sample and someone ought to write a real logger or something like that.
http://irclogs.foswiki.org/bin/irclogger_log/foswiki?date=2012-03-01,Thu&sel=481#l477
[02:54]
pharveywhat's involved in writing a syslog impl? [02:56]
SvenDowideitpharvey, not much [02:56]
gac410Probably a CPAN module :D [02:56]
SvenDowideitLog::Dispatch is a nice one
and i found something else for the Logger/listener thing
[02:57]
pharveyI'm never very confident messing with logrotate.d
I'm wondering if we're needlessly wasting CPU cycles in the rotate procedure... but I assume we're not (!)
[02:57]
SvenDowideitwe are [02:59]
gac410Log::Dispatch looks very nice Sven. Wooeee... You can log to email. That would be ... um.. voluminous [02:59]
SvenDowideitif its 'read it all in to memory' as gac410 says [02:59]
gac410Just for the once a month rotate. [03:00]
SvenDowideity, thats really a crap thing todo [03:00]
gac410It sure does. I fought with that for hours trying to get foswiki.org to rotate [03:00]
SvenDowideitthere's no need to read into memory to rename a file ffs [03:00]
gac410It "splits" it for some reason. [03:00]
SvenDowideitwhat i don't like about L::D is that its totally fixed log levels
bummer, we don't have treeplugin on f.o
[03:00]
gac410Hm. Log::Dispatch has log levels or I don't understand your comment.
$log->alert('Strange data in incoming request');
[03:03]
SvenDowideityes, _fixed_ log levels [03:03]
gac410oh. So you mean you can't log "fred" leve.
l
[03:04]
SvenDowideity
and its string based
with a hint that structured doc logging is possible
[03:04]
gac410yah. has a warning about calling with utf8 characters [03:04]
SvenDowideitbut i've not tested to see what i can do
the restrictions make sense for lowest common denominator
but its a pest for what i have in mind
SvenDowideit is going to try to use it, or something abstracting it to replace the listeners
as i want to avoid releasing them :(
but for today, I'm struggling with superfish, loaded from treeview
works well, its just that it wraps
[03:04]
gac410I assume CDot's change from events.yyyymm as the "current" log to use events.log - was to always have a consistent current log. But the rotate requirement is a disaster. [03:06]
pharveysuperfish is a pain to get right [03:06]
SvenDowideitand un-not needed
there are better ways to log to the right file :(
so, i can select $('li')
and i can $('li:first')
but how can i select the last 3 li's
mmm, even that will be a pest :(
[03:06]
.... (idle for 16mn)
gac410hmpf events.log rotated fine. debug.log and error.log both failed to rotate.
The saving grace for existing sites is they have probably stayed on the Compatibility logger
[03:25]
SvenDowideitargh!@
can't use superfish in the sidebar
overflow:hidden and all sorts of gruble
[03:27]
gac410Well, Log::Dispatch looks really good at first & second glance. Log::Dispatch::File::Stamped - does what our Compatibility logger does, Log::Dispatch::FileRotate does what PlainFile does. and you get the added things like DBI, Email, Syslog, etc.
There is even a Win32 even log logger, and a Jabber logger.
[03:34]
SvenDowideitand a mongodb loger [03:36]
gac410Log::Dispatch::Win32EventLog [03:36]
SvenDowideity
thats why i was heading towards it
for 1.2
[03:36]
gac410Seems like making it LogDispatch as a pluggable logger in parallel with Compat and Plainfile would be good. Could even be an extension rather than core, since loggers are pluggable.
So for 1.1.6, Tell people to use Compat logger on busy sites?
[03:37]
SvenDowideitfor 1.1 yes [03:38]
gac410Compat logger is probably safe, avoids all the rotate issues. [03:39]
SvenDowideiti'm actually wondering if we can ship it in1.2
y
didn't we tell people not to use the other before?
[03:39]
gac410which one - compat?
