#foswiki 2012-08-22,Wed

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

WhoWhatWhen
***kip3f has left [00:35]
......... (idle for 43mn)
Cervator has left [01:18]
................................ (idle for 2h36mn)
gac410Hmm... We ought to use Date::Calc instead of Foswiki::Time::parseInterval. Our code assumes 1 Month = 30 days :( [03:54]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Time [03:54]
....... (idle for 34mn)
AlexisHazellHow long before 1.2 is scheduled to be released? [04:28]
gac410No real plan yet. There is some pending work that has been held up for various reasons. [04:29]
AlexisHazell*nod* So is there changes to the datetime handling scheduled for 1.2?
s/is/are/;
[04:30]
gac410no. [04:30]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/ZBp-iA
[foswiki/master] Item12023: LogDispatchContrib work - GeorgeClark
[04:31]
***GithubBot has left [04:31]
FoswikiBothttp://foswiki.org/Tasks/Item12023 [ Item12023: write a Log::Dispatch based Logger ] [04:31]
AlexisHazellOkay. The reason i ask is that i'd like to add support for some more datetime formats. [04:31]
gac410It was just an observation. There is one bug that would be nice to fix, but not the interval code. Not much uses that code, and adding a "month" is a somewhat dubious exercise. [04:31]
AlexisHazellIn putting together this TMS, i've had to write several workarounds for the fact that the format i really needed wasn't supported. [04:32]
gac410What format? [04:32]
AlexisHazellWell, something along the lines of 2012.08.22 or 2012-08-22.
The closet i could get was 2012.08.22.23.59.
[04:32]
gac410yyyy-mm-dd whthout the time? [04:33]
AlexisHazellYep. [04:33]
gac410hm. Spreadsheet, Or other macros? [04:33]
AlexisHazellWell, i don't remember finding anything suitable, particularly in the context of dealing with the values within JavaScript.
The TMS has now been rolled out; i'm going to give it another week or so of usage in order to find any uncaught major issues, then i plan to release it as a Contrib.
[04:34]
gac410The Foswiki::Time::formatTime() function certainly supports a $year.$month.$day or $year-$month-$day format. [04:36]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Time [04:36]
AlexisHazellSo it will be easier for others to see what i've been trying to do, and if i've taken completely the wrong approach. :-)
Yeah, it's support within dataforms that was the specific issue.
i'm not using the native FS interface for entering datetime values in forms.
[04:36]
gac410Ah. yeah dataforms - There are a couple of proposals for improved date handling in forms. IIRC. Have not seen any work. Things have been really quiet lately. Hopefully just summer vacations. [04:38]
AlexisHazell*nod* [04:39]
gac410One proposal - http://foswiki.org/Development/CreateISODateFormat [04:41]
AlexisHazellYes, i read through that; it was the ideas discussed there that i thought might have been in line for 1.2. [04:41]
gac410Depends upon the Committed Developer's plans. Need to talk to ArthurClemens -
FoswikiBot: seen ArthurClemens
[04:42]
FoswikiBotgac410: ArthurClemens was last seen in #foswiki 56 days 9 hours ago saying "nice excel like js editor: http://warpech.github.com/jquery-handsontable/". [04:42]
***gac410 has left [04:45]
.......................... (idle for 2h6mn)
ChanServ sets mode: +o MichaelDaum [06:51]
...................... (idle for 1h47mn)
PeterSvendsenHi all, I have transfered my old foswiki installation with content to a new server... it seems to work but e.g. the Redirect pludin doesnt work.... when I press a topic that are redirected it point to the old server.... where should that be changed, any good hints? [08:38]
...... (idle for 27mn)
***ChanServ sets mode: +o Babar [09:05]
..................... (idle for 1h41mn)
jastMichaelDaum: in WikiWorkbenchContrib there's a file pub/Applications/twb.js in the MANIFEST. unfortunately Foswiki 1.1.5 doesn't seem to accept that path (in fact it dies with a perl error) since it doesn't contain both web and topic path parts...
