#foswiki 2012-09-18,Tue

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

WhoWhatWhen
pharveymoo [00:53]
gac410howdy pharvey [00:54]
pharveyhi [00:54]
............................ (idle for 2h17mn)
SvenDowideit: worth mentioning the new Mongo 2.2 locking situation in jira? Or were we exploiting a "bug" against older versions? Seems annoying
Oooh, TTL Collections - nice.
SvenDowideit: maybe there's a way we could do collection-per-web rather than db-per-web... muhahaha!
[03:11]
.......... (idle for 46mn)
***gac410 has left [04:02]
SvenDowideitpharvey i don't think we care
end of lifeing mongo for foswiki is probly more useful
[04:14]
pharveythinking more short-term [04:15]
SvenDowideitfundamentally, the mongo design is that if you hit a query for which you have not optimised, and you have a large dataset, return zero results
that fundamentally makes it impossible to work around
collection per web doesn't change the issues
[04:15]
pharveycollection-per-web might work around the db-lock problem [04:17]
SvenDowideiti can check, but is there anything in 2.2 that we really need?
tbh, zero results for de-optimised queries is the big design flaw that if i had understood / or even expected, would have made me discard it
SvenDowideit will need to blog that, but may not as its hard to write without calling people rude things
[04:17]
pharveywell, they keep telling me that you just need to add .limit() on to the end of your .sort() and you can magically sort as many non-indexed records as you like
'course, it's hard to know what the .limit() should be unless all your objects are the same damn size
[04:20]
SvenDowideitmmm
plus when you limit and skip, it falls over in s screaming heal
[04:21]
.............................. (idle for 2h27mn)
***ChanServ sets mode: +o MichaelDaum [06:49]
MichaelDaumSvenDowideit, and performing a .limit() _after_ a .sort() won't cut it with regards to performance: why sort the rest of a result set when you already found the first n top hits.
sort-filter-skip-limit shouldn't be separated for performance reasons. and databases normally don't.
[06:51]
pharveyin MongoDB, sorting a limited resultset (without indexes; which is the use-case giving us problems) only sorts the first N hits, you need the entire set sorted before you return the first N.
but it might all be moot if Sven can whip DBIStore into shape over the weekend :P
The limitation that results.sort().limit() works-around is that, if you don't have indexes turned on for the attribute being sorted, then you need your response to fit into a single mongo/BSON object (max size: 4 or 16MB, depending on version).
The idea is that you use .skip()/.limit() to truncate the resulting BSON object down to something less than its maximum allowable size
and this is really really crappy, but whenever I've complained that perhaps Mongo should be smarter about this to avoid the extra guesswork involved (how many results shall I fit inside my BSON object today? depends on the data!) you just get insulted for not thinking webscale-enough...
#mongodb, almost as bad as #perl
[07:04]
flexibeastOoh, that reminds me of a very sarcastic clip about "webscale" databases .... [07:14]
pharveyIf it's the clip I'm thinking of, I found it more annoying than funny :P [07:15]
flexibeastSynthesized voices + swearing? [07:15]
pharveyindeed [07:16]
flexibeastThat's the one. :-) i appreciated the underlying sentiment, if nothing else. :-) [07:16]
pharveytrue
not that we need mongo's webscale-ness, we just wanted convenient ad-hoc querying of JSON-ified data
it just sucks that their major use-cases all assume you're planning and tuning your indexes... which means you've got a reasonably stable schema to query against
which flies in the face of "schemaless"
[07:16]
flexibeast*nod* [07:19]
PeterSvendsenHi, I would like to upgrade my jquery plugin... but the server dont have access to the internet..... how can this then be done? Where to download the plugin? [07:24]
pharveywhich version of Foswiki do you have?
Latest JQueryPlugin is at http://foswiki.org/Extensions/JQueryPlugin (scroll to bottom)
[07:25]
PeterSvendsenPharvey: I use the latest version......
Ahh I just download JQueryPlugin.zip and install?
