#foswiki 2012-01-08,Sun

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

WhoWhatWhen
gac410Hm. I've added a test to release11 - runs fine on trunk, Fails trying to expand macros if I've used $this->CreateNewFoswikiSession()
Can't call method "dispatch" on an undefined value at /var/www/foswiki/trunk/core/lib/Foswiki.pm line 3228.
[01:18]
pharveyah, I had a fix on a branch somewhere I've not yet committed [01:19]
gac410Excellent - I couldn't see that I was doing anything in my changes that was trunk or branch dependent. [01:19]
pharveyIIRC the createNewFoswikiSession() tries to finish the current session if one isn't there
or maybe I committed a fix on trunk but didn't take it to release
tries to finish the current session *even if* one isn't there*
[01:19]
gac410I'll compare the createNewFoswikiSession. Trying to sync up the rewrite of USERINFO macro
and add a test for the anti-spam cloaking.
yeay - that was it. Thanks pharvey. can't afford to loose the hair. ;-)
[01:20]
pharveyhehe [01:24]
gac410I'll commit the changes - unit tests running now. thanks again - that saved me a boatload of time. [01:37]
pharveysorry I didn't fix it [01:37]
gac410nbd - you were here and had a real fast answer. [01:38]
......... (idle for 41mn)
GithubBot[foswiki] foswiki pushed 3 new commits to Release01x01: http://git.io/Rf792Q
[foswiki/Release01x01] Item11410: Sync trunk to release11 - GeorgeClark
[foswiki/Release01x01] Item11135: Sync changes in UnitTestContrib to rel - GeorgeClark
[foswiki/Release01x01] Item11410: Implement USERINFOisTooRestrictive fix - GeorgeClark
[02:19]
***GithubBot has left [02:19]
FoswikiBothttp://foswiki.org/Tasks/Item11410 [ Item11410: Implement USERINFOisTooRestrictive ]
http://foswiki.org/Tasks/Item11135 [ Item11135: store2: Development.SimplifyTheStoreMetaSemantics ]
[02:19]
..................... (idle for 1h43mn)
dj_segfaultHi. How can I put text across multiple lines but appear on one line in the output? Most wikis use \ or \\ but that doesnt seem to be working, and I don't see any mention of that in the formatting docs. [04:02]
gac410With WYSIWYG or with raw editor [04:03]
dj_segfaultraw
I like my steaks and my editors raw.
[04:04]
gac410hm. What context. In simple text, just use newlines and they all wrap together in regular html. [04:05]
dj_segfaultOK, I think I'm having the same confusion as the other day. When I go into the editor I press the Wiki Text button, and I'm entering Wiki syntax.
So neither WYSIWYG or Raw; wiki text.
[04:06]
gac410wiki text is what oldtimers refer to as raw
It used to be called editing in raw mode. It's what you see if you put ?raw=on on the url.
[04:08]
dj_segfaultOK, I see what happened. It turned my \ into <br/> when I saved.
I think I can get it to work right if I don't put that in. Lemme try a few things.
[04:09]
gac410yeah you don't need any special syntax unless you are insde a table. Then you need the \ to continue lines without breaking table.
Using the "Edit wiki text" link at the bottom of a topic is more direct than entering Wysiwyg, then pressing wikitext button. In some unusual situations the latter can corrupt things - if the transition to tml -> html -> tml can't deal with some of the markup
[04:10]
dj_segfaultOk, this is less than ideal. I entered a few bullet list elements with each element spanning multiple lines, and it displayed correctly with the entirety of each bullet on one line, but when I edited it again, I see the editor ate the line breaks and everything is combined onto one line in raw. I'll try your idea.
Ahhhhhhhhhhhhhhh that works better. Thanks.
[04:15]
gac410yeah maybe a bug. I just entered 4 line "this\nis\na\ntest" It displays "this is a test" Wysiwyg ran it all togther as "thisisatest" and hitting wikitext button left it that way :-(
gac410 shouts pharvey
hm. Not as bad on release11. trunk runs it all together, no spaces. release11 merges the lines but at least preserves some spaces.
