#foswiki 2012-11-05,Mon

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

WhoWhatWhen
timotheit does exactly what I said. there's a routine (actually several) called stripTraceback. They get the result $@, and either do or don't strip the traceback fro mthe string. [00:00]
SvenDowideityup, clear as mud
they do that, but i don't know when, and how that affects me
i'm being user-ishly obtuse, but its true nonetheless
[00:00]
timotheThen you're not an expert, and shouldn't open that tab. [00:00]
SvenDowideitexpert != i've read the code
expert ~~ i've read the docco provided
i know what you mean tho
[00:01]
timotheIn configure v111, we can add "true expert" mode. That's the one where you've read the code too. [00:02]
SvenDowideitoh dear v111 [00:02]
timothefor now, let's get v2 (or is it v3) out. [00:02]
SvenDowideitsurely we should just integrate configure into a chrome pane by then >:} [00:02]
timotheBlock of plexiglass in a museum, I hope. [00:03]
SvenDowideitit seems to me that DebugTraceback is a developer thing
and developers are supposed to set $ENV{FOSWIKI_ASSERTS} = 1;
and #$ENV{FOSWIKI_MONITOR} = 1;
[00:03]
timotheIt is a developer thing that I want support people to be able to turn on if a support call requires it. The dependent clause is why I put it in the .spec. [00:04]
SvenDowideitin their LocalLib.cfg
SvenDowideit mumbles about implied knowledge :(
and looks to see that he hasn't
y, and i'm sort of asking if you realise that you may has duplicated something we already have and use for that
but in another sense, its not identical, just similar
[00:04]
timotheI removed information that support had before in the interests of clarity, and provided a means for them to get it back.
Without letting a user loose on a config file with vi.
[00:05]
SvenDowideitoh, and here's a q - does that setting only apply to configure?
as i said - the user is expected (required) to go edit the bin/LocalLib.cfg
using vi or whatever
s/user/admin/ - darned hard to have a name for that thats not slightly overloaded
[00:06]
timotheAt the moment, yes. But others can certainly use it -if the semantics are maintained. I'd rather teach configure to make the necessary edit to LocalLib than use the edit as an excuse for not making diagnostics easily accesible.
editors are what create corruption in config files which is what creates support headaches.
[00:07]
SvenDowideitwe cna't edit LocalLib.cfg from configure [00:08]
timotheAnd why not? [00:08]
SvenDowideitits bad enough that we have a web app that modified a file in th eLIB path
but that file is in the BIN
which makes sysadmins scream
the reason that the file is there, is to stop foswiki from searching the path for the LocalSite.cfg
and from guessing wrong and in the process causing other security messes
[00:08]
timotheSo we add a symlink - apacheconfig already has follow symlinks on. [00:09]
SvenDowideitnot everywhere it doesn't
we deploy to lotes of webhosts where you have to be much cleverer
s/cleverer/sneakier/ ?
[00:09]
timotheI'd rather solve problems than make excuses. And if we can solve them for a useful subset of environments, that's better than for none. [00:10]
SvenDowideitplease, don't forget that we've had 12 years of dealing with the truely disgusting mess that is deployment
but adding non-doccoed duplications of settings will make support harder
i'd rather not make it harder for the difficult cases
[00:10]
timotheAnd I've had over 4 decades for more than just this product. Life is hard. But that's no reason to cave. [00:11]
SvenDowideitbecause we've made it simpler for the already simple cases
i'm saying that focussing on the 'subset' _is_ caving
mmm, oh well, flaky net perhaps
timothe - if you're reading the logs - i'm curious if by 4 decades you mean that no-one should question the assumptions and dissagree with you
or just that in asking questions we're denegrating that experience
cos we're not, we all have things we bring, and they're all very useful - that that includes questions
[00:11]
pharveyis this about enabling asserts from configure? [00:22]
SvenDowideiti think so, but without looking at the code, (ie, looking at the help for the setting) you'd never know :)
na, its more about actually seeing a useful stacktrace
as it seems to get filtered
mmmm, fasinating
run a search over 10 webs
one has a js issue
so you get on results
who coded this
[00:23]
pharveyI must admit I've wasted a lot of time trying to get a stacktrace sometimes
seems more difficult with mod_perl setups somehow
or just the way I set them up in apache at least :P
[00:25]
SvenDowideitmod_perl? mumble :)
i've heard perl people talk about that the same way some do about taint and locales
[00:26]
pharveypeople are against locale?!
the anti-taint sentiment is really, really surprising to me... I don't understand it.
[00:26]
SvenDowideitin following the perl5 porters mailing list for a few years
well, its simple - most perl != web stuff
and most of them don't turn on taint, as it 'gets in the way'
and so development and fixes for taint where not high on the list when perl when through its 'bad years'
[00:27]
pharveyI know, I've had irritating conversations on #perl that end with something along the lines of "taint doesn't make your code safe", haha.
I see
[00:29]
SvenDowideitand now, those working on performance just see it as an annoying slowdown thing that doesn't really help
i'm incorrectly paraphrasing, but its that kind of multi-step mess
where its like legacy code that people were afraid of for so long that no-one really recals the point of it, but touching it is 'scary'
argh! how do i get mongo to give me a real stack trace from $where code
(answer: you should not be using $where, its slow)
SvenDowideit mumbles more
[00:29]
pharveythe mongo.log [00:33]
SvenDowideitthat does not give a stacktrace [00:33]
pharveyit does if you increase the verboseness doesn't it? [00:33]
SvenDowideitif you're lucky it tells you the wrong function call in the $where
and the error that is actually not there
[00:34]
pharveyoh yeah, this rings a bell [00:34]
SvenDowideitSvenDowideit won't look, as i think i see the error :) [00:34]
pharveywe had mystery crashes in the d2n() code, and the error log wasn't entirely helpful... or was it? [00:34]
SvenDowideityup, both obfuscated and enough to let me debug for a few hours [00:35]
pharveyI should really stop skipping breakfast [00:35]
SvenDowideityes, that is correct
thankyou
SvenDowideit potters off to have coffee and breakfast
[00:36]
pharveySvenDowideit: Individually most of these are trivial, and I've got branches with some of these already started, but in order of effort required easiset-last: AddDataFormCreatorWizard, MacroToListInstalledFormFieldTypes, CustomNewUserTemplates, FormatArbitraryTimes, ConfigurableCookieNamesAndPaths, AddENDINCLUDEandSTOPSECTION,
and I actually have a number of attempts at AddHideOptionToSTARTSECTION where I thought it was non-trivial but then realized it wasn't so hard except now I can't seem to locate that work anywhere and will have to re-travel that journey all over again
[00:41]
SvenDowideitAddDataFormCreatorWizard - what i want to do is simplify the one thats there
so it comes down to just editable on the definition (for now)
as there are some more new features needed to make it wizzy
but i would like to see what you have - so commit and we'll reduce scope after
MacroToListInstalledFormFieldTypes is coded and commited
so refinements can be called bug fixes, unless you're expanding further
mmm, darn, thats the limit of my battery - about an hour and a bit
[00:43]
....... (idle for 30mn)
InitForm test failures - in trunk? [01:18]
gac410No Release116, which I'm trying to build for alpha. See the email: see http://fosiki.com/Foswiki_Release01x01/ for output files. - Tested with perl v5.14.2, from /usr/bin/perl 2 failures: InitFormTests::test_tmpl_form [01:18]
SvenDowideitmmm gac410 so why do those tests fail on rel?
i guess thats the point :/
[01:18]
gac410yeah. I think it happened when you reverted the form default value stuff. I've tried to fix, but was not successful. [01:19]
SvenDowideitah, roger [01:19]
gac410When I dont really understand the test, I hate to "fix the test" when maybe I should "Fix the bug" six / half dozen I guess. [01:20]
SvenDowideity, i detest changing tests
especially when they've been there for ages
v scary
[01:20]
gac410yup. [01:21]
SvenDowideiti guess i'm not really getting anywhere with this mongodb issue, i'll potter to the other computer [01:22]
.... (idle for 15mn)
mmm, sorry, looks like it'll take some time before i can get to my computer again, bbiab [01:37]
gac410seeya [01:37]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/5GHDYw
foswiki/master fbd23c1 TimotheLitt: Item12180: Merge bursts of unsaved changes updates to reduce traffic and (hopefully) make the interface seem more responsive. Provide initial status from main screen if LSC is valid...
[01:41]
***GithubBot has left [01:41]
FoswikiBothttp://foswiki.org/Tasks/Item12180 [ Item12180: Implementation for AJAXOnDemandCheckersForConfigure ] [01:41]
...... (idle for 27mn)
SvenDowideitgac410, so the initform unit test is from trunk?