[07:27]
MichaelDaumpharvey, wow this is a shocking insight into mongodb :( ... they can't properly sort. so wtf. [07:29]
CDotyou mean they can't sort properly >:-) [07:29]
pharveyPeterSvendsen: gac410 put a lot of work into the command-line installer. You *could* "just" unzip it, but if you use the CLI, it follows any pre/post-install steps, and makes a nice backup of your current plugin in case you need to revert
pharvey looks for the CLI installer instructions
[07:30]
MichaelDaumah the one-legged man doing tai chew is back [07:30]
pharveyhttp://foswiki.org/System/InstallationGuide#Installing_extensions [07:31]
MichaelDaumMichaelDaum trying to get this tarantinoesque picture out of his head again [07:31]
PeterSvendsenpharvey: thx I try..... [07:33]
flexibeastSo, any thoughts on the 'FoswikiFeatures' list i've got so far? http://foswiki.org/Book/FoswikiFeatures [07:38]
pharveyflexibeast: yeah, it's worth mentioning http://foswiki.org/Development/ParameterizedVariables next to EasyMacrosPlugin - they're built-in to Foswiki 1.1.2+ (or was it 1.1.3?)
also, have you used TagsPlugin and TagMePlugin yourself? They both need major workovers IMHO
[07:40]
PeterSvendsenpharvey: I did the easy part by copy paste..... and in the configuration changed to JQuery version 1.7.1 but when I look at the plugin on the link you send me it say version 4.33... where can I look that it is the coirrect plugin I have installed? [07:41]
flexibeastpharvey: Okay, will check out the ParameterizedVariables stuff. [07:41]
pharveyah, there's two "versions" going on. 4.33 is the Foswiki JQueryPlugin package version. jQuery 1.7.1 is the version of the javascript library itself [07:41]
PeterSvendsenokay.. the reason why I ask if because I still get warnings.... e.g. Warning: Autocomplete has been deprecated. Please use the autocomplete plugin part of the jQuery-ui package. The new plugin is not backwards compatible. See the JQueryPlugin documentation for further details. [07:42]
pharveyflexibeast: parameterized variables are included in Foswiki 1.1.4+. See Item10187 and Development:TrimAllSpaces and Support:BestPracticeTip16 [07:43]
flexibeastMy org does use TagMePlugin - i agree it could definitely be improved, but the basic functionality is there. i haven't used TagsPlugin personally - didn't want to have to set up a db just for this. [07:44]
FoswikiBothttp://foswiki.org/Tasks/Item10187 [ Item10187: Support simple macro parameters ] [07:44]
pharvey(as examples). http://foswiki.org/Support/BestPracticeTip16
BestPractice16 is probably good because it shows an EasyMacros vs Parameterized example of the same thing
(I think).
[07:44]
flexibeastSo i'm comfortable at least leaving TagMePlugin there; i would prefer to do so, because tagging is something i suspect a lot of people will be expecting to be able to do, nowadays [07:45]
pharveyoh, no it doesn't. BestPracticeTip16 is just EasyMacrosPlugin. [07:45]
flexibeastHeh, i must say i was slightly confused there! [07:46]
pharveyI thought I did all those TODO items. Story of my life :) [07:46]
flexibeast*chuckle* [07:47]
pharveyI guess there's no good link to our parameterized variables feature - except the notes in the http://foswiki.org/System/PreferenceSettings doc.
http://foswiki.org/System/PreferenceSettings#Parameters even
[07:47]
flexibeastOkay, thanks; will see what i can come up with. :-) [07:49]
pharveyI just wish TagMe stored its data in DataForms rather than its own secret sauce. Not to mention the code is... well. In any case, MichaelDaum's Nat* ecosystem has things like ClassificationPlugin, which work in a more DataForms-tastic way [07:49]
flexibeast*nod* i initially played with ClassificationPlugin, but it was way too heavy for our needs. [07:50]
MichaelDaumfeel free to register a demo account at http://demo.michaeldaumconsulting.com [07:51]
pharveyI really enjoy seeing some effort on the book front, by the way. I've always tried to add doc whenever I can't point somebody to a solution.
probably more desperately need is a re-org of our existing doc. But... we're all volunteers, hard to get janitors for free :P
[07:51]
flexibeastYes indeed!
My hope is that maybe the book can provide a skeleton structure for thinking about how to reorganise the docs.
[07:52]
pharveygreat idea [07:53]
flexibeast(If nothing else.) [07:53]
MichaelDaumTagMePlugin was once invented by Pth. it did not really implement free tagging last time I looked. plus the way it stores tags out of band is problematic. its search does not scale and is unfriendly to wiki apping. so I wrote 2006.