(using install via configure, at least)
[10:46]
MichaelDaumdoesn't accept? what does it say? [10:46]
jasterror in line 1346 of lib/Foswiki/Configure/Package.pm (1.1.5) due to the variable not being defined [10:47]
MichaelDaumwow never seen that one [10:47]
jastthe MANIFEST line is matched against a regex that assumes that pub/* lines are of the form pub/Web/Topic/file
the check isn't... quite ideal if it crashes on bad lines, I suppose :)
[10:47]
MichaelDaumyea well the twb.js should actually be located at pub/System/WikiWorkbenchContrib/twb.js ... my bad [10:48]
jastyeah, but might want to improve the check, too [10:49]
MichaelDaumbut the check "segfaulting" on anything non-compliant ... thats outchie [10:49]
jastyeah [10:49]
CDotMichaelDaum: do you have any store benchmarking scripts you can share? [10:51]
MichaelDaumnyprof
and some wrapper around it
lemme sea
I did some perf tests for the new page cache ... how did I do that? hum.
http://foswiki.org/Tasks/Item11549
here's a wrapper for stress-testing foswiki using siege http://pastebin.com/k1RjSwJT
[10:51]
PeterSvendsenHi all, I have transfered my old foswiki installation with content to a new server... it seems to work but e.g. the Redirect pludin doesnt work.... when I press a topic that are redirected it point to the old server.... where should that be changed, any good hints? [10:56]
MichaelDaumCDot, so I first did an overall RL-test scenario, and then went into details using nytprof on one innocent topic [10:58]
CDotI want a way to compare the performance of different store implementations across a range of scenarios [10:58]
MichaelDaumyea well [10:59]
CDotThere has to be a way to eliminate variability in the web server [10:59]
MichaelDaumthen keep this variable constant across all test
the interesting bits are in the ratios not the abs nrs
abs nrs play a role in showing off ... which is important too, but not when working on the thing
[11:00]
CDotyes of course. There are factors - such as process copying - which are important to take into account in a webserver context. However there are ways to eliminate, or at least minimise, those overheads that are nothing to do with the store implementation
what I don't want to do is to set up a test environment that makes one store implementation look stupidly bad. I want to try and compare "best" against "best"
[11:04]
MichaelDaumpsgi would be cool for that
as thats the way to plug in the app into a test env instead of a web server
thats the way to go to automate benchmarks per release and build
besides creating a standardized set of jobs to measure
benchmarking the store with all of its locking and protocols going on, mixing views, edits and saves makes sense too due to caches being invalidated by one or the other process.
[11:13]
..... (idle for 22mn)
PeterSvendsenI try to enter the plugin config page through "InstalledPlugins" but when I try I get the message "The website declined to show this webpage" I am logged in as admin!! [11:39]
..... (idle for 22mn)
***Teridon has left "PING 1345636884" [12:01]
.................... (idle for 1h38mn)
CDotThinking about it further, I'd like a set of benchmarks that let us answer questions like "is my site working correctly". An end user site should be able to pick up a benchmarks set and with a minimum of fuss, run it on their install. Results should indicate performance under different test conditions.
CDot speculates whether the logs from a large site could be used to replicate a "typical" load.
[13:39]
gac410Hm... We'd need a much more detailed log I suspect. capture / replay for workloads sounds challenging. [13:41]
CDotI was thinking of replicating the frequency and spacing of different types of request
though how to eliminate robots.....
[13:41]
gac410well, they are part of a regular workload on public sites. [13:42]
CDotpublic sites, yes, but a typical foswiki site is intranet
of course many such intranet sites will have appliances too, but..... that should be something you can switch in/out
[13:43]
gac410Maybe write an event analyzer that will extract statistics on arrival rates, and transactions without reporting any details. And ask some intranets to run it. [13:44]
CDoty, that's the sort of thing I had in mind
CDot is pulling down the f.o logs to have a ply with
[13:44]
gac410Though I've found that there are events that are not logged. And really the biggest workload I suspect on many sites is search which wouldn't be captured.
ugh. Just ran trunk unit tests. 142 failures.
I guess we don't have the nightly tests running any more.
Most of the RCS Handler tests failed (108 failed out of 126). And a bunch of Manage.pm and Store tests too.
[13:45]
..... (idle for 22mn)
Not looking good. Issues with White space folding , topic META , Search results, HTML Validation, META validation, RCS tests are a disaster.
Doesn't anyone run the unit tests? I ought to run git bisect and then revert the failing commits :(
[14:10]
not my configuration. cleaned the install, pseudo-installed a default config, and reran. 146 failures. [14:17]
CDotgac410: is that ion release branch? or trunk? [14:19]
gac410trunk [14:19]
CDotis release branch ok? [14:19]
gac410I'm starting to step back in time now to see when all the breakage happened. [14:19]
CDotCDot has a suspicion [14:19]
gac410I have not tried release branch, but there have been very few commits, so probably okay sure hope so. :)
gac410 steps into wayback machine for June 1st 2012 and reruns the FoswikiSuite
[14:19]
CDotinteresting, and somewhat surprising. The ratio of views to saves on f.o is 58:1
CDot expected it to be higher
damn, I wish we logged searches.
[14:23]
gac410yeah. we really could use a lot more optional instrumentation down in core functions. Search and runtime too! [14:26]
jastis foswiki.org down intentionally?