SvenDowideit is missing the conditional stuff, will have to put it in
[02:08]
dj_segfaultHi, guys. When I go into /bin/configure install extensions on my 1.1.4, I get 'Invalid version format (non-numeric data) at /home/neagile/foswiki/public_html/lib/Foswiki/Configure/UIs/EXTENSIONS.pm line 283.' (pharvey, we were talking about this earlier). I put in debugging to write out each plugin it visits in the loop, and the last one it reports is SetVariablePlugin. That plugin is disabled. The next plugin i [02:13]
gac410hm. I'm running the release11 InitForm tests ... I think. Just started the suite a bit ago.
dj_segfault: message was truncatd .. The next plugin i
[02:13]
pharveydj_segfault: configure still processes plugins even if they're disabled, FWIW. Have you tried upgrading it? [02:17]
SvenDowideitand that error is not to do with plugin code
its to do with the cfg values in spec files
and in LocalSite.cfg
[02:18]
gac410Sven ... you talking about InitForm or dj_segfault's error [02:19]
SvenDowideitdj_segfault's
initform i have fixed, but am struggling to commit
for un-godly computer issues
[02:19]
gac410dj_segfault: If you like to live on the edge, you could try the lib/Foswiki/Configure/Dependency.pm from latest Release01x01 code. [02:20]
SvenDowideitin adding the defaul column to the unit tests for 1.2, i changed the results [02:20]
gac410Well as long as the test is broken and not 1.1.6, i can keep going on a 1.1.6_001 alpha build. [02:21]
SvenDowideitcorrect
i confirm that it is the test that is wrong
[02:21]
gac410excellent. [02:21]
SvenDowideitand its commited
now, where was i :/
oh yeah, bugging pharvey :(
[02:22]
gac410Now should I do the hack to protect <nop> when entered into a <input ...> tag. [02:22]
SvenDowideitpass :/ [02:22]
gac410:P [02:22]
SvenDowideiti really don't have an opinion on how dangerous it might be [02:23]
dj_segfaultSorry, wife pulled me aside. [02:24]
gac410I'm pretty sure it's safe from a rendering perspective. It's more the overhead of 3 more regex scans of the topic during render to allow html-inside-html [02:24]
SvenDowideitSvenDowideit sticks fingers in his ears and yells lalalalalala [02:24]
gac410when it's really not necessary any more. s [02:24]
dj_segfaultcan't find a file called specfile [02:26]
SvenDowideitthere are lots
find foswiki/lib -name Config.spec
[02:27]
gac410Before fix for Item11480, <input ... text='<nop>WikiName' > the nop was required or we'd render the wikiname into the text box. Now it is not needed, and will break the input field without my hack. [02:27]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to Release01x01: http://git.io/0bqdCw
foswiki/Release01x01 f463424 SvenDowideit: Item11966: add conditional test results - 1.1 does not have the 1.2.0 deafults column...
[02:27]
***GithubBot has left [02:27]
FoswikiBothttp://foswiki.org/Tasks/Item11480 [ Item11480: Contents of HTML input blocks should not be rendered as wikitext. ]
http://foswiki.org/Tasks/Item11966 [ Item11966: add another a Default column to DataForms ]
[02:27]
gac410dj_segfault: What causes that?
I missed something. Did you end up trying 1.1.6 Dependency.pm?
[02:27]
dj_segfaultgac410: When I go to Install or Update Extensions [02:28]
SvenDowideitpharvey, darnit :p
there's the problem
[02:28]
gac410Anytime? [02:28]
SvenDowideityou're dehydrated [02:28]
dj_segfaultpharvey: <Arnold> It's not a tumor</arnold> [02:29]
SvenDowideitmmm, ok,
Invalid version format (non-numeric data
[02:29]
gac410dj_segfault: Anytime on vanilla 1.1.4, or after you tchanged something. [02:29]
pharveyhah, good point. My home office is far from climate-controlled.
dj_segfault, hehe
pharvey finds h2o
[02:29]
dj_segfaultgac410: It happened after I added a few extensions and enabled them. I' [02:29]
SvenDowideitthere's a sig for you - no, you do not take a tablet for dehydration [02:29]
dj_segfaultI've since disabled them with no change [02:30]
gac410di-hydrogen monoxide. Terribly dangerous stuff 1000's die every year.
dj_segfault: what ext. did you add. Let me try to break my 1.1.4 system.
[02:30]
SvenDowideitdj_segfault, thats why i'm pretty certain its a spec file that is incorrect [02:32]
gac410Or a messed up version string, since we load all the plugin modules to check the version, even if disabled. [02:32]
SvenDowideitoh do we? deadly [02:33]
gac410That's another think timothe fixed for 1.2.
We also used to load Foswiki.pm under configure, which loaded LSC a 2nd time,
[02:33]
dj_segfaultIIRC RemoveTrashPlugin EditChapterPlugin and possibly TagMePlugin [02:33]
gac410Okay. I'll try one at a time. [02:34]
dj_segfaultI got that off of ls -alt in lib/Foswiki/Plugins/
Thanks
[02:34]
gac410well so far, all installed, no configure errors. I'll try enabling them all.
hm... Still okay. IIRC one issue that can cause strangeness - we tweaked Dependency.pm a few times to deal with it, is if there are missing dependencies, and the attempt to determine the version has a compile error, sometime strangeness happens.
[02:37]
dj_segfaultWhat are the odds that upgrading to 1.1.5 would (1) be painless (2) fix the problem? [02:42]
gac410without knowing the issue, it's hard to say.
gac410 generally doesn't like the shotgun approach to debugging. .... but it works
[02:42]
dj_segfaultThat's why I started shoving debug statements in it. [02:43]
SvenDowideitone thing to try is to rename all the Config.spec files [02:43]
gac410I've confirmed my system is 100% vanilla. [02:43]
SvenDowideitmmm, no, it can also be the cfg file itself now ? [02:44]
gac410I'm still not following. You cannot get into configure at all, or you cannot access the ExtensionsInstaller.
iirc, the extension installer itself does not use the Config.spec file at all.
Except to copy it during the install.
bin/configure (not install) finds all Config.spec file and merges them with LocalSite.cfg.
[02:44]
dj_segfaultAh. I have another possible candidate. I recently installed TaskManagementContrib which is brand new. I remember I didn't see it in the list afterwards, even enable/disable
Can you try that one?
[02:47]
gac410okay... hang on. [02:47]
dj_segfaultI found that via ls -alt public_html/working/logs/configure [02:48]
gac410BINGO... That's it. [02:49]
dj_segfaultExcellent. What do I do now? [02:54]
gac410I'm trying to figure out a solution. [02:55]
dj_segfaultCool. Thanks. [02:55]
gac410Copying over a newer Dependency.pm doesn't help. [02:56]
I don't think it's actually TaskManagementContrib. It does not install any modules. Only topics. But it pulls in dependencies, and it might be one of them [03:01]
SvenDowideitor its version? [03:02]
gac410There is no place for configure to get the version. iirc, it depends upon the lib/Foswiki/Contrib/....pm [03:05]
dj_segfaultInteresting. One of the dependencies is SetVariablePlugin, which is the last one my debugging code prints out
According to the sparse documentation at http://foswiki.org/bin/viewauth/Extensions/TaskManagementContrib
[03:05]
gac410Oh crap!!! It's using the new style version strings. which obviously breaks old foswiki. That's not good. [03:07]
dj_segfaultWhich? TMC or SVP? [03:07]
gac410SVP
dj_segfault: Edit lib/Foswiki/Plugins/SetVariablePlugin.pm Comment out and insert:
#use version; our $VERSION = version->declare("v2.30");
our $VERSION = '2.30';
[03:08]
dj_segfaultIs there some magic that knows not to install version >$x for release $y? [03:09]
gac410no.
This is bad.
gac410 needs to find out how badly things break.
yeesh. This breaks 1.1.6 too.
gac410 is looking pretty bad right now.
[03:10]
dj_segfaultSorry, I'm confused. Both of those lines are already in the file. What am I commenting out and what am I putting back in? [03:14]
gac410Comment out the line "use version; our $VERSION = version->declare("v2.30");
and add the line "our version = "2.30";
er... but with the upper case
ie. get rid of the "v" and get rid of the declare stuff.
gac410 does not understand how if ( $ext->{installedVersion} eq 'HEAD' ) { ... which is line 283 - can trigger that error
[03:15]
dj_segfaultThat fixed it. Thanks.
Should I now uninstall those two plugins?
[03:22]
gac410Up to you I guess. If you don't need them..
I think I need to revert all the version stuff. This is a disaster.