There is a warning message saying you should convert. But that was probably bogus advice.
hm. Choosing Compat doesn't fix it. You would need to edit LocalSite.cfg and add in the old log settings :(
[03:39]
..... (idle for 20mn)
SvenDowideitargh!
1.1.5 tinymce combins sequential verbtims
[04:03]
gac410I think it's always done that. [04:04]
SvenDowideitits never done it to me before :/
there were 4 empty lines between them
[04:04]
gac410It's intentional, because tmce creates new verbatimes when you hit a newline
er. 1.1.5 or the TMCE released on Extensions web?. Actually it's WysiwygPlugin, not TMCE.
There is explicit code in Wysiwyg to combine verbatims. One fix I made for the latest Wysiwyg was to not combine them if they have different classes.
[04:05]
pharveynext time a get a couple of weeks off I'm doing Foswiki::DOM :P [04:08]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::DOM [04:08]
gac410One "bug" I fixed which I now regret. There were several open tasks to treat links as "real" links in Wysiwyg mode - ie lighting the link icon. But that means that if you change a wikiword, you also need to click the link button and change the link to it. [04:12]
SvenDowideitpharvey, doom on you, doom on you
er, DOM on you
[04:12]
gac410Just noticed that. So that behavior sucks. [04:12]
pharveyok. I'm going to poke the log rotation code. Must be a better way than slurping the file into memory [04:14]
gac410Very good. It really needs some help. [04:15]
pharveyhrm. At first glance, it doesn't seem to be slurping the whole file.
oh, yes it is. Silly me.
[04:16]
gac410I was going to say... did I loose my mind? [04:17]
pharveyIt's incrementally concatenating a string. I was looking for something else :} [04:17]
gac410If a line is not a complete log record it really doesn't like it. I tried to hack around that. [04:19]
pharveyI guess I should make a pathological testcase. [04:20]
.... (idle for 16mn)
SvenDowideitmmm, well, I maybe should update the fatwilly theme to allow the use of treeplgin to generate a superfish menu
it looks pretty slick
and then make the slicksitemap work on 1.1, and we get a magic menu driven by parent relationships that ar drag&drop modifyable
worse, I just re-implemented JQueryMenuSkin using one single topic :(
[04:36]
pharveyfun-tastic [04:41]
SvenDowideitnite :) [04:41]
........... (idle for 51mn)
GithubBot[foswiki] foswiki pushed 3 new commits to Release01x01: http://git.io/9bvP0A
[foswiki/Release01x01] Item12022: Address perlcriticisms - PaulHarvey
[foswiki/Release01x01] Item12022: Remaining bareword opendir - PaulHarvey
[foswiki/Release01x01] Item12022: Add a test to measure log rotation perf - PaulHarvey
[05:32]
***GithubBot has left [05:32]
FoswikiBothttp://foswiki.org/Tasks/Item12022 [ Item12022: Silently rotating foswiki logs inside a view request breaks busy sites ] [05:32]
GithubBot[foswiki] foswiki pushed 3 new commits to master: http://git.io/Fq1c3Q
[foswiki/master] Item12022: Address perlcriticisms - PaulHarvey
[foswiki/master] Item12022: Remaining bareword opendir - PaulHarvey
[foswiki/master] Item12022: Add a test to measure log rotation perf - PaulHarvey
[05:33]
***GithubBot has left [05:33]
pharveyoh dear. The biggest killer in the log rotation is... wait for it... drum roll: Foswiki::Time::formatTime [05:34]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Time [05:34]
pharveyFoswiki::Time::parseTime is high up there too [05:34]
SvenDowideitwhy would log rotation need to call that more than twice?
is the numpty parsing every line?
[05:34]
pharveyindeed
but wait
[05:35]
SvenDowideitffs [05:35]
pharveyno, hang on
I NYTProf'd the whole TestRunner.pl
which includes first logging 15,000 times to the logger API
so I'm probably most likely quite wrong there
so, false alarm. Don't panic :P
[05:35]
SvenDowideitok, real q
are you going to do Log::Dispatch?