I wanted to open another issue ;)
[14:34]
gac410Hm... apache is down. I didn't do it. CDot? Are you tweaking the server? [14:35]
CDotnope
I just copied the logs onto a local machine, but that shouldn't have killed apache
[14:35]
gac410starting it now. [14:36]
jastcheers ;) [14:36]
gac410It's back up. [14:36]
jastyay [14:36]
gac410That's strange. There are two root users logged in - both from same IP. [14:37]
CDotme
I had one process zobied, so started another
[14:37]
gac410okay. Then it probably was a glitch. anyway it's back up now. [14:37]
jastMichaelDaum: another one for you (sorry ;)): http://foswiki.org/Tasks/Item12049 [14:43]
gac410CDot: Stepping back to June 1st, tests appear a lot better. But also running **much** longer. Still not done. [14:46]
MichaelDaumjast, thx. fixed and released. [14:46]
jastcheers! [14:52]
CDotit looks to me like the RCSHandlerTests were written to assume that certain operations do not modify the topic, and that assumption no longer holds
specifically, it's clashing over the insertion of META:TOPICINFO lines
[14:52]
gac410What's causing that? Is that a bug or a (missing?) feature request? [14:53]
CDotCDot suspects MichaelDaum's changes, but is investigating [14:53]
gac410The unit tests are all now in sync from trunk to release11, Differences need to be committed to both branches with conditional tests. :(
Or at least they were in sync after a lot of hard work by pharvey.
[14:54]
CDoty, confirmed that it's the insertion of a META:TOPICINFO line in a topic that didn't previously have one [14:56]
gac410Is it actually modifying the stored topic, or inserting a phantom TOPICINFO [14:57]
CDotnot sure yet [14:57]
gac410MichaelDaum: have you been running the unit tests? Any ideas what's going on? [14:57]
MichaelDaumis this trunk or rel branch? [14:58]
gac410trunk [14:58]
MichaelDaumyes ran them [14:58]
gac410And they passed? [14:58]
MichaelDaumof course. yes the store now explicitly treats the txt file as a cache: there's a cachetopicinfo() method for that making sure all checkins update this info accordingly [14:59]
CDotoh, no, the META:TOPICINFO is being inserted *into an attachment* [14:59]
gac410yikes ... that's not good [14:59]
MichaelDaumoutch
which test is that, CDot
[15:00]
CDotn.m. I was wrong; it's in a topic [15:00]
gac410The RcsHandlerTests are mostly all failing. 146 fails. [15:00]
CDotsub verify_Item3122 { [15:00]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/EPLuEA
[foswiki/master] Item12049: fixing VIEW_TEMPLATE in user profile to use Local instead of Set - MichaelDaum
[15:01]
***GithubBot has left [15:01]
FoswikiBothttp://foswiki.org/Tasks/Item3122
http://foswiki.org/Tasks/Item12049 [ Item12049: NewUserPlugin: view template setting in System.NewLdapUserTemplate too intrusive ]
[15:01]
CDoty, a call to $this-._cahceMetaInfo is adding META:TOPICINFO into the topic before it is saved
the test, however, does not recognise that
[15:01]
MichaelDaumnother bit rot [15:02]
gac410How were you running the tests to get them to pass? [15:02]
CDotdid you forget to merge something from a branch? [15:02]
MichaelDaumI told my computer "computer, run all unit tests"
let me fix
[15:02]
gac410Stepping back to June 1st, everything passes except VCStoreTests::verify_Inconsistent_getRevisionAtTime_VCStoreTests_RcsWrap And some configure tests fail because foswiki.org was down.
MichaelDaum: If the test change is dependent upon Release 1.1 vs. 1.2, need to make it conditional, The 1.1 and 1.2 suites should be identical.
[15:04]
MichaelDaumgac410, why's that
we have trunk/UnitTestContrib and one in rel branch
[15:05]
gac410We are planning on dropping the per-branch unit tests. There will only be one version to ensure consistency. [15:05]
MichaelDaumboth check different cores
sounds messy
[15:05]
gac410pharvey has done mega-work to implement that. [15:06]
MichaelDaumwhen was that discussed [15:06]
gac410Keeps things consistent. It's been going on gradually for many months. [15:06]
MichaelDaumsux [15:07]
gac410Finds clear differences between branches. People were sloppy with tests, and changing end user visible results between branches and just fixing by changing the test. [15:07]
MichaelDaumsee, that does not work out.
when there's ongoing development on trunk, you will always have unit tests being created and maintained on _trunk_ as well
[15:08]
gac410It's been that way for ages now. pharvey keeping them in sync. Right. New tests only run on trunk, unless they should run on both branches. [15:09]
MichaelDaumas new features arent there on rel branch, the same unit tests are _bound_ to fail applied to a release that hasnt got these changes
sounds pretty essential
[15:09]
gac410He added some simple tests to allow tests to be skipped depending on the release.