[03:25]
SvenDowideit:/
what is it that causes the crash?
[03:26]
gac410I don't understand .. version/vpp.pm is throwing the error. Something with the way configure is calling the versionm as the string is correct, and it doesn't throw errors when using the module. [03:27]
SvenDowideitso it might be resolveable with a wrapperere
mmm, complicated
[03:27]
gac410yeah.... I don't understand it, ... AHHH... I it's the compare.
I think perl knows it's a real version string and when we say ''eq HEAD" it says WHAT?
with perl version objects, it "knows" how to compare them, all built in. Us doing strange things like expecting a VERSION string to be HEAD, is just going to not work.
I obviously didn't understand the implications of using a version object.
[03:28]
dj_segfaultI get the sense I shouldn't file a bug for this because it's a larger-than-that-plugin issue, right? [03:37]
SvenDowideitgo on, raise a bug:p
YAY.
[03:38]
gac410Yeah. I'll mea culpa it. Basically the bug is in UIs/EXTENSIONS.pm it needs to determine if the version isa version object. and not compare it to HEAD if it's a real version object.
And then we need to provide a hotfix for the older code. Or we abandon the whole idea of using real versions.
[03:39]
SvenDowideitand then there's the problem of older installed :/
na, not abandon, just er, yeah, we may need to mumble.
ie - add the isa code
oooo, can we um detect what version of foswiki?
[03:39]
gac410where [03:40]
SvenDowideitso that the plugin's code has a 'you're dumb' fallback
ie, in MyPlugin.pm
our $VERSION = your new hotness
[03:40]
gac410Ah... if > F115, our $VERSION = yea. good idea [03:41]
SvenDowideit$ VERSION = 'craptackual' if (foswiki_too_dumb)(; [03:41]
dj_segfaultYou mean hot mess [03:41]
SvenDowideitie, make the if (<1.1.6)
or even 1.2
so that we're downgrading to support old crud
but default to new hotness
[03:41]
gac410If we can get to real version strings, we can discard a bazillinon lines of rather arcane comparsion code
I'll play with that a bit.
[03:42]
SvenDowideityup [03:42]
dj_segfaultI don't understand how you get away with not having version/release mapping. So if a plugin gets upgraded to work with release X, when an X-Y install goes to upgrade plugins, there's nothing stopping the install from upgrading the plugin to a version that won't work? [03:45]
SvenDowideitwe don't really 'get away with it'
non-core plugins are developed adhoc, by 'other people'
[03:46]
dj_segfaultExhibit A [03:47]
SvenDowideitand the pain is that we basically can't be everwhere all the time and do everything
yay!
[03:47]
gac410We have proposals to do " real versions and dependency checking" but it needs a developer with lots of ambition [03:49]
dj_segfaultGetting all Cathedral and the Bazaar on yo ass [03:49]
SvenDowideitdj_segfault, you mean you're going to help and do it :p [03:50]
dj_segfaultYou need an actuarial student [03:50]
SvenDowideitpharvey, er
i just ran FuncTests using mongo and without -CSAD and it completed
i hope i'm very confused
[03:50]
dj_segfaultdj_segfault runs away ducking [03:54]
SvenDowideitdj_segfault isa Duck?
or just duck-typed :D
[03:56]
gac410SvenDowideit: The premise appears valid. We can define $VERSION as a real version, or a string depending upon something. Have to be careful though because 1.1.6 and beyond, the foswiki version is also a version string. [03:56]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/KUVUGQ
foswiki/master 710d8cc TimotheLitt: Item12210: Add EMAILADDRESS checkers, enable for {WebMasterEmail} and remove old checker...
[03:57]
***GithubBot has left [03:57]
FoswikiBothttp://foswiki.org/Tasks/Item12210 [ Item12210: Easier generic checkers ] [03:57]
gac410hi timothe .. you've missed the excitement. I've utterly botched the new module version stuff for backwards compatibility. Installing a new "real version string" plugin breaks configure. [03:57]
timotheArthurClemens: If you svn up to at least 15905, look under mail -> General -> {WebMasterEmail} First "Test" class button.
gac410: Well, you said you like to hear people scream...
[03:58]
gac410:) [03:58]
timotheWhatever you did should be fixable.
Tell me a story...
[03:59]
gac410Configure Extension list compares the version to HEAD .. to detect if it's pseudo installed. Once $VERSION is a version object, the comparison barfs because HEAD is non-numeric.
I think we have a solution. in Plugins ( If old foswiki, $VERSION = "some string" else $VERSION = version->declare("v1.2.3");
And then in new foswiki we need to figure a way around using HEAD to detect pseudo install.
[03:59]
timotheThat doesn't sound too hard to fix. But don't forget the version stuff in CGISetup.pm also looks at version strings.
The way I detect pseudo-install is with -l
But we could (and maybe should) teach it to write a .pseudoinstalled in the entity's 'lib/'
[04:01]
gac410Yes. that's what Configure/Dependency.pm does. Assigning {installedVersion} a value of HEAD is legal. but comparing {installedVersion} eq 'HEAD' when {installedVersion} is a version object wil lcrash
The issue is not the pseudo installed stuff. It's comparing a non-pseudo-installed version to "HEAD".
We could do something like return version 0.0.0_000 if pseudo-installed.
[04:02]
timotheSure, but isn't it if( A eq 'HEAD' or B eq 'HEAD' ) then manual compare else $version ->? [04:04]
gac410No. If installed eq "HEAD" then it's pseudo installed and no furhter comparison is done. [04:04]
timotheActually, pinstall implies a developer, so probably 9999.99_999 would be better. Or 99999.svnrev (use svnversion cmd) [04:05]
gac410We don't attempt to display or compare pseudo versions.
Yeah that works too. Any string that is also a legal version when the installed version happens to be a version object.
[04:05]
timotheif you make pinstall look 'newest', then developers can live dangerously [04:05]
gac410good idea. Thanks. I'll play with that. [04:06]
timotheNo problem. Have a look at my latest commit; some will be familiar, but it's the first non-'validation' feedback button. [04:06]
gac410will do... But right now, anyone who installs SetVariablePlugin will totally disable install or remove extensions. Need to fix that really fast. [04:07]
timotheOh, well. I'm off to other things. Have fun. [04:08]
gac410Excellent. 9999.99_999 works perfectly SvenDowideit. So Dependency returns that string for pseudo installed, and configure compares for it, and no more crash.
And the If / Else in modules will keep older versions happy.
[04:09]
GithubBot[foswiki] FoswikiBot pushed 2 new commits to master: http://git.io/plN3aQ
foswiki/master 2b21f12 SvenDowideit: Item9715: Don't run a single unit test in a worker thread. ever....
foswiki/master edf095c SvenDowideit: Item9715: hunting down bugs in op ref for mongodb...
[04:10]
***GithubBot has left [04:10]
FoswikiBothttp://foswiki.org/Tasks/Item9715 [ Item9715: Add entry points for deep store listeners ] [04:10]
SvenDowideitpharvey, this is maddening
on this ubuntu box i just set up an lxc container on to run mongodb
FuncTests pass without -CSAD
i'm running Fn_SEARCH atm to see if things are as i expect, and if so, will run the suite to get a total
but basically we're in a weird place.
[04:19]
dj_segfaulthttp://foswiki.org/Tasks/Item12225 Plugins using new version system break older versions Foswiki [04:20]
SvenDowideitcruel! [04:20]
dj_segfaultSometimes the truth hurts. [04:21]
SvenDowideitgrin [04:21]
gac410I can take it. [04:21]
dj_segfaultI could have phrased it as "this plugin breaks 1.1.4", but that's not the heart of the issue [04:21]
gac410if (substr( $Foswiki::VERSION, 0, 1) eq "v") {
print STDERR "Using new version crap\n";
[04:21]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::VERSION [04:21]
dj_segfaultSorry, it's an Agile thing. We like to be honest with ourselves [04:22]
gac410Believe me, I'm not offended in the least. I feel really horrible that this is causing any issues, but thankfully we have an easy fix. [04:22]
dj_segfaultExcellent. Always willing to help debug stuff. [04:23]
SvenDowideityup, 28 test failures in Fn_SEARCH, i have progress
<pharvey> SvenDowideit: Individually most of these are trivial, and I've got branches with some of these already started, but in order of effort required easiset-last: AddDataFormCreatorWizard, MacroToListInstalledFormFieldTypes, CustomNewUserTemplates, FormatArbitraryTimes, ConfigurableCookieNamesAndPaths, AddENDINCLUDEandSTOPSECTION,
<pharvey> and I actually have a number of attempts at AddHideOptionToSTARTSECTION where I thought it was non-trivial but then realized it wasn't so hard except now I can't seem to locate that work anywhere and will have to re-travel that journey all over again
you need to pick one and doit, else, you'll never get one done
[04:24]
gac410The hotfix to older releases for 99% of the users is to just delete the test for HEAD. For people running with pseudo-install, use "9999.99_999" instead of "HEAD" to detect pseudo-install.