[05:36]
pharveyI was hoping for a cheap fix in the existing code
after setting up my perf test, I'm finding some quick hacks in the existing _rotate code isn't showing much promise
oh well, I hope the perf test is useful to someone
[05:36]
SvenDowideitum
ok, this is mad
you have the old log open
and when you detect that you want a new file
you (used to be able to) log the new entry to a new tmp file
then rename the old log
oh, wait, we're not holding stuff open are we
y, dumping to debug.log is pooey
oh well
[05:37]
pharveyoh, wait
I WAS RIGHT THE FIRST TIME
oh, capslock
my $eventTime = Foswiki::Time::parseTime( $event[1] );
[05:39]
SvenDowideitsnigger [05:40]
pharveythis is done on every line processed in _rotate [05:40]
SvenDowideitchange that to a simple regex match on the day date
and you should have a nice speedup
with little effort
SvenDowideit is hoping that you'll one-up me with an even faster one :)
[05:40]
pharveyhrm. I wonderz if the date format is fixed
that's probably why somebody used parseTime in the first place
[05:41]
SvenDowideitno, but you don't care
use formatTime to make the regex for you
or use the first line to make it for you
(after testing that its the right day)
that way you only parse/format once (or twice)
oh dear, Log::Dispatch will be hrd to use to _get_ the logs back
[05:42]
pharveywell, using my $eventTime = 1; brings the perf from 0.440275 to 0.142348 :}
I'm still a bit uneasy about how multiple processes trying to rotate the same logfile at the same time is actually... handled
[05:44]
SvenDowideiti'm not surprised
SvenDowideit gives up on today a little, still sick
[05:44]
pharveyget well sooner [05:45]
.......... (idle for 46mn)
GithubBot[foswiki] foswiki pushed 2 new commits to master: http://git.io/3qNvUg
[foswiki/master] Item11998: tell the user if Empty* is missing too (could be git per Contrib problem, or could be I'm running from foswiki/core dir.. - SvenDowideit
[foswiki/master] Item11998: don't add single quotes around the name of the extension wherever we're replacing the Empty* name (pm's everywhere) that breaks everything - SvenDowideit
[06:31]
***GithubBot has left [06:31]
FoswikiBothttp://foswiki.org/Tasks/Item11998 [ Item11998: create new extension missing MANIFEST in EmptyContrib? ] [06:31]
...... (idle for 25mn)
SvenDowideitWEEEEEEEE
and now foswiki logs to syslog
[06:56]
pharveyw00t :)
I'm doing an if ($event =~ /^\d{4}-(\d{2})-\d{2}T[0-9:]+Z\b/) {} else { ...parseTime ...} hack and it's much faster
[07:11]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/ZpmWOA
[foswiki/master] Item12023: Log:Dispatch based logger which can log to DBI, Syslog, file, MongoDB, email, twitter.... - SvenDowideit
[07:16]
***GithubBot has left [07:16]
FoswikiBothttp://foswiki.org/Tasks/Item12023 [ Item12023: write a Log::Distch based Logger ] [07:16]
pharveyWell, not that much faster. 0.443983 vs 0.11991 [07:16]
CDot£9.5bn from the public purse, and not a single gold medal to show for it. Pfffft.
pharvey: that's 4 times faster! That's quite a difference
[07:18]
pharveyyep, I think I'll commit it
fundamentally, I still need to do something different
I don't understand how the PlainFile logger handles my situation where 5 different processes are all trying to rotate the logs all at once
thankfully fcgid kills them all after 30s so none of them complete, but still
£9.5bn - that's a pretty big splash
Sydney 2000 was the reason I stopped watching television. I was so utterly saturated with Olympics This, Torch Bearer that. It was so horribly flogged to death, you could see it in the reporters' eyes they had it even worse, being forced to generate so much airtime leading up to the actual olympics, haha.