[04:17]
pharveyThe normal way to continue a list item across multiple lines is to simply prefix any following line with spaces
* this
is
a
single
line
[04:23]
gac410dj_segfault: If you have blocks of tml that you want to remain exactly as entered, surround them in <sticky> .... </sticky> [04:23]
dj_segfaulttml == wiki markup? [04:24]
gac410yes [04:24]
pharveyIf escaped newlines are disappearing, that deserves a bug item, but the design of HTML2TML makes it hard to fix
so far trunk can preserve newlines inside %MACRO expressions only (IIRC)
gotta run
[04:25]
gac410seeya pharvey. the newlines are completely eaten on the trunk tmce editor from what I could tel. [04:25]
...... (idle for 27mn)
***gac410 has left [04:52]
MarianBanicahei guys... who is here? :) [04:57]
............................................................................. (idle for 6h24mn)
padraig_lennonMorning all.. Anyone here using textboxlists in data forms? I am trying to create a separate topic to System/JQueryAjaxHelper but it does not seem to want to work.. Its a brand new 1.1.4 install [11:21]
pharveypadraig_lennon: are you logged in... is jQuery-1.7.1 the selected version in configure [11:23]
padraig_lennonYes to both..
To be honest it does work somewhat.. i.e. I am able to define the data set and have updated the dataform but the autocomplete is VERY slow.. Basically does not appear until after full matching text has been added if ever.. I tried testing the group autocomplete function and that is very fast..
I think I'm missing something in my topic which is not in JQueryAjaxHelper but for the life of me a don't know what.. The JQueryAjaxHelper is VERY confusing..
I can paste what I have in the data topic if that helps?
http://pastebin.com/riY9pxqe
Does the dataform field textboxlist load ui:autocomplete by default?
I have autocomplete disabled via configure as it is stated to be defunct..
[11:23]
............ (idle for 56mn)
JQGridPlugin - ERROR: can't load jQuery plugin Grid: Can't locate Foswiki/Plugins/JQueryPlugin/GRID.pm in @INC Don't know what up here.. AM I missing something? [12:24]
.... (idle for 15mn)
ArthurClemensis grid enabled in configure? [12:39]
padraig_lennonArthurClemens - Yes..
Full error.. http://pastebin.com/hHvSM0Lx
[12:49]
ArthurClemens - Must have been a misconfiguration of my lib/LocalSite.cfg.. Have it working now.. [13:00]
ArthurClemensok
perhaps new config settings are not copied over to LocalSite
[13:00]
padraig_lennonI think it must have got messed up somewhere in the foswiki 1.0->1.1 upgrades.. Thought it better to start again.. Thanks for your help.. [13:04]
..... (idle for 24mn)
AuthurClemens/pharvey could one of you guys do a quick check for me? If you use GRID within a TABPANE the width of the grid table is messed up is it? [13:28]
pharveypadraig_lennon: define "messed up" [13:29]
padraig_lennonBasically the table does not expand the width of the tab.. On the extension page it says it should automatically expand.. [13:30]
pharveyreally? I didn't know that. But mine aren't expanding even outside any tab pane [13:31]
padraig_lennonAlso seems to interfere with other JQICON references in the same topic... They don't get rendered.. Without the grid they are fine.. [13:31]
pharveyHrm, I must admit I'm not running current JQueryPlugin [13:31]
padraig_lennonMine expand outside tabpane all right.. [13:31]
pharveyIf you create a task with a screenshot, I'll see to it when I upgrade our own site in the next week or two [13:32]
padraig_lennonok will do.. Will test more before I add one in case its something on my side.. [13:33]
pharveyFWIW my grids and tab panes seem to be working okay, but I'm using jQuery-1.6.4 [13:33]
padraig_lennonI'll test it with that version.. [13:34]
pharveyif the grid is too wide, they will spill over the right-hand edge of the skin we're using in an ugly way, I'm not sure of any simple solution to that (use smaller font? :) [13:34]
padraig_lennonIgnore the JQICON note above.. I forgot to update {JQueryPlugin}{IconSearchPath}
pharvey no difference with 1.6.4. Really weird with the grid width.. It actually works once but if you refresh the page it messes up..
[13:36]
........ (idle for 36mn)
gac410hi CDot - Didn't see that tidbit about the password manager coming :-D Rather useless tuning I'm doing?