So I'm wondering if it's worth adding If / else logic to every plugin to be backwards compat, vs. a simple patch to Configure/UIs/EXTENSIONS.pm
[04:26]
SvenDowideityou don't need to add it to all plugins [04:27]
gac410But it does show.. I need to add test cases to the Dependency tests, to test modules using "real" version objects. [04:27]
SvenDowideitjust to the EmptyPlugin and EmptyContirb tempaltes [04:28]
gac410It needs to be added to all non-core plugins. [04:28]
SvenDowideitand any that are modernised [04:28]
gac410Ah.... yes true. [04:28]
SvenDowideitthat should be a small set [04:28]
gac410I'm just going to look for "v" in first character of Foswiki version. since we've never used v* numbering. [04:29]
SvenDowideitlooks like given the progress, I'm about to call feature freeze for 1.2.0 [04:29]
gac410gac410 scrambles to review FeatureProposals one more time. [04:29]
SvenDowideitor to be more accurate, no-one has given me sufficiently concrete implementation commitments to suggest that i should wait for them [04:30]
gac410:P
So you are not going to move CPAN to a sep. contrib.
[04:30]
SvenDowideiti'll talk to jast tonite hopefull and pharvey will have some time too
there are so many things i'd like to do
that its probably better not to
when did 1.1.0 come out?
[04:30]
gac410I really think we would be better to find a solution to not ship any CPAN. [04:31]
SvenDowideitthat is simply unrealistic [04:31]
gac4107a5d10a Item000: BUILD Foswiki-1.1.0 at Mon Oct 4 15:25:59 2010 GMT [04:32]
SvenDowideitwe ship CPAN for people that have a server that has no access to cpan [04:32]
FoswikiBothttp://foswiki.org/Tasks/Item000 [ Item000 ] [04:32]
SvenDowideitand i know more than enough of those [04:32]
gac410cpanm can bootstrap itself with a single .pm module. [04:32]
SvenDowideitfrom where?
their servers have no net access
they copy over the zip
and for the same reason they rarely use plugins
plus, they use windows on net connected machines, and non-windows for the foswiki server
[04:32]
gac410We give them a tool to grab the latest CPAN .tgz files. Then they use bootstrap cpanm and local zipfiles. [04:33]
SvenDowideitthat is a reason they don't upgrade
rather than one zip, you now want what?
i really fail to grok what is so bad about shipping them
but then i also fail to grok why anyone would not 1. be using debian and 2. use the debs
[04:34]
gac410I'm more concerned with shipping so many pre-built deps. Some of the LogDispatch deps use XS modules by default. [04:35]
SvenDowideiter, i didn't know that
i'd have blocked it from core if i'd realised
[04:35]
gac410When I built the lib/CPAN/lib I made sure there were no XS versions included [04:35]
SvenDowideitor at least thought more about it
so its not using XS by default
[04:35]
gac410ie. Some of them will use XS if available, If you build them on a system with the xs avail, they install the XS version and do NOT install the pp version. [04:36]
SvenDowideitso?
ok, let me ask this differently
[04:36]
gac410So I had to jump through hoops to get the code for lib/CPAN/lib [04:36]
SvenDowideitfoswiki.zip is huge [04:36]
gac410yes [04:36]
SvenDowideitthe CPAN dir is miniscule in proportion
and the CPAN dir is ignored by default
[04:37]
gac410er... Wait until Arthur adds DateTime [04:37]
SvenDowideitand used to help people that otherwise could not install
thats not 1.2, so lets not borrow trouble
and even then, compared to the FamFamFam dir, irrelevant
so, given that your concern can't be causing most users that don't use it to use an outdated version
and you can't be concerned about zip file size
what exactly is the problem you're highlighting?
[04:37]
gac410I'm just very concerned. I don't have every env. to compare. we ship lib/CPAN/lib ... (some purePerl thing).. Customer locally installed ... xs version, which *removes* the local .pp version. Is there any chance that you end up with mixed versions.
as we have sumFunc.pp and they have someFunc.xs
I just don't know. As I obviously didn't know enough for the VERSION change either.,
[04:39]
SvenDowideitwe ship lib/CPAN/lib which for most users is _not_ in their perl LIB paht
mumble, path
[04:40]
gac410Sure it is. Foswiki AUTOMATICALLY adds it to the end of the path.
You don't need to do anytihng at all to activate it. Only settings needed if you want it at the front of the path.
[04:40]
SvenDowideitah, so isn't in their perl path when they install cpan packages [04:41]
gac410right. [04:41]
SvenDowideitits not in the __perl__ LIB path [04:41]
gac410correct. It is however at the end of the Foswiki LIB path. [04:41]
SvenDowideitand its last in line for foswiki, so anything newer gets there first
not as nice as what i would prefer, but even so
[04:41]
gac410Which means there is a *potential* maybe small that some local and some fw get used. because when xs is detected and built, pp is NOt installed. (but we have the pp version) [04:42]
SvenDowideityup, which is why i'd ship it un-enabled
but that makes life painful for hosted server newbies
[04:42]
gac410ship what un-enabled. [04:43]
SvenDowideitthe lib/CPAN/lib auto add to end of LIB
ut even so
weve had this for so very many years
[04:43]
gac410Yes. but with a very small number of simple libs. We have EXPLODED the list for 1.2 [04:44]
SvenDowideitand yes, you are right that there is a tiny window of possible pain, but can you count the number of times we've seen it? [04:44]
gac410never seen, but the old list had no XS aware modules to my knowledge. [04:45]
SvenDowideitexploded to 93 files
mmm
[04:45]
gac410The XS vs. PP gets determined when you build the CPAN module. I had to play games to get the PP version.
Anyway... I really want to fix SetVariablePlugin and find which others got broke.
We have time to figure out the lib/CPAN/lib issue
[04:45]
SvenDowideitya
and feature freeze doesn't mean we can't change our minds
[04:46]
gac410right. And we're not really talking about a feature, it's a packaging question. And could have multiple solutions. [04:47]
SvenDowideitwe do, and y, i want more - that was what the separate into contrib really is for [04:47]
gac410Pulling lib/CPAN/lib into a Contrib is a simple change. No impact on the ultimate install zip. [04:47]
SvenDowideitjsut like i'd like a lightweight jquerycontrib
debian already has a libjquery
the problem is (and why i want to call feature freeze) is that there are hundreds of tiny things like that that i want to do
and if i don't call a stop, 1.2.0 will not happen
whereas if i call 1.2.0 now, we have cool stuff for 1.3.0 in 6 months
[04:48]
gac410I think Koen's dependency_installer.pl is a interesting approach, but way to complex / boil the ocean. Need similar concept, that can install a local lib/CPAN/lib module for when the package installer isn't available. [04:49]
SvenDowideitthe store commit i made in the last few days was a really good wakeup to 'duh, should release more often' [04:49]
gac410yes. We take way too much time. [04:50]
SvenDowideitless time ~~ more dicipline to say, yes, thats nice - next time
i assign pharvey the task to fix the bugs in wysiwyg formfields
(if any)
[04:50]
gac410And also more discipline to say "No that's a feature. Not in a patch release!" [04:52]
SvenDowideithehe
there's a good reason to have more than one rm
and to have them working on opposing releases
so they can point fingers at each other >}
or was that shake bones
[04:52]
gac410I think we need to "Branch EARLY". So trunk stays "experimental" NextRelease takes features, and ThisRelease ONLY is bugfix. [04:53]
SvenDowideitthere is only one reason not to
once branched, the cool and sexy kids won't test the nextrelease
i used to agree with you
[04:53]
gac410what, cool and sexy stay on trunk? [04:54]
SvenDowideityeah
plus, the branch early process lost when people decided it was too hard to work in more than one
[04:54]
gac410I think the challenge is that trunk gets too many "really big" changes that take too long. Maybe separate git branches will help solve it. [04:55]
SvenDowideit_but_ this was in svn days
i think it has
thats why store2 is not in trunk
and unicode isn't either
[04:55]
gac410great minds think alike. Branches are just too easy with git. I must bounce between branches a dozen times a night. [04:56]
SvenDowideitso, i'm proposig that _if_ we declare freeze tonite [04:56]
dj_segfaultUbuntu model: Long Term Support releases mixed with more cutting edge releases [04:56]
SvenDowideitthat i will branch as soon as f.o is running 1.2, which i would prefer to do in 2 weeks time
so ner
dj_segfault, i'm going for a perl release like thing
lts requires people we just don't have
we basically need to be able to drop support for code that is more than a year or 2 old
as plugin developer's rarely support even that
[04:56]
gac410I really need to get f.o on 1.1.6 [04:58]
SvenDowideitcorrect [04:58]
gac410So we need to get 1.1.6 out. [04:58]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to Release01x01: http://git.io/ponHzw
foswiki/Release01x01 cf3f9f6 GeorgeClark: Item11267: Don't use "HEAD" to detect pseudo install....