[07:19]
CDotI made sure to avoid as much olympic coverage as I could before it actually started, so I'm not saturated yet; but getting pretty sick of it already, I must admit. [07:32]
pharveyyeah, I'm sure I would have been more interested if I hadn't been over-saturated before it even started [07:33]
CDotFundamentally PlainFileLogger is the simplest thing that might work; I wouldn't have thought it handles multiple rotations at all well. [07:33]
pharveymaybe SvenDowideit will commit is syslogger and I can switch to that :P [07:34]
GithubBot[foswiki] foswiki pushed 2 new commits to Release01x01: http://git.io/JamkYQ
[foswiki/Release01x01] Item12022: avoid parseTime in PF logger _rotate() - PaulHarvey
[foswiki/Release01x01] Item12022: Update comments on test - PaulHarvey
[07:46]
***GithubBot has left [07:46]
FoswikiBothttp://foswiki.org/Tasks/Item12022 [ Item12022: Silently rotating foswiki logs inside a view request breaks busy sites ] [07:46]
GithubBot[foswiki] foswiki pushed 2 new commits to master: http://git.io/GQxEBQ
[foswiki/master] Item12022: avoid parseTime in PF logger _rotate() - PaulHarvey
[foswiki/master] Item12022: Update comments on test - PaulHarvey
[07:47]
***GithubBot has left [07:47]
pharveyCDot: If you have any further wisdom on Item12022, feel free [07:47]
..... (idle for 23mn)
CDotpharvey: I commented. The current code is very dumb. [08:10]
.......... (idle for 47mn)
SvenDowideitweird
going to watch the cycling and the mtb at sydney2000 olymipcs was fun
and well wroth it
i recon its mad to not go to your local(ish) event
just for pharvey - the syslog logger _is_ commited
i'll clean it up more over time, but LogDispatchContrib should already be semi-functional
(it logs to foswiki/bin/:)
and to syslog
and to stderr
with different filters
mmm, pharvey shall i paste that again for you, or are you readinlogs
[08:57]
pharveypharvey read logs [09:02]
SvenDowideitits really nice that i have an suexec foswiki set up with LogDispatchContrib and it logs straight to syslog [09:02]
pharveyooh, ta :)
hrm, do I dare add some locking shenanigans to PlainFileLogger
[09:02]
SvenDowideitno
:p
add a tiny amount of code to LogDispatchContrib
and make it use Log::Dispatch::File::Locking
and make a feature req to add LogDispatchContrib to core for 1.2
and Log::Dispatch >:}
[09:04]
pharveyI would, but I'm just wondering if there's some more low-hanging fruit in our core logging impl [09:06]
SvenDowideiti wonder if we can re-do foswiki buildcontrib etc for 1.2 to use local:lib and some PAR like thing [09:06]
pharveypharvey reviews htaccess to see how it's done [09:06]
SvenDowideitthere will be lots
but Log::Dispatch appears pretty freaking low hanging too
i'm pretty pissed that i didn't just write it years ago when i put it on my tasks likst
SvenDowideit goes to make dinner :)
[09:06]
pharveyI will definitely want to try to switch to Log::Dispatch, I'm just mending-as-I-go [09:07]
SvenDowideitlaters [09:07]
pharveythanks sven, seeya
pharvey digs through pidgin settings, looking to kill off smilies >:/
Log::Dispatch
yay
[09:07]
CDotSvenDowideit: it wouldn't be mad if there was any chance of tickets. I got nada in the ticket lottery. [09:17]
***ChanServ sets mode: +o MichaelDaum [09:26]
.......................... (idle for 2h6mn)
SvenDowideitCDot, we didn't get any in the lottery
we went to the things that were sold afterwards, because the seats were empty on the day
and to things like the road race, where we just went
and the paraolympic opening / closing
and mmmm, just walking around the venues without going in
as sydney was a nice open area
very very nice atmosphere
[11:32]
....... (idle for 33mn)
AlexanderSthi there, is there any know issue about using iwatch to trigger solrindex? I have a virtualhosts environment and i include a web with a symbolic link. changes in the linked web do not trigger the solrindexer properly
i played around with the iwatch.xml and it seams that the first <path> entry is not used.