I was sure somewhere along the way I had seen a discussion / docs about how fastcgi / fcgi would cache passwords :-(
[14:16]
CDoty? don't recall that. But all is not lost, if you genuinely cache.
though TBH I'd probably go for the splitting up the htpasswd file myself
reading the passwd file is very fast - so long as it isn't too big
keeping a large PW file in memory is just wasting that memory, IMHO
[14:18]
gac410it does seem very worthwhile to cache. The parse times can be very long.- [14:20]
CDotCDot has no numbers for the read times, but it's such a simple file..... [14:20]
gac410When I changed it from a regex to a "split" it knocked seconds off of the load time for a large file. hm... I documented that somewhere. [14:21]
CDotsure - a "large" file
but a small file needn't take so long
a balanced hash splitting the PW file into - say - 16 evenly-sized bits could work wonders
again, no numbers to support that idea :-(
[14:21]
gac410Sven would argue use a db. [14:22]
CDotyeah, but a db => a new dependency
... and micha would argue for the normalisation of DB-base "temporary" storage
[14:23]
gac410yes. And iirc, his db implementation uses a cpan module that is extremely lmited in hash support. [14:23]
CDotbut if you want a fast load, how about cacheing using Storable?
(of course, you still get the scaling problem)
[14:24]
gac410.htpasswd should be compatible with mod_auth ... so our format is rather fixed. [14:24]
CDotCDot isn't proposing changing the format
the format remains the same. It's just we store the pw for each user in a different file, selected based on a hash of their username
if the hash is balanced, then the users get split evenly over a whole bunch of smaller "buckets"
[14:25]
gac410I was referring to storable. That wold be a different format wouldn't it? [14:27]
CDotdetermining if a user exists still requires a blanket read, of course, which might be the crunch
nah, Storable would just be used to cache the text file
the text file would still be required
[14:27]
gac410seems like since we already have a hash, preserving it across sessions would not be a big effort. [14:28]
CDottrue. However splitting up .htpasswd would speed up *new* HTTP processes, as well as those which already have the cache
also, no issues with cache refresh etc.
[14:29]
gac410Had not thought of user exists - yes that would need reading the whole file. [14:29]
CDotother issues - "findUserByEmail" [14:30]
gac410User exists tests if the user has a hash in the password entry. A user defined with a null password does not exist. [14:30]
CDotUsers::Password does not have an "exists" method
so searching by email would be the only reason to load the whole DB
what's the range of characters allowable in htpasswd usernames?
[14:30]
gac410The mapper tests exists by asking Users::Password for the hash. iirc. I looked to confirm that the user would not be able to reset a disabled entry.
CDot - don't know about characters.
[14:32]
CDotok. So since it asks for a "login", then the hash algorithm would hit the right bucket, and again, the read would be minimal.
If you cached the buckets between sessions, you get the best of both worlds.
[14:32]
gac410Making the hash a persistent global seems simple. Splitting the files is higher risk. Does install have to pre-allocate them? Babar pointed out that some installs won't allow file creation in the location of .htpasswd so needs to be preallocated. [14:34]
CDotyeah, they would be preallocated.
named .htpasswd_1 .. .htpasswd_8 or some such
just bear in mind that the persistent-global approach will work, but is burning memory. Only a very few entries in .htpasswd will ever be used.
no point in doing *any( of this without some benchmark data, however.
otherwise it's just guesswork - and we've had our fingers burned by that several times before.
[14:36]
gac410I'm trying to find my loadtime benchmarking. I tested it fairly extensively to decide to change from regex to split.
I know I documented it somewhere.
[14:39]
CDotw.r.t to the DB approach, I'd use a lightweight DB such as BerkeleyDB and cache, rather than replace, the text file [14:41]
CDot has gone to make rice wine.... [14:48]
Babarsqlite?
sake made by CDot... can't wait to try that one out :)
[14:55]
gac410gac410 hates it when he can't find benchmark results ...
I *know* I tested loading large .htpasswd files - when I was benchmarking the impact of doing auto-recognizing the encoding. Auto added 1ms per 1000 entries on my system. But I also had per/1000 timings to load.
[14:58]
..... (idle for 22mn)
gac410 was mis-remembering. it wasn't seconds. On 1.0.9, loading/parsing 4000 .htpasswd entries - ~60 ms. trunk without auto-detect 35ms. Split vs. regex. [15:22]
Babarso micro-optimisation [15:22]
gac410Autodetect added 4 ms. or 1ms/1000 [15:22]
Babaranybody not using a database with more than 4K entries is... [15:22]
gac410We are at a little over 3000 now on foswiki.org [15:23]
Babarand do we see any impact? [15:23]
gac410Of course. If you are trying to get to sub-second response time, 30-40ms is still 30-40ms.
given the password file changes 2-3 times a day, and is loaded for *every* transaction, I can't see why we would not want to cache. Every bit helps.