[05:11]
***GithubBot has left [05:11]
FoswikiBothttp://foswiki.org/Tasks/Item11267 [ Item11267: Migrate from subversion to git ] [05:11]
SvenDowideitweeee pharvey 125 incorrect results
and 36 are HoistTests that i have not updated for the op ref changes
[05:11]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/ShMbsw
foswiki/master 3361c64 GeorgeClark: Item11267: Don't use "HEAD" to detect pseudo install....
[05:12]
***GithubBot has left [05:12]
gac410I assume I need to bump the version of SetVariablePlugin to fix the Version string
Confirmed fix works back on Foswiki 1.1.3
[05:13]
SvenDowideitassume sounds right [05:14]
gac410Okay. SetVariablePlugin updated to Version 2.31. Has new version check code. [05:24]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/WuW0-Q
foswiki/master 403192d GeorgeClark: Item11267: New version breaks old Foswikis...
[05:25]
***GithubBot has left [05:25]
gac410So disaster averted. SetVariablePlugin is the only one actually released with the new version string. I'll work on fixing the rest of the plugins with the new conditional version string tomorrow.
Also need to add some unit tests for the new perl version string... if I can figure out how.
Yes. I've added a bunch, so have others.
[05:29]
SvenDowideitya
as a testcase topic, but its better than nothin
[05:31]
gac410on. never mind [05:32]
SvenDowideitnope, completely differnetly done
and ssp has been changed to use a hash of function pointers
but the test topic, that has real value to me
[05:32]
gac410yeah I saw that work pass by. I watch the commits now and then. [05:32]
SvenDowideitits hard when one change takes 20 commits to land tho [05:33]
gac410They also added a %CALCULATE% macro that's a registred tag.
yeah. that practice of one file per checkin is completely absurd. Anyone can see it's just for padding commit counts.
[05:33]
SvenDowideitonce file, worse, its more like 4 commits per file
into different branches
[05:34]
gac410oh, yeah.
gac410 still has a svn checkout. occasionally running a svn up.
:)
[05:34]
SvenDowideitoh dear - maybe its the same as excel? >:}
mmm, to use it, i need to add the CALCULATE macro - which might also not be shite for compratiblilty
[05:39]
gac410XOR: Each integer is treated as true/false and then logically xor'd together. The acutal value of the integer other than 0 or non-zero doesn't matter.
I added BITXOR so that we actually have a bit-wise XOR
[05:40]
SvenDowideitmust admit, i'd use a logical xor more than a bitwise one
(these days)
mmm, ok, atm, i think i might add the testcase file and CALCULATE to 1.2
and in the process have a measure of where we are compatibility wise
waddayarecon
[05:41]
gac410I think CDot had an alternative implementation of a registered tag. but commented out. [05:43]
SvenDowideitSvenDowideit is impressed that peter has added tests
or hidden somwhere
[05:43]
gac410Ah... my mind is refreshed. It
it is NOT the XOR that was broken. The BITXOR did not do a bitwise XOR
[05:44]
SvenDowideitgiggle [05:44]
gac410If passed a character string, operates in TWiki compatbility, performing a bitwise NOT operation on the string [05:44]
SvenDowideitah, v confuslein [05:45]
gac410I think I discovered it when adding to unit tests. However a bitxor became a not, I have no idea about that particular logical conclusion.
illogical conclusion?
I've probably broken some other compatibility as well. I definitely found / fixed bugs as I added unit tests.
anyway. time to sleep. Maybe I'll get 1.1.6_001 built and installed on F.O tomorrow.
[05:46]
SvenDowideitsweet
sleep well :)
[05:50]
gac410ty have a nice day. [05:50]
dj_segfaultSame here. Gnight [05:54]
***gac410 has left [06:01]
...... (idle for 25mn)
ChanServ sets mode: +o OliverKrueger [06:26]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/-BPiow
foswiki/master dc02643 SvenDowideit: Item8417: add CALCULATE macro (Crawford's experimental SSP macro) to allow evaluation in the normal order (named for compatibility) and add Peter's TestCases to show where new SSP formulae have been added...
[06:39]
***GithubBot has left [06:39]
FoswikiBothttp://foswiki.org/Tasks/Item8417 [ Item8417: When using CALC in search, CALC in format will not be executed ] [06:39]
.......... (idle for 47mn)
***ChanServ sets mode: +o MichaelDaum [07:26]
......... (idle for 40mn)
jastSvenDowideit: I'm not sure if I can deliver an all new locale system this week, so you might as well count me out for 1.2 [08:06]
Babarjast: may I see the task for the locale bug?
'cause trunk uses locales, and -T, afaik. So I did not understand George's statement that -T ought to be removed when running under locales
[08:14]
jastit's not quite up-to-date tbh
well, perl has a fairly big bug with locale+taint checks, in that regexes taint the capture vars in many more cases than they should
as a result, we have http://foswiki.org/Tasks/Item11953
example script: http://jk.gs/locale-short.pl.txt
[08:14]
Babarhum... as I said earlier, we had this issue for utf-8, and I thought we worked around it [08:17]
jastutf-8 is a different issue, and it's not due to a perl bug
(afaict)
[08:18]
Babarwait... your example doesn't make sense
$1 is nothing. So it could be pretty much anything
[08:20]
jast$1 is 'foo' [08:20]
BabarI wouldn't call that a bug
ah, missed the ?
yeah, this good old bug...
that $1 isn't tainted until you access it
[08:20]
jastit's not _that_ old. didn't exist in 5.8 ;) [08:21]
Babar$ perl -T -Mlocale -MScalar::Util=tainted -le'"foo" =~ /[xy]?(.*)/; warn$1;die if tainted($1)'
foo at -e line 1.
but:
Died at -e line 1.
$ perl -T -Mlocale -MScalar::Util=tainted -le'"foo" =~ /[xy]?(.*)/; die if tainted($1)'
[08:21]
pharveygood.. OLD... bug? [08:21]
jastoutputting tainted vars isn't forbidden [08:21]
Babarold as in last month when I discussed it with jast [08:22]
jastalso, Scalar::Util::tainted is broken [08:22]
Babarjast: I never said that :) [08:22]
pharveyScalar::Util::tainted is broken though [08:22]
Babarnot really... [08:22]
jastI can consistently make it return wrong results [08:22]
BabarI just did :) [08:23]
jastat least in 5.14 [08:23]
pharveywell, either that or our taint method is busted [08:23]
Babarwithout the warn, it doesn't crash
with it, it does.
[08:23]
jastthat's because of a bug in Scalar::Util::tainted
use a different implementation and it'll always die
[08:23]
Babaryou're not following...
your implementation will make it crash because you copy the string, therefore access it
Scalar::Util checks the string itself, which isn't marked as tainted, until it is copied
[08:23]
jastthen what marks it as tainted?
if that information isn't present in the initial string
[08:24]
BabarI'd say it's part of the bug, not related to Scalar::Util [08:24]
jasthow could it possibly be? the resulting string is only tainted with the [xy]? present
so that 'tainted' info must be stored *somewhere*
[08:25]
Babaragreed.