[12:06]
jastI don't think iwatch follows symlinks [12:11]
AlexanderSti think iwatch does, because if i watch a symlinked web it does work, if it´s not the first path entry [12:13]
jasthmm, interesting
my bad. following symlinks actually seems to be the default in inotify calls.
[12:14]
AlexanderStbut how is "following a symlink" treated in iwatch? do i maybe have to define different watchlists?
shit...it´s the same with different watchlists..only the last entry is taken into account if you have a symlinked web
MichaelDaum, do you have experience with iwatch and symlinked webs?
[12:18]
ModAcOstYes, Iwatch will only trigger one event
You will need to trigger two solrjobs for the directory
[12:23]
AlexanderStpuh...
after all i think i will watch the whole vhosts directory an trigger the solrindex job with the host=all preference
[12:25]
ModAcOstDo that, I will do it proper later [12:28]
MichaelDaumAlexanderSt, I have got a path entry for the symlinked data directory. so no, I am not relying on symlinks.
only this one is a host=all job then
[12:29]
AlexanderStthat is what i just did...but i´m watching the whole data dir of the wiki with some webs who are symlinked...
merci bien
MichaelDaum, there is one other question...who do you handle to restart? iwatch if you create a new web?
[12:32]
MichaelDaummy first path in iwatch is /.../foswiki/trunk/core/data, then comes the vhosts like /.../vhosts/localhost/data
the vhosts are symlinking /home/www-data/vhosts/localhost/data/[System, _SomeTemplates, Applications]
the while /.../foswiki/trunk/core/data triggers a host=all solrjob, the others are per vhost only
restart iwatch?
you mean when creating a new vhost
[12:37]
AlexanderStMichaelDaum, i also watch vhosts/global/data e.g. . if the data dir get´s a new subdir (a new web is created), it´s not getting watched by iwatch.
thats the behaviour i can see here
[12:46]
MichaelDaumah okay
nother cronjob: service iwatch restart
well this definitely makes things more and more uggly
[12:47]
AlexanderSti agree :-) [12:48]
ModAcOstYou could let iwatch watch for created,isdir and have iwatch restart itself [12:49]
MichaelDaumhm just tested it
mkdir data/System/test ... perfectly triggers a new solrjob
as does rmdir data/System/test
this is iwatch-0.2.2 on ubuntu 12.04
so basically I can't repro the observation AlexanderSt
[12:49]
AlexanderStahhhg [12:51]
MichaelDaumdid you test a plain manual mkdir?
inside an iwatched dir
[12:51]
AlexanderSti think i have the reason...
maybe events="close_write" is not right?
[12:52]
MichaelDaumtry pastebin for your iwatch.xml [12:53]
ModAcOstIf you write a txt file it will trigger a close_write, that should be fine [12:54]
AlexanderStyes i did test it yet with mkdir
i cannot pastebin the iwatch.xml since i´m sitting behind the fence...
[12:55]
ModAcOstThe config will not respond to simple directory-creation, it will only watch for .txt files [12:56]
AlexanderSt...not with my computer of course ;) [12:56]
MichaelDaumthen your iwatch.xml is too restrictive wrt the events it reports [12:58]
ModAcOstI don't see any problem with it. If you create a new web you put sth. in it and thats a txt file. [12:59]
AlexanderSt<path type="recursive" events="close_write" alert="off" syslog="on" filter="\.txt$" exec="bla">path</path>
even without filter it´s not working...