I can't buy the memory argument, since it is loaded into memory for every transaction, it would take the same amount of memory cached or not. right?
[15:25]
Babarmain diff is that if it's not cached, it takes this amount of memory on every load, and maybe not even... do we load the full list everytime? [15:29]
gac410Yes. Full load every time.
Anyway... I have some things to do today ... I'll take this up later, and either revert my change detection code, or look into how to make it into a real cache.
it loads the cache so that findUserByEmail can search it without asking to load the whole file.
[15:30]
Babarso why would splitting the file help? It should be the opposite [15:33]
gac410agreed. [15:33]
Babaranyway, I should review the code then... but for now, I'll have a bath :) [15:33]
gac410Only helps if we delay loading the file until it's really neede. [15:33]
Babar'xactly [15:33]
gac410IMO the caching is allready built, and I've added the detect change/invalidate code. Now just need to figure out how to make it persistent.
which gets into perl module loading and globals that falls into "magic" category of my understanding.
[15:34]
padraig_lennonPing ArthurClemens [15:39]
ArthurClemenshi [15:39]
padraig_lennonHi.. Just a quick one for you.. I installed the FormPlugin via configure but it installed the test cases as well.. Should I raise a task to omit them?
How would one do this? I see the same for StringifierContrib
[15:40]
ArthurClemensthey are in MANIFEST
I can remove them no prob
[15:42]
padraig_lennonJust remove there and commit the changes?
I can raise a task for you.. Not a big thing.. Just OCD...
[15:42]
Babarbut...
if you remove them from MANIFEST, how are you going to run them? :p
[15:43]
ArthurClemensfrom svn [15:43]
Babar*how*? [15:43]
gac410Unit tests should never be in the manifest.
pseuo-install figures out to link them anyway.
pseudo
[15:43]
Babarunit tests are run only from core/test/unit, and if they're not in the MANIFEST, you won't have anything there, even if you pseudo-install.
how does it do that?!
[15:44]
padraig_lennonBabar wouldn't you be running the tests from a svn/git checkout? [15:44]
gac410I don't know, they always work for me. [15:44]
padraig_lennonBabar if you use a checkout for testing it wouldn't need to look at the manifest would it? I thought that the manifest is only for creating the packages? [15:46]
gac410okay - make a liar out of me. [15:46]
ArthurClemensstrange indeed. TablePlugin has no tests in MANIFEST, but I do remember I ran the tests [15:46]
gac410How the heck have I been unit testing. I *know* I test ClamAVScanPlugin and it also does not have tests in MANIFEST.
Did pseudo-install change recently. I'm sure it used to install tests
[15:46]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/kBDeTQ
[foswiki/master] Item11412: remove mac system directory - ArthurClemens
[15:47]
***GithubBot has left [15:47]
FoswikiBothttp://foswiki.org/Tasks/Item11412 [ Item11412: Remove Mac system directory ] [15:47]
ArthurClemenspseudo-install.pl -link does link the test dir [15:49]
Babarok, it does indeed
line 733
[15:49]
gac410strange. It does for ControlWikiWordPlugin, but didn't for ClamAVScanPlugin. [15:50]
Babarif the module has a test/unit directory, then it links them all
so, you're right padraig, it's useless to have them in the MANIFEST
my bad
[15:50]
gac410so another mystery then - why is that not working for ClamAVScanPlugin [15:51]
padraig_lennonI'll raise a task for FormPlugin and StringiferContrib so.. I don't suppose I can create one task for both can I? [15:52]
Babaryou could create it for all modules with unit tests in MANIFEST, but... if there are just 2... [15:52]
padraig_lennonI'll check it out.. [15:53]
ArthurClemensno, I've just created one for FP [15:53]
padraig_lennonok.. [15:53]
gac410Ah. It must be test/unit/<NameOfPluginContrib>/testfiles
It will not link tests that reside directly in the unit directory.
[15:53]
Babaryes gac410, just arrived to the same conclusion :) [15:54]
padraig_lennonWould it not be a better approach to automatically remove these when packaging using BuildContrib? [15:55]
gac410If someone added them to the MANIFEST maybe they want them? [15:56]
Babarexactly
maybe write some warning in BuildContrib that it's not needed
but I wouldn't filter them out
[15:56]
padraig_lennonWhy would they need tests though?