ClassiPlug
[07:54]
flexibeastSomething that occurred to me is that maybe i should mention the Book effort on -discuss, as there might be people experienced with FW who aren't coders but would like to help the project out in other ways.
MichaelDaum: Well, i'm happy to mention ClassificationPlugin also!
[07:55]
pharveyI made an ajax skin for TagMe which allows free tagging
I suppose that ajax skin doesn't work any more
[07:56]
MichaelDaumI was more answering the tagging thread here. not the book undertaking. [07:56]
flexibeastFair enough. :-) [07:59]
PeterSvendsenAnybody who knows how to install Crypt::PasswdMD5? I use strawberry perl [07:59]
flexibeastOh wait, sorry, i was confusing ClassificationPlugin with TCAO! [08:02]
MichaelDaumTCAO is Lynnwood's baby, predating both TagMe and ClassificationPlugin [08:03]
flexibeastOkay.
i'd rather not link to the ClassificationPlugin page, since all that page says is "Work in progress" (despite also linking to your site).
[08:04]
SvenDowideitMichaelDaum mongodb suffers from the reality that the marketing is vague in places where someone like me expects that if they're going to be vague, its because that part is well and truely sorted
when the reality is that they're coding as fast as they can for the people that are handing them money, and the rest of us are just benefiting from it
basically, they support either webscale, _OR_ adhoc
but can't do both, because any query hitting data that has not been pre-optimised with suitable indexes, they (essentially) thor an exception
the premis is that if its not been planned for, then it should not slow down the legitimate uses
even that would be ok, if there were an infinite number of indexes, but that would cause more problems than it would solve.
sitting in on DBI development means I'm getting a completely different look at life tho
[08:10]
MichaelDaumit really sounds like mongodb are reinventing the wheel in some parts. problems did not change no matter what tool you throw at it. ignoring adhoc is so crazy that it is cool already ... so they think at least of themselves. [08:15]
SvenDowideitthats the point!
noSQL is entirely about reinventing the wheel
[08:15]
MichaelDaumI like wheels, i.e. as I get older. [08:16]
SvenDowideitme too [08:16]
MichaelDaumpoint is it is called Rock _and_ Roll. not just Rock. [08:17]
SvenDowideitits pretty reasonable to choose a rock if that is all you need [08:17]
MichaelDaumas as consequence you dont get far in a rock chair. you need wheels on it for that.
unless you are prepaired to make a fool out of yourself down the streets.
[08:17]
SvenDowideitthe funniest thing is that noSQL is older than SQL
but we don't want to talk about prime and other really old stuff
[08:20]
pharveyI really liked the look of MarkLogic, but it's not open$ource$
pharvey abandons ship
[08:26]
SvenDowideitYAY: Apple iPad 2 As Fast As the Cray-2 Supercomputer
i love modern HW
[08:28]
CDot"as fast as" == meaningless [08:38]
MichaelDaumin the beginning there was no sql [08:43]
laenIn the beginning there was no computer. [08:52]
.... (idle for 18mn)
pharveyin the beginning, there was only the void
which is where zero-point energy comes from?
[09:10]
***FoswikiBot has quit IRC (Ping timeout: 260 seconds) [09:22]
MichaelDaumyet there was no sql. so sven is right. [09:23]
........ (idle for 38mn)
***ChanServ sets mode: +o Babar [10:01]
jastso, mongodb needs actual planning if you're going to use it in big applications? like, I don't know, sql databases? unpossible!