[12:59]
MichaelDaumtry without the events param [13:04]
.......... (idle for 49mn)
and? [13:53]
AlexanderStMichaelDaum...no even without event praram it´s not working
...maybe it´s because we built iwatch on the server on our own...
[13:55]
MichaelDaumAlexanderSt, rhles or sles ;( [13:59]
AlexanderStrh
5.7 :(
[13:59]
MichaelDaumcorps dont get it that those two are doomed for any serious web dev [14:00]
AlexanderStbut just got the same issue on a debian virtualhosting server [14:00]
MichaelDaumwith those "enterprise" editions you always end up knitting your own binaries, or drown in searching for rpms on the net trying to resolve the never eneding dependencies between packages. [14:01]
ModAcOstIf i leave out the filter and events it does honor new webs, but with them it will just ignore them until iwatch is restarted [14:02]
MichaelDaumrecently a rh support hotline supposed to install your own local perl in parallel to the official rpms ... no kidding
just to get a perl fit for unicode
gawd, why oh why should I use _any_ distro at all then
[14:02]
AlexanderSt:) [14:04]
MichaelDauminsisting on rhles or sles is like insisting on ie7
they will never ever get something build on top of html5 user media (http://neave.com/webcam/html5/)
anyway. enuf for today. football time. see ya.
[14:04]
ModAcOstMichaelDaum, are you the devil? You have a talent describing my pains... [14:06]
alis there a specific reason that the debian packages from fosiki.com are not yet included with debian?
anything one could help with?
[14:13]
jastal: go talk to debian folks if you want to have it in debian. the main challenge is finding someone who knows both foswiki and debian packaging and is willing to maintain the packages. [14:19]
laenMmm, hows foswiki related to twiki?\ [14:19]
jastlaen: foswiki is a fork of twiki (with virtually all of the original developers having sided with foswiki) that was created when the original creator of twiki decided to close it off and turn commercial [14:20]
laenOw, okay. Which kinda explains why the twiki channel is empty ;). [14:20]
jastI believe even wikipedia has a section about it [14:20]
laenI was hoping to get some pointers on replication, even though i barely have any experience with either. [14:21]
aljast, ok, thanks [14:23]
jastessentially the answer is probably "copy files around" ;)
hardly the cleanest possible thing to do, but by far the easiest
[14:23]
laenWell, i was thinking about rsync, but i'm kinda surprised to how easy it's setup -- if i understand it correctly.
It would just require both pub/ and data/, or is there more that needs copying?
(It only needs to be read only on the other side).
[14:24]
jastnot unless you have something very special [14:25]
CDotno, that's enough [14:25]
laenOkay so on the other side, i basically just need a simple bare TWiki install. [14:25]
CDot.... or Foswiki [14:26]
laenWell, that'd be my preference.
But i didn't make the descision, there already is a TWiki at the current company here. But i'll certainly bring FosWiki to the attention of the company here.
[14:26]
CDotUnless you have used any incompatible features of TWiki, you should be able to use a Foswiki on the same data as the TWiki runs. Mind you, I'm not sure if there have been any significant features developed in TWiki since the fork. [14:33]
laenI doubt it. [14:34]
Henry78Hi folks! I've got a weird problem: A blank login page. Anybody seen this already? I'm trying to find the root cause for hours...
I't a quite new foswiki installation, that worked fine some weeks ago. Now I got a call from a user that just gets a blank page when trying to login. I'm excluding apache/server configuration 'cause I don't get any output when trying: 'cd foswiki/bin; perl -T login"
[14:37]
***Henry78 has left [14:40]
rfreytagI am having a little trouble installing LatexModePlugin - would someone here have time to troubleshoot? I am running Foswiki-1.1.3, build 11475 and I just downloaded LatexModulePlugin from foswiki.org.
all the components are owned by the web user (in my case www-data).
no errors or warnings coming from <server>/bin/configure at all.
(all the right people appear to be here but perhaps I should ask this question somewhere else - would you prefer I put it on the foswiki support pages?)