$ perl -T -MB::Deparse -Mlocale -MScalar::Util::PP=tainted -le'"foo" =~ /[xy]?(.*)/; die if tainted($1)'
Died at -e line 1.
so the non-XS version of Scalar::Util works
(it uses kill 0)
[08:25]
pharveyoh SvenDowideit - Item8417 - cooooool :) [08:34]
FoswikiBothttp://foswiki.org/Tasks/Item8417 [ Item8417: When using CALC in search, CALC in format will not be executed ] [08:34]
SvenDowideitpharvey, :p
jast, i need to talk to you about it in more detail, but won't have time for another 2-3 hours
basically, if we can make a workable plan that we will deliver to, then i can work on wiggling
but without a plan that we progress on, its just risking a delay for zero result
i'm also willing to feature freeze everything except locale
_if_
but i'll talk to you about it later (hopefully today)
[08:39]
jastall right [08:43]
I've updated the task to include some more details [08:51]
..... (idle for 20mn)
***ChanServ sets mode: +o Babar [09:11]
....... (idle for 34mn)
SvenDowideitpharvey did you see the updated list of test fails i pasted on the mongodb topic? [09:45]
ok jast, so your plan is, create Foswiki::Locale which has 'use locale' [09:50]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Locale [09:51]
SvenDowideitand then imports the the foswikiRegex's and uses them from a function
and then to use that module everywhere that matters
leaving the old regexs defined for use by plugins (which can be fixed in their own time) and places in the core where its unimportant (for now)
thats very very similar to what we did in Foswiki::Sandbox when that was made
[09:51]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Sandbox [09:53]
SvenDowideit(so i'm pretty comfortable with that idea)
if you can make a something as vague as the Foswiki::Sandbox::untaint
ie, Foswiki::Locale::regex(RegexNameinHash, functionToCall)
we might be able to defer any problems that we didn't think of beforehand
_if_ my guesses above are inline with your ideas
then all i need, is to know by when you think you could _realistically_ commit 99% feature complete by
and depending on that estimate, we can decide :)
[09:53]
jastSvenDowideit: the problem with that is that locale is applied to the regex at the point where the regex is defined [09:59]
SvenDowideitso you copy it? [10:00]
jastso I'm not sure you can go like "take this non-locale regex and apply it with locale enabled"
haven't tested, though
[10:00]
SvenDowideitmind you, our stored regex's are often built up from strings
so there's scope there too
[10:00]
jastlocale doesn't affect strings, though... fortunately :)
thank the celestial teapot for small favours
[10:01]
SvenDowideitdon't count those chickens
perl 5.20 hasn't landed yet :p
plenty of time to break stuff
[10:01]
jastby perl 5.20 I'm confident we can have the migration to PHP complete :p [10:02]
SvenDowideiter, wierd
i was migrating to either c++ or javascript
[10:02]
jastnode.js? :} [10:02]
SvenDowideiti did some of that too yes [10:03]
jasttagline: the fun way to block other requests! [10:03]
SvenDowideitbut more 'all in browser' with data svaing direct to apache webdav
i don't mind the single event thing, the code style is more mumbly
[10:03]
jastthere's a fairly nice framework for writing web apps in C++ [10:04]
SvenDowideitna, didn't use that [10:04]
jastserver-side ones, that is [10:04]
SvenDowideitused the boost spirit library and some fast eval [10:04]
jastthe fun part is that it abstracts away client/server presentation issues [10:04]
SvenDowideitso it does all the easy stuff :>} [10:04]
jastthat is, you can write logic and it'll automatically generate corresponding javascript code, too [10:04]
SvenDowideitooo, thats more interesting [10:05]
jastwith graceful degradation and all that [10:05]
SvenDowideittemplates based?
mmm, no, i shouldn't look, i'll ask you again after 1.2.0's out
[10:05]
jast:) [10:05]
SvenDowideitso
i need as estimate
[10:05]
jastto quote the website: The API is widget-centric and uses well-tested patterns of desktop GUI development tailored to the web. To the developer, it offers abstraction of web-specific implementation details, including client-server protocols, event handling, graphics support, graceful degradation (or progressive enhancement), and URL handling. [10:06]
SvenDowideitelse i cna't decide anything [10:06]
jastright...
there is no real complexity in what I've suggested
it's basically just busywork
[10:06]
SvenDowideitso we agree on complexity >:} excellent
except
for it to succeed, there are lots of tests to write
[10:07]
jastless tests than with the initial approach, though :) [10:09]
SvenDowideitna, same number of test
just less stress if we miss some
[10:09]
jastit might be tricky to write stable tests, since we'll either have to assume that all test systems have a specific locale in common, or we'll have to do horrible, horrible things [10:10]
SvenDowideit:p
why?
it will be turned on by a cfg
[10:10]
jastwell, test results will depend on the specific locale being used [10:10]
SvenDowideitwe have a way to set different cfg values and then restart the session [10:10]
jastif the locale we decide to use in tests isn't installed on the system, all tests will fail [10:10]
SvenDowideitoh, thats ok [10:10]
jastright [10:11]
SvenDowideitas we detect that and skip test for tha tlocale
pick a set of them, use any that are installed
[10:11]
.... (idle for 17mn)
ok, 1.2.0 feature freeze it is
jast you have til the end of the month assuming its as non-complex as we both think it should be
pharvey that means your list of todos is shorter
about half what you listed this morning is basically done, so you can finish any bits i may have missed :)
and ffs, this wide char thing is horrid
my ubuntu box has perl 5.14.2, mongodb2.0.7, and 0.45 driver
but no crash, no hint of a complaint about wide char
[10:28]
.... (idle for 16mn)
jastcan't talk much right now; wi-fi is being extremely unstable :( [10:48]
SvenDowideitno worries - i've commited you in email :p [10:55]
.... (idle for 16mn)
jastwhat exactly do we want to test w.r.t. locales, anyway? I'm not completely clear yet on how we can test it without spending most of the effort pretty much only testing that perl's locale support works [11:11]
SvenDowideitwe want to add tests that will show up failure if someone breaks locales
both our code, and perls
more ours, but still
in essense, write tests that fail today but will pass when you are done
test driven development will also tell you when you've had enough :)
mmm,
my @tokens = grep { !/^!.*$/ } @{ $searchOptions->{tokens} };
[11:14]
pharveyboo [11:15]
SvenDowideitheya :)
putting %SEARCH{"*" type="word"}% seems to 'work'
[11:15]
pharveywoot [11:16]
SvenDowideitbut putting * into the WebSearch topic results in horrid red lines of crap [11:16]
pharveyjs error? [11:16]
SvenDowideitna, its a regex bang in Meta.pm
its clearly a bug, but i want a test case ffs
it fails for good reason
$text =~ /*/ is after all a bug
[11:17]
pharveyDon't you hate it when you accidentally clear your browser history. I think I had nearly a year's worth, now I have to type things in full [11:18]
SvenDowideitna, i hate it when my browser can't cope with the 100 tabs i have open (to read later, honest)
and i have to delete it manually
happens about once a month :/
i really should go get the source for firefox and fix that
and fix those bugs in apache that annoy me
but no, i volunteer to RM 1.2
ah, the crash is beacuse we try to highlight the searched term?
do we have that?