anyway. here's an actual question. how would I go about profiling a slowish foswiki setup? ideally without throwing tracers everywhere
[10:15]
SvenDowideitmongodb's a bit less mature in approach
if you ask a mature storeage system a bad question, it tells you the answer, just a bit slower
mongodb just tells you not to ask
which is fine, if thats how you docco and market it
but they docco and market it as capable of adhoc queries on big data
and its not
profiling foswiki - mmm, i use nytprof
[10:25]
jastI'm gonna look into that, cheers [10:27]
SvenDowideit:) [10:27]
jastchances are a good part of the sluggishness is due to the client's virtualisation, but it can't hurt to check anyway [10:27]
SvenDowideitits very good
yes, in that case nytprof does not help
you need to get on the real box, and measure how often the vm is starved by other htings
i've failed several times to show anything conclusive
other than that foswiki on the same real HW shits all over the vm'd version
even when nothing else was on the vm server
SvenDowideit finds that aspect of virtualmachines very depressing
[10:27]
jastthere's no such thing as getting on the real box [10:29]
SvenDowideitthere is for me :) [10:30]
jastquite a few of our customers have an IT department that doesn't really know anything about anything
chances are they wouldn't know how to give us that kind of access... not that they would give it to us if they could
[10:30]
SvenDowideitsadly, even those that do can't do much about it
vm's basically raise the complexity to the point where sudden stalls in response are untrackable, or unreproducable
[10:30]
jast"but VMs lower costs and... I dunno but they told us it'd be great so we bought it" [10:32]
SvenDowideiti loooove vms
i use them everyday, on all my systems
but they are shite if you actually need consistent response
[10:33]
jastI'm typing this through a vm [10:34]
SvenDowideithehe
i'm on my notebook, using an xpra virtual session for xchat, running on a vm downstairs, talking to a znc irc relay server on my main servers vm
mmm, perhaps i've overdone it a little
[10:34]
jastoh, yeah, the server on which this irssi runs is virtualised, too [10:35]
SvenDowideitxpra means i have a persistent x session for each app
so i can go to the other room, continue chatting on the same xchat app, and connect to my chromium then walk back and continue on
shame xpra only does one client, i have to re-connect 4 or more x-apps each time i move
SvenDowideit is very enamoured of cheap pc HW
[10:36]
jastI don't really use that many computers
so I tend to pool all of the money in one system
[10:42]
SvenDowideithehe
i spent $200 on my latest 3 servers
[10:46]
jastI have a desktop computer and a notebook. that's it. [10:47]
SvenDowideitdual dual core xeons and 32GB ram
and about that on my current notebook
i'm buying hardware from ~2006
and its way fast enough, and extremely disposable
[10:47]
jastmakes sense [10:48]
SvenDowideitmmm, spent much more on my lcds tho
:)
laters - i think i might be sleeping soon
[10:48]
..... (idle for 22mn)
MichaelDaumquestion: why does Foswiki::Meta::setEmbeddedStoreFrom only remove _one_ newline from the end of the file after all %META records? [11:10]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Meta [11:10]
MichaelDaumcalling setEmbeddedStoreFrom twice will remove first one then another newline.
that's why MetaTests::test_validateMetaTagsInText fails
[11:10]
SvenDowideitremoving the newline is such an increadible pain in the bum
but, what changed to make the test fail?
or is it a new test?
[11:18]
MichaelDaumsetEmbeddedStoreFrom is called once more than before part of a reload
instad of s/\n$//s, the clean up should actually be s/\n+$/\n/s
[11:21]
SvenDowideityes, the question is - the test didn't used to fail
a few months ago, trunk has unit tests that all passed
is it a new test, or a new bug?
[11:22]
MichaelDaumit is a new bug
or better: an old bug that now pops out his head.
[11:23]
SvenDowideitor its old code that had to be called a certain way, and someone's now calling it incorrectly [11:24]
MichaelDaumI will add another test that calling setEmbeddedStoreFrom twice should be idempotent [11:24]
SvenDowideit(the legacy shit is really shit, and most of it crawford and I, and a few others have had a bad time trying to fix, only to have to unfix it because the fix broke something else [11:25]
MichaelDaumy [11:25]
SvenDowideitit would be better to fix the code that revealed the problem
so that I can replace the entire thing properly
atm, i can't actually commit my 5000 times faster replacement
because i dont know what tests are doing
[11:25]
MichaelDaumas I said: calling setEmbeddedStoreFrom twice should do no harm. yet it does as it changes the text consecutively. the fix is the fixed s/.../.../ above. [11:29]
SvenDowideityes, should
but does
and does so because its horrid ccode who's bugs some shittly bit of code relies on
i _think_
and unlike previously, we don't have enough legacy release managers helping us find them
i'm a bit confused tho
setEmbeddedStoreForm should read from a txt
so it won't be called on data its been called on before
as its input is readonly
except..... -
why the FK would you call:
my $topictext = readFile()
$meta->setEmbeddedStoreFrom($topicText)
$meta->setEmbeddedStoreFrom($topicText)
so, to ask again.
what recent code change reveals this issue?