[14:52]
jastnothing wrong with asking here, I believe, but people may take some time to respond
Henry78: anything useful in working/logs/error.log?
rfreytag: same question :)
[14:57]
rfreytagthanks - understood.
you trying to install LatexModulePlugin too? heck of a coincidence.
[14:57]
jastno, different problem, same question :} [14:58]
Henry78no, nothin in logs, neither foswiki nor apache [14:58]
jastmy next best guess is that some template got clobbered
anyway, I have to leave for half an hour or so
[14:58]
Henry78I removed all custom templates, the others weren't changed since it worked the last time
no luck
[14:59]
jastdid anything special happen recently? power failure, system update, etc. [15:01]
ModAcOstIf apache-user cannot write to working/logs/error.log you get a similar behaviour. [15:01]
Henry78a colleague changed the apache config to SSL, but I reverted any changes; and as told above: bin/login doesn't spit out anything when run from command line
all logs (working/logs/*) have rw for www-data
[15:02]
....... (idle for 31mn)
: [15:34]
.... (idle for 15mn)
AlexanderStexit [15:49]
AlexisHazellArgh. Why are jquery-1.7.min.js /and/ jquery-1.7.1.js both getting loaded? i think it's causing a datepicker call to fail. [15:56]
........ (idle for 35mn)
laenSo i hear you're all spreading fud. [16:31]
CDotlaen: are we? howso? [16:36]
laenPeterThoeny> fyi, the twiki community has to deal with a hostile fork that is spreading lots of false rumors, like "twiki renamed to foswiki" etc PeterThoeny> this is fud PeterThoeny> i am the primary developer with the post checkins
I didn't make it up ;).
[16:36]
CDoterm, I don't recall anyone saying twiki was renamed to foswiki. It wasn't, twiki was forked. [16:37]
laenThat's what he said, not me. [16:38]
CDotas for "lots of false rumors" without knowing what he's talking about, that's hard to contest [16:38]
Henry78I'll give up for today, login page is still blank...I'll check in here tomorrow or so. [16:38]
CDotbut as far as most of us are concerned, twiki is history, and we don't really care *what* they say or do. [16:39]
Henry78Thank all for the hints! [16:39]
laenIt's just a bit funny and sad at the same time, i'm not chosing sides. He's now making quick phone calls about it i believe. [16:39]
CDotI wonder who he's calling? Your boss, probably. [16:39]
laenMy only concern was that i wanted some easy support, and #foswiki won ;).
My boss? :P Why would he do that.
[16:40]
CDotwell, you are welcome to continue getting support here, even if you continue to use twiki. [16:40]
laenCDot: that sounds kinda personal. I really don't care what the company wishes to use, i am tasked with having to work with it. I wouldn't chose Twiki personally.
Peter is now providing me with all kinda URL's that prove it's all fud. Not sure why.. you guys already answered my question heh.
[16:41]
CDotheh; well, have a fun surf. [16:42]
laenThaaaanks! I'll hang around. [16:42]
CDotoh, I just realised you might have misinterpreted what I said. What I meant is that we don't have any problem with twiki users getting support from here. Or at least I don't. But the degree of support we can offer is limited to areas where the codebase hasn't shifted too much (in either project). [16:46]
AndreU_as we all now, TWiki is not moving anywhere codewise
docs and icons of TWiki are a moving target thought
[16:52]
......................... (idle for 2h0mn)
rfreytagok - I found my problem with LatexModePlugin - I updated the wiki with this simple instruction for those similarly in need of a clue (present company excepted): "Finally, enable LatexModePlugin: go to the configure interface. under the 'Extensions'->'EnabledExtensions' tabs find the checkbox (probably unchecked) next to {Plugins}{LatexModePlugin}{Enabled}. Check the box and save and then LatexModePlugin should properly display its examples on you [18:53]
............ (idle for 58mn)
***al has left [19:51]

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