So it is very easy to write that a test for a new feature only applies to branch > blah
[15:09]
MichaelDaumthe longer these unit tests are exposed to programmers, the more complex the suite gets. [15:10]
gac410And also very easy to assert only depending upon release. [15:10]
MichaelDaumnow, I can fix RcsHandlerTests for trunk, but they won't pass on rel branch
there's too much difference here
[15:11]
gac410nope. Not how it's done. [15:11]
MichaelDaumnope he said [15:12]
gac410Hm... pharvey documented it all somewhere. [15:14]
foo, he didn't document it. There is a new skip_test_if function that can conditionally skip tests ... used for new features missing in older releases.
And individual tests can be conditional using $this->check_dependency('Foswiki,<,1.2') )
I guess we need a feature request. I assume we'll have your objection MichaelDaum and have to go to a vote :)
[15:23]
MichaelDaumyes because that leads towards spaghetti unit tests [15:26]
gac410but it helps prevent the mess we have with bugs in 1.1 not fixed in 1.2 because the unit tests were not synced up. And the huge extra work trying to keep the tests in sync.
Your Store changes are targeted for 1.1.6 according to the task, so you'll have to sync the tests anyway if that goes ahead.
if ( $this->check_dependency('Foswiki,>=,1.2') ) { assert that topic has meta }
[15:26]
***ChanServ sets mode: +o Babar [15:36]
gac410The test framework was changed in Item11479, back in January. Documentation is in test/unit/FoswikiTestCase.pm, with examples, for both skip_test_if and check_dependency [15:37]
FoswikiBothttp://foswiki.org/Tasks/Item11479 [ Item11479: Add ability to skip tests/suites ] [15:37]
gac410Also a new function check_conditions_met [15:40]
MichaelDaumall those unit tests operate on botched test data [15:54]
gac410Right. So by fixing test data so that it passes on trunk and fails on Release branch, it points out a bug on release that needs to be fixed before the next release. [15:55]
MichaelDaumnone of the topic texts have a well formed format
y
all those fixes wrt the store havent been merged to the rel branch
[15:55]
gac410The tests should still be in sync, even if they are failing on release branch, as it points out a missing fix that should not be skipped.
If it will be a long time before the fix, then they could be marked expect fail if release < 1.2, to allow the suite to complete.
[15:56]
.... (idle for 18mn)
MichaelDaumdown to 6 failures [16:15]
fixed
gotta run see you tomorrow
[16:26]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/ocxpKQ
[foswiki/master] Item11983: fixed unit tests to expect well formed topic data once it got stored - MichaelDaum
[16:30]
***GithubBot has left [16:30]
FoswikiBothttp://foswiki.org/Tasks/Item11983 [ Item11983: prevent excessive calls into the revision system ] [16:30]
....... (idle for 32mn)
gac410Wrote up feature proposal Foswiki:Development.UnifyUnitTests - reflects pharvey's work already implemented. [17:02]
FoswikiBothttp://foswiki.org/Development.UnifyUnitTests [ UnifyUnitTests ] [17:02]
..... (idle for 24mn)
gac410Still 38 tests failing. [17:26]
***WikiRingBot has quit IRC (*.net *.split)
Guest87207 has quit IRC (*.net *.split)
[17:33]
verne.freenode.net sets mode: +oovo gmc Babar FoswikiBot SvenDowideit
verne.freenode.net sets mode: +o julian_work
verne.freenode.net sets mode: +v WikiRingBot
verne.freenode.net sets mode: +oo MichaelDaum_ ColasHome
[17:39]
gac410CDot: Foswikirev:15063 breaks FormattingTests::test_mailWithoutMailto [17:42]
FoswikiBothttp://trac.foswiki.org/changeset/15063 [ Changeset 15063 – Foswiki ] [17:42]
................... (idle for 1h33mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/usYxCA
[foswiki/master] Item11860: Mark the tests as expect fail - GeorgeClark
[19:15]
***GithubBot has left [19:15]
FoswikiBothttp://foswiki.org/Tasks/Item11860 [ Item11860: SEARCH Pager not working when zeroresults parameter passed to ViewTemplate ] [19:15]
...... (idle for 27mn)
***Teridon has left "PING 1345664529" [19:42]
.............. (idle for 1h5mn)
verne.freenode.net sets mode: +o ChanServ [20:47]

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