[11:30]
MichaelDaumthere's one additional $obj->unload(); $obj->loadVersion(); which triggers the additional setEmbeddedStoreFrom() on itself.
this will extract the $text of the latest version and set it back once again ... and call setEmbeddedStoreFrom() yet again.
as a natural consequence of reloading the obj
this happens as part of a new+save. another save on the same obj will remove yet another newline thus creating a bogus change
there's no such test I guess testing two saves on the same obj ... making sure the topic stays the same of course.
somethign silly like new() save() save()
[11:42]
SvenDowideitthis will extract the $text of the latest version..
that means its more a bug in the getEmbeddedStoreForm
so - why is there a new additional unload&loadVersion?
mmm, i still don't know what changed to cause the unit test failure
SvenDowideit is pissing in $meta->put() :(
i wonder if anyone actually uses it :/
[11:50]
MichaelDaumhard to fix this newline at the end of a text [11:53]
SvenDowideitMichaelDaum do you actually know why the test went from passing to failing? [11:54]
MichaelDaumyes [11:55]
SvenDowideitwhat was it? [11:55]
MichaelDaumhttp://trac.foswiki.org/changeset/15268 last diff [11:55]
SvenDowideitwhat a terrible thing to do to perf [11:56]
MichaelDaumthere _are_ some tests that expect an extra newline at the end of the text even though it wasnt part of the text stored. [11:56]
SvenDowideityup [11:56]
MichaelDaumhow to we treat this
imagine a txt file "text text text\n%META" and "text text text\n\n\n%META"
how many \ns are in the text of (1) and (2)
and what do we want to happen on a save()
[11:56]
SvenDowideityup, there's a fun legacy
and it (used to) be worse due to wysiwyg
[11:58]
MichaelDaumoh ha [11:58]
SvenDowideitlets put it this way
i recall setting it to s/\n+$/\n/s
and that broke something, but i bet that was before unit tests
cos i agree, its totally bizzare
but whats weirder, is that $meta->embeddedForm might change things
as its not like that code is called on it own data
thing is, i have code to replace it
except i bet its incompatible wrt that \n too
mmm, no, i still need to make sure no-one uses $meta->put(
[12:00]
does anyone have a topic that has multiple TOPICMOVED meta in it?
and is it useful
mmm, shit.
WorkFlowPlugin might actually be using the array mess of meta->put
[12:16]
yup, crap. CDot whatapain
WORKFLOWHISTORY
[12:27]
.... (idle for 18mn)
ASSERT( $keyName, join( ',', keys %$args ) ) if DEBUG;
what is that asserting
[12:45]
ModAcOstEvery meta must have a "name" key
So %META:BLALBA{name="..." ...}%
[12:47]
SvenDowideitModAcOst no
you and i wish this were true
but it has never been the case, and while crawford and MichaelDaum and i agreed we'd be happy to do it
i found a plugin (workflowplugin) that exploits how the code actually works
so now, i have to find a way to support that, and then to measure how much that slows things down
[12:49]
MichaelDaumthe core does require a name. otherwise it cant extract the meta data records. [12:50]
SvenDowideitno, the core does not
it can, and workflowplugin uses that
[12:51]
jast3... 2... 1... fight! :} [12:51]
SvenDowideitthankfully, only about 5 people seem to have realised it
but i'm not going to remove the functionality unless it makes my fast parser significantly slower
[12:52]
MichaelDaumname is used to identify meta records. if there's no name attr in a meta record, then the store indeed still puts in those records. there are however some operations that dont work anymore for unnamed meta records. [12:55]
SvenDowideitsome.
however, as meta records are used by plugins, you can put & find them even without name
[12:55]
MichaelDaumsuch as pubKeyed, remove, putAll, copyFrom. just to name those I see atm. [12:56]
SvenDowideitsadly irrelevant
as meta used does not need them
[12:56]
MichaelDaummerge [12:57]
SvenDowideityup, totally un-necessary :( [12:57]
MichaelDaumanything formish uses names
same for attachments
[12:57]
SvenDowideityour are mentioning totally irrelevant things [12:57]
MichaelDaumeven plain put has got {name} hardcoded into it [12:57]
SvenDowideitplugin code can, and _DOES_ use put and find
so i can't just remove it
there's a pile of special sauce needed, and i'll bet theres no unit tests :(
putAll does not need the name
just to make life easier and harder
but gawd, what a heap of slow code to do something so simple
and all deoptimised for named access, and optimised so that writing to a file is easier
[12:57]
MichaelDaumquestion: what is reprev in TOPICINFO? [13:02]
SvenDowideitits an indicator that some time in the history of a topic, it was saved twice without the rev increasing
or something similarly useless
[13:03]
jastisn't reprev the parameter used for restoring an old revision? [13:04]
SvenDowideitdifferent context [13:05]
jastoh, term overloading then. nice. [13:06]
SvenDowideitdunno, i'm not looking up what the urlparam for restoring is :) [13:08]
and MichaelDaum no, copyFrom does not need a name attr either
SvenDowideit is going through re-writing them in order atm, and well, its a really really great eg of pre-de-optimisation
luckily, its past my bed time, so i'll continue some other time, but argh!