well, my my, we do rock more than we used to don't we
[11:19]
pharveyyep :)
I think Arthur did it
[11:22]
SvenDowideitwhy am i so stupid as to put bad inputs into our UIs 'just beacuse i wonder what happens'
pharvey when you see me in sydney, kick me please
[11:23]
pharveywill do [11:24]
SvenDowideitYES! i was just restricting the test to a 'reasonable' subset of topics
now i have a test, i can 'just fixit' :) guess that'll keep me busy til thrusday :/
[11:30]
MichaelDaumthe fix to the flawed fix to http://foswiki.org/Tasks/Item11480 is even more flawed [11:37]
jastphew, for a moment I was afraid you'd be talking about the CompareRevisionsAddOn issue :} [11:38]
MichaelDaumnaw, this about my good morning shock "where are all my <input>s?" [11:39]
SvenDowideitMichaelDaum wondersful :/ can you throw some tests at it? [11:39]
jastsounds minor [11:39]
pharveySvenDowideit: I'm about to commit ./pseudo-install.pl -E<config> <module>, Eg. ./pseudo-install.pl -ESeleniumAutoBuild UnitTestContrib or Eg. ./pseudo-install.pl -EAutoBuild MongoDBPlugin [11:39]
MichaelDaumSvenDowideit, yes I would if I had time [11:39]
SvenDowideityou'd use less time writing them than finding the surprises every morning :) [11:40]
pharveyit adds lib/Foswiki/(Plugin|Contrib)/<module>/<config>.spec to LocalSite [11:40]
SvenDowideitpharvey oh, very cool [11:40]
MichaelDaumMichaelDaum fighting back things pushing onto the stack ... a typical monday morning [11:40]
SvenDowideitMichaelDaum yup, sameo
thats why i've declared feature freeze
otherwise i'd never get the F8ker out
this way we have simple and not so simple things for 1.3.0 or 2.0 in june
[11:40]
MichaelDaumhm [11:41]
SvenDowideitpharvey i'm not sure '.spec is the right name [11:42]
MichaelDaumthere's stuff in NatEdit I was going to release somewhere in between [11:42]
SvenDowideitif they are bugs fixes, the no problem [11:42]
MichaelDaumnope feature, and payed work [11:42]
SvenDowideitif they are more, then thats ok in the next few weeks
feature freeze - as in no new features, finish what is started
[11:42]
pharveySvenDowideit: it's exactly like Config.spec, but... extra bits of .spec you can summon by name [11:43]
MichaelDaumfeature freeze is feature freez [11:43]
SvenDowideit(plus talk to me about it&risks) [11:43]
MichaelDaumit adds an in-edit attachment link + upload [11:43]
SvenDowideitConfig.spec defines types and docco [11:43]
pharveyok then, I'll make them .cfg [11:43]
SvenDowideiti'm imagining atm that -E imports _values_ [11:43]
MichaelDaumthis once _was_ a feature of natedit but rottened [11:43]
SvenDowideitthat may be closer to the truth :) [11:43]
pharveyyou know me, I'm terrible at naming things :P [11:43]
SvenDowideitMichaelDaum don't stress - tell me about it when you're close
i'm still hoping that jast can make it for 1.2.0
[11:44]
MichaelDaumme2 [11:44]
SvenDowideitfreeze, plus things i've been convinced will help
wrt natedit, well, i'm pretty wiggly
as its totally new anyway
[11:44]
MichaelDaumnot really ;) [11:45]
SvenDowideitit is totally new to core [11:45]
pharveyit's easier to sneak in features to plugins, specially TinyMCEPlugin [11:45]
SvenDowideitso this time is the first time that its not as subject to release process [11:45]
MichaelDaumnatedit is only 5+ years old [11:45]
SvenDowideitquite separate from dealing with core release process
where we trade off stability of pre-existing things
the less time its been in core, the more option we have to change it
[11:46]
jastI'll welcome some help with the whole locale tests thing
but I'm fairly sure I can get the core functionality done in time
[11:47]
SvenDowideitonce you get started with a few tests, you should be able to get help from other people that care about locales
that kinda being important to them >:}
not to say i won't try to help, but i fail to see where i will have time
i essentially have under 2 months to test everything
and fix anything i finx
neither will be acheived, but that is the job of rm
[11:48]
pharveyrm 0r [11:51]
SvenDowideitpoof :)
course, i won't be doing much til wednesday - nice ful day tomorrow
up at dawn, kindergym, library, park, dunno
[11:52]
pharveysounds fun [11:55]
jastdo we need a feature proposal for the whole locale thing? [12:08]
BabarSven said he was ok to have it in trunk
jast: installing 5.8.8 to see if it fails
[12:09]
jastthat one specific issue reported should not happen with 5.8.8 [12:12]
BabarI could bisect it to see when it appeared
if I got the perl source tree :)
[12:13]
jastI spoke to #p5p, someone theorized that it happened when the code for character classes in regexes got an overhaul
keep in mind 5.10 has a partial version of the problem and 5.12 and up have the full problem, so you'd have to bisect twice :)
[12:14]
SvenDowideitjast, proper international everyting support has been on the roadmap for this codebase since about 2001
so i think i can justify it if someone complains :)
SvenDowideit goes to bed, see you all laters :)
[12:17]
jastgood night! [12:20]
Babar5.123.1 works for me
s/3//
5.14 crashes though
[12:25]
jasttry with this regex instead: /\W?(.*)/
according to docs, it shouldn't taint
[12:30]
Babarthis taints on 5.12.1 and 5.14.2
and 5.8.8
[12:36]
jastfor me it taints on 5.10.1, too
hmm, didn't taint in 5.8.9 when I tried
[12:37]
Babarthat one taints on everything [12:37]
jasthmm [12:38]
Babar$ perl -T -Mlocale -MScalar::Util=tainted -le'"foo" =~ /\W?(.*)/;die"1" if tainted$1;warn $1;die"2" if tainted$1'
foo at -e line 1.
2 at -e line 1.
that's 5.8.8
[12:39]
jastinteresting. I thought I'd seen a different result before, but now I get the same thing
anyway. lunch break!
[12:39]
Babarof course, if you use your taint method, it will die the first time, as you already access the string [12:40]
...... (idle for 26mn)
gac410MichaelDaum: I've checked in a fix I had proposed for the <nop> inside <input> But it won't handle <input ... text='<literal>...</literal>' [13:06]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/IAAF8w
foswiki/master 760708f GeorgeClark: Item11480: Fix for the <nop> inside a <input> tag...
[13:11]
***GithubBot has left [13:11]
FoswikiBothttp://foswiki.org/Tasks/Item11480 [ Item11480: Contents of HTML input blocks should not be rendered as wikitext. ] [13:11]
gac410MichaelDaum: looking at Item11480 and http://trunk.foswiki.org/Sandbox/InputTest rendered on trunk, (refresh the cache) ... the fix seems okay for those examples.
But this is a very fragile solutoinl
[13:14]
MichaelDaumoh it is the nop tag inside? [13:15]
gac410yes. [13:16]
MichaelDaumomgee [13:16]
gac410And except for our "special non-html html" nested tags are illegal [13:16]
MichaelDaumyea [13:17]
gac410And the whole purpose of the nop was to prevent rendering that this fix prevents in the first place. [13:17]
MichaelDaumtrue. [13:17]
gac410the much more difficult one is text='<literal> *bold* </literal>'. We handle literal tags differently. Change them to <!--literal(n)--> which is just as bad, and can't be fixed by the same trick.
but again, the fix eliminated the need for <literal> in that case anyway.
I think the eventual conclusion is writing a HTML parser using regex is fraught dangers.
[13:19]
MichaelDaumMichaelDaum wonders when t.f.o updates
it does work fine over here localhosty
[13:27]
gac410It's updated. Refresh the cach.
cache
[13:27]
MichaelDaumah the page cache [13:29]
gac410yeah.
I re-released your SetVariablePlugin
[13:29]
MichaelDaumthanks for that
though the "fix" looks awful
I wonder whether it would have been better to revert it to old style again
[13:30]
gac410I'd prefer to have people patch older foswiki. the HEAD stuff is developer only anyway.
I think longer run, there is an advantage to getting to perl version objects. since perl "knows" how to compare versions, we can discard a lot of complex code
[13:31]
MichaelDaumis there any hotfix possible for legacy foswikis to deal with new-style VERSIONs? [13:33]
gac410The easy fix for installs that don't do pseudo install is to edit EXTENSIONS.pm and change 'HEAD' to '9999.99_999'
The problem is not in the extensions. It's that configure extension installer compares plugin::$VERSION to 'HEAD' and the version.pm dies
[13:33]
MichaelDaumic [13:34]
gac410Do we expect that users can patch lib/Foswiki/Configure/UIs/EXTENSIONS.pm s/HEAD/9999.99_999/ or mangle $VERSION in new extensions, or abandon real perl versions?
We could also release a "FixVersionsContrib" that does the hotfix of EXTENSIONS.pm .... but if a user has already installed an extension with a new version string, it's too late.
[13:41]
....... (idle for 33mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/LL05Dw
foswiki/master d22987d TimotheLitt: Item12210: Provide a more useful (and less boring) test message....
[14:16]
***GithubBot has left [14:16]
FoswikiBothttp://foswiki.org/Tasks/Item12210 [ Item12210: Easier generic checkers ] [14:17]
gac410Item11860 is an urgent trunk task which says "note: need to test on 1.1 to make sure this is only a 1.2 trunk issue. " The test fails on Release11. Does this mean it's really a Release01x01 blocker? [14:27]
FoswikiBothttp://foswiki.org/Tasks/Item11860 [ Item11860: SEARCH Pager not working when zeroresults parameter passed to ViewTemplate ] [14:27]
foswiki_irc5Hi i got a problem with the plugin KinoSearch
i dont know how to get it into my web
[14:33]
i always get this error if i want to include it through %KINOSEARCH%
Can't locate KinoSearch1path in @INC (@INC contains: C:path . C:path C:path C:path C:path C:path-x86-multi-thread C:path C:path)
[14:38]
gac410the "path" strings are a obfuscated list of your perl INC lib path. So failures don't reveal your server file system organization to the world.
So it's saying that it can't find the KinoSearch tool in the searchable libraries
I don't know KinoSearch though, so not sure if it could be configuration or a missing dependency
[14:40]
foswiki_irc5:( [14:45]
gac410Is the "kinosearch" tool installed on your server?
(not the foswiki plugin)
What release of Foswiki are you using?
[14:46]
foswiki_irc5hm
i have to install more than the plugin?
i am using 1.14
1.15 i mean :/
@ gac410
[14:52]
gac410foswiki_irc5: KinoSearch is a "hook" for the Kino Search engine. http://search.cpan.org/~creamyg/KinoSearch-0.315/
It might require a c compiler to get the kino search engine installed on Windows.