Just wondering..
[15:56]
BabarUnitTestContrib? :) [15:57]
gac410you beat me to it. [15:57]
padraig_lennonok out of my depth.. :( [15:58]
gac410pseudo-install auto-links *directories* in test/unit. UnitTestContrib ships files in /test/unit, so unless ./pseudo is changed, it needs them in the manifest.
probably just history. someone had a good reason sometime in the past ... or maybe not.
[15:59]
Babarnah...
UnitTestContrib is supposed to be unit tests for core, so they shouldn't be in some special directory
it was just the way to split unit tests from normal core code back in the days, I guess
[16:00]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/gMqXZg
[foswiki/master] Item11413: remove test files from MANIFEST - ArthurClemens
[16:01]
***GithubBot has left [16:01]
FoswikiBothttp://foswiki.org/Tasks/Item11413 [ Item11413: Remove test files from MANIFEST ] [16:01]
padraig_lennonThanks Arthur.. [16:03]
.... (idle for 15mn)
GithubBot[foswiki] foswiki pushed 1 new commit to Release01x01: http://git.io/lPIYyA
[foswiki/Release01x01] Item11411: double quotes to single quotes; integrating some earlier updates from trunk - ArthurClemens
[16:18]
***GithubBot has left [16:18]
FoswikiBothttp://foswiki.org/Tasks/Item11411 [ Item11411: Implement proposal: HTML attributes should use single quotes ] [16:18]
padraig_lennonFollowing extensions also have test files StringifierContrib/ContributorsPlugin/SetTopicValuesPlugin/JSCalendarContrib/RestPlugin/VirtualHostingContrib/OpenLayersPlugin/MongoDBPlugin/SchedulerContrib [16:21]
gac410Hm. The Extensions field in the Support QuestionForm does not seem to be working. It shows the little busy spinner for a few seconds, but never pre-fills anything. [16:29]
padraig_lennonI was getting the same issue on my site which i asked about earlier today.. Thought it was just my install.. [16:33]
...... (idle for 28mn)
gac410padraig_lennon: did you update to 1.1.4 - the autocomplete and textbox jquery code has changed - might be related. [17:01]
padraig_lennonYa.. Re-installed from scratch again this morning.. [17:02]
.... (idle for 17mn)
gac410Check out Foswiki:Support/JQueryMigrationGuide for details on changes needed for forms using autocomplete. [17:19]
FoswikiBothttp://foswiki.org/Support/JQueryMigrationGuide [ JQueryMigrationGuide ] [17:19]
padraig_lennonI'm really only using textboxlist at the moment.. [17:21]
gac410padraig_lennon: never mind - looking at the logs, you seem to have things covered.
textboxlist did change, though I don't understand the migration issues - pharvey was going to add some notes at one point.
[17:21]
padraig_lennontextboxlist - is a very useful addition. The only problem I found about using it over searches is that does not remove obsolete values. For example. If I had 10 server topics and an topic points to one of these. If the server topic was removed the user would be forced to update the value.. Might be a personal preference but it does force the docs to be correct.. Because with textboxlist the values are stored as text this won't [17:24]
.................................................... (idle for 4h19mn)
SvenDowideitgac410 and CDot
wrt large htpasswd files
my option1 is if you have lots of users, you would be dumb not to move to one of the larger scale passwd managers - like the HTTPDUsers one
but if you really really need a file (and the HTTPDUsers db and dbfile , ndb etc options are all apache compatible, so its a bit weird )
then imo what we should do, is write a grep based reader to read only the one line you need
cos parsing is pointless
grep/grep-like depending on benchmarks
once we have a real, and sufficiently tested (ie, cpan, not roll our own) l2 like cache system, then things will be even more interesting
mmm, might call that store2.1 :)
gone :}
[21:43]
..................... (idle for 1h40mn)
MarianBanicahei guys.
new entry in TOP 5 submitters :D
i mean for translations :D
[23:27]
gac410Great! Making progress on Romanian?
33% - good show!
[23:28]
MarianBanicalittle by little :d
it will take some time but i promise that i will do the whole work :d
i mean till 100% :D (maybe not the whole work by me only)
:d
[23:33]

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