[13:13]
....... (idle for 30mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/dNTqjg
[foswiki/master] Item11983: fixing old bugs poping up now - MichaelDaum
[13:46]
***GithubBot has left [13:46]
FoswikiBothttp://foswiki.org/Tasks/Item11983 [ Item11983: prevent excessive calls into the revision system ] [13:46]
...... (idle for 29mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/VK-2SA
[foswiki/master] Item12079: twisty preferences added to the wrong zone - MichaelDaum
[14:15]
***GithubBot has left [14:15]
FoswikiBothttp://foswiki.org/Tasks/Item12079 [ Item12079: twisty preferences added to the wrong zone ] [14:15]
............. (idle for 1h3mn)
foswiki_irc2hello
my question is the follow: I would like to use the commentPlugin. If a new message is posted, the WIKINAME is written to the topic. BUT I would like to have a search result as well
e.g.: "JohnWayn wrote 3 messages"
How is it possible? :)
I would be very glad, if you can answer my question
[15:18]
..... (idle for 20mn)
MichaelDaumgosh there's so much smell in the store code.
for one Foswiki::Func::saveTopicText uses Foswiki::Meta::setEmbeddedStoreFrom to save the new text. alas this change will be reverted as the {_latestIsLoaded} flag is still false.
[15:42]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Func http://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Meta [15:43]
MichaelDaumFoswiki::Meta::new() called with a $text param will do both
but any other call to setEmbeddedStoreForm($text) is in danger of loosing its data as it does not set the {_latestIsLoaded} flag.
CDot, what do you think?
[15:44]
CDotvery dangerous territory [15:45]
foswiki_irc2I am not sure, wheater I understand this solution... [15:45]
MichaelDaumCDot, aah now that you say :/ [15:46]
CDotIIRC I deliberately didn't set _latestIsLoaded from setEmbeddedStoreForm because it *isn't*. _latestIsLoaded is meant to show that the latest on disk is now in memory.....
but if it hasn't been written to disc yet, the flag can never be true....
I was concious of trying to downplay saveTopicText because in my eyes it is a horrible, nasty, dangerous hack
[15:46]
MichaelDaumbut then any make-sure-you-run-on-latest-data using loadVersion is wrong potentially and there's no way to do it right. [15:47]
CDoty, I didn't say I had solved all the problems. Just trying to illustrate the thinking behind latestIsLoaded [15:48]
MichaelDaumI will add a {_latestIsLoaded} = 1 to Fun::saveTopicText ... just in case the horror isnt strong enuf [15:48]
CDoterm, I suspect that a lot is going to break. But I could be wrong. [15:49]
MichaelDaumyet I still wonder why it doesnt make its way down to the disk [15:49]
CDotI guess it will, given that this is in saveTopicText
so you may be OK
CDot promises to visit MichaelDaum in hospital if he isn't OK
[15:49]
MichaelDaumMichaelDaum isn't afraid of a one-legged [15:50]
CDotc'mere, and I'll bite yer knees off! [15:52]
............................................ (idle for 3h36mn)
***Heikkila has left [19:28]
...................................................... (idle for 4h28mn)
SvenDowideitmmm, so micha's commit last night looks like its dangerous, but has no unit tests to show its ok
and then the commit that shows the issue is troubling too
the worst thing is that without peter and kenneth to test ancient topics and styles, we're probly not going to notice the problems until some user hits them
so worst case, he's fixed a logical error that breaks un-unit tested backwards compatibility
which makes me think, stuffit, if we don't have a test for it, then we can't really hold back our work for it
MichaelDaum_, I noted 2 things
[23:56]

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