(from some google searches. I have no first hand info unfortunately)
[14:55]
foswiki_irc5so i have to install Kinosearch via install Kinosearch ? [14:57]
gac410Yes the Foswiki KinoSearch plugin lists a dependency for "KinoSearch1" which is an external (to Foswiki) tool.
Ah.. I gave you the wrong CPAN link.. http://search.cpan.org/~creamyg/KinoSearch1-1.01/
[14:59]
foswiki_irc5hm am i able to install that cpan thing manually? cos i always get as maketest: no [15:01]
gac410I think that it needs a C compiler on Windows. [15:02]
foswiki_irc5hm [15:03]
gac410Microsoft Visual C++
or something
[15:03]
foswiki_irc5in generall for cpan installations? [15:03]
gac410No. CPAN very seldom needs a c compiler [15:04]
jastexcept in the many cases it does ;) [15:04]
gac410Ah... googling around ActiveState has a KinoSearch1 package. Not sure about Strawberry. [15:04]
foswiki_irc5ok so i will install microsoft visual c++ and do i have to link cpan to it in any way? [15:05]
gac410http://code.activestate.com/ppm/KinoSearch1/
I have no idea. I don't do any windows development at all.
[15:05]
foswiki_irc5:(
i am using strawberry
well i am using the perl from xampp
so i guess its strawberry?
[15:05]
gac410I'm really not sure. I have a windows XP vm that I occasionally boot for a quick vanilla foswiki install test, but nothing else windows here. [15:08]
foswiki_irc5ok [15:08]
............................ (idle for 2h19mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/aOF0Yg
foswiki/master ef806cf TimotheLitt: Item12180: Template and expand diagnostics for feedback protocol error messages....
[17:27]
***GithubBot has left [17:27]
FoswikiBothttp://foswiki.org/Tasks/Item12180 [ Item12180: Implementation for AJAXOnDemandCheckersForConfigure ] [17:27]
.... (idle for 15mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/shXPGg
foswiki/master 7bd5007 TimotheLitt: Item12180: Typo in comment....
[17:42]
***GithubBot has left [17:42]
...... (idle for 26mn)
gac410Foswiki-1.1.6_100 alpha is built. I'm just going to install on foswiki.org. I won't be making it available though. [18:08]
Any objections to my running the upgrade script on foswiki.org [18:20]
......... (idle for 44mn)
well that was strange. did someone just shut down apache on foswiki.org? [19:04]
Babarsomething I guess. [19:14]
gac410I didn't restart it, it just came back. And I was doing some local stuff in my own account, but nothing that should have effected anything [19:15]
Babarwhat do you mean it just came back?
Monit monitors that it's constantly running
so you cannot shut it down like that, you need to have monit do it, otherwise it will come back
[19:15]
gac410I didn't do anything.
I didn't shut it down either. but "top" showed almost nothing running for a while. "who" showed only me logged in.
[19:16]
Babarotherwise, as I said, we have scripts that monitors the load, and if it gets too high, it restarts it again [19:17]
gac410Ah... didn't think ot check email. There it is. checkload messages [19:17]
Babaryes
the load is at 7
your cp is killing it :p
[19:18]
gac410hm. my running a cp will cause apache to restart?
anyway, my cp is done. And if things are sane, time to run the upgrade script.
[19:18]
Babargac410: it could, yes.
cp might artificially increase the load, as it slows down the machine
and if the load goes too high, checkload will restart apache
[19:20]
gac410when in doubt, blame apache I guess [19:20]
Babarapache is pretty much the only thing running there :) [19:21]
gac410Though at the time of the checkload errors, I don't think the copy was running. I was doing a trial upgrade, but all that does is diff the files in the distribution and print the results. [19:21]
Babarcould be just bad timing :)
if the load was already high, just a bit more would be enough
[19:23]
gac410Okay. I'm going to go for it - Copy 1.1.6_001 onto f.o [19:23]
Babargood luck.
oh, EditTable is broken on trunk btw
[19:24]
gac410:P Is it broken on Release01x01 branch too? [19:24]
Babardid you just restart it?
dunno. Didn't test :)
[19:25]
gac410Or is that something I'm about to find out. No, I've done nothing yet.. Still building up courage [19:25]
Babartry EditTable locally :)
ok, so it got restarted once more :)
I guess that proves it wasn't you
[19:25]
gac410hm. Yeah.. this time I was only pondering the command I was about to enter.
gac410 now goes off to figure out EditTablePlugin
[19:26]
Babarhttp://www.80legs.com/webcrawler.html is hammering the site
man... I hate it when stupid people program stupid software
[19:27]
gac410well, Twisties are not working on my test 1.1.6 install. Sigh... I guess no upgrade right now. [19:29]
jastBabar: introducing... fail2ban ;)
(seriously, it's a great help with running public servers)
[19:31]
gac410yup. But not available the way we are currently hosted, iirc [19:32]
Babarjast: read the page. fail2ban won't help [19:32]
jasttoo distributed? [19:32]
Babarand gmc promised me EONS ago (plus a week) to install it.
yes
[19:32]
jast:( [19:33]
BabarBlocking our web crawler by IP address will not work. Due to the distributed nature of our infrastructure, we have thousands of constantly changing IP addresses. We strongly recommend you don't try to block our web crawler by IP address, as you'll most likely spend several hours of futile effort and be in a very bad mood at the end of it. You really should just include us in your robots.txt or contact us directly.
I already changed the robots.txt with a nice comment
[19:33]
gac410Babar, EditRowPlugin - It seems functional on 1.1.6, except I can't drag rows. Twisties fixed by refreshing browser. [19:34]
Babaryou see the buttons? [19:35]
gac410Add row, save, cancel, delete last etc?> [19:35]
Babaryes
didn't see them on trunk yesterday
[19:36]
gac410yup - work on 1.1.6... and I didn't understand move. Click and clikc on target works. I was trying to drag. [19:36]
Babaryeah. click-o-drome :) [19:37]
gac410so I guess I can run the upgrade script. I'm also going to have it remove files that were not shipped with new foswiki. WikiReferences, and old JQueries [19:37]
Babargood [19:38]
gac410okay up and running on 1.1.6_001
Sanity testing welcomed :)
hm. restarting again?
seems slow but it might be the bots.
[19:48]
Babarhow many freaking crawlers are there out there?! [19:58]
gac410more than we want. [19:58]
Babarhow it's ahrefs, and googlebot
oh, and baidu too :)
[19:58]
............ (idle for 55mn)
GithubBot[foswiki] FoswikiBot pushed 1 new commit to Release01x01: http://git.io/1z8ZfQ
foswiki/Release01x01 bd49aee GeorgeClark: Item9526: xgettext run...
[20:56]
***GithubBot has left [20:56]
FoswikiBothttp://foswiki.org/Tasks/Item9526 [ Item9526: Translation work (needed for Pootle auto-checkin - please don´t close) ] [20:57]
Babarhum... why didn't pootle commit back the translations?! [20:59]
gac410No idea. Should I hold off on a community announcement for translations? [21:00]
Babarseems Stefanos changed something not so long ago
let me double check it
[21:00]
gac410okay thanks [21:00]
Babarit seems tobe working ok
Foswikirev:15681 proves I was mistaken: It did commit it!
[21:02]
FoswikiBothttp://trac.foswiki.org/changeset/15681 [ Changeset 15681 – Foswiki ] [21:02]
Babareven though the timing looks odd
anyway... it's up and running and looking good
so feel free to declare string freeze whenever you want
[21:03]
gac410okay great. Yup... email will go out shortly. [21:04]
........ (idle for 39mn)
ArthurClemensI really have no idea why (again) I log in to pootle and I can only suggest not save
I remember this happened last time, but no idea what solved it
[21:43]
gac410Are you in the right project [21:44]
ArthurClemensok, so "Everything" is not for me
just "Foswiki" works
[21:44]
gac410You need to be in the foswiki project. Yes.
Not sure who / what is everything. Maybe the start of being able to translate extensions?
[21:44]
Babaryes
but I never finished it
I agree it's misleading
[21:46]
gac410It's a race. Lars has 60 words left, Arthur, 75 [21:55]
ArthurClemensprogrammer talk are the hardest
"Plugin module was not defined."
done!
[22:00]
gac410Excellent. Thanks Arthur [22:07]
............ (idle for 56mn)
I've been pondering the issue we've encountered using HEAD as a version to flag a pseudo-installed extension. What about creating a dummy Contrib with a PREINSTALL exit which patches the EXTENSIONS.pm and Dependency.pm
It could have a PREUNINSTALL which reverses the patch.
[23:03]
....... (idle for 33mn)
BabarLars is done too :)
now let's see if it will get committed back in 17 minutes...
[23:37]

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