#foswiki 2016-02-15,Mon

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

WhoWhatWhen
***ChanServ sets mode: +o gac410 [00:37]
ChanServ sets mode: +o gac410 [00:51]
........... (idle for 53mn)
rathier_afk has quit IRC (Ping timeout: 276 seconds) [01:44]
............ (idle for 58mn)
gac410 has left [02:42]
ChanServ sets mode: +o Lynnwood [02:56]
..... (idle for 24mn)
ChanServ sets mode: +o gac410 [03:20]
..... (idle for 24mn)
ChanServ sets mode: +o gac410 [03:44]
.... (idle for 15mn)
ChanServ sets mode: +o gac410 [03:59]
.... (idle for 17mn)
ChanServ sets mode: +o gac410 [04:16]
..... (idle for 24mn)
ChanServ sets mode: +o gac410 [04:40]
................................ (idle for 2h39mn)
GuilainC is now known as GuilainC_away [07:19]
ChanServ sets mode: +o CDot [07:27]
....... (idle for 33mn)
ChanServ sets mode: +o MichaelDaum [08:00]
............................................... (idle for 3h52mn)
favioflamingo has quit IRC (Ping timeout: 264 seconds) [11:52]
.......... (idle for 47mn)
msn has quit IRC (Remote host closed the connection) [12:39]
............ (idle for 58mn)
ChanServ sets mode: +o gac410 [13:37]
...... (idle for 26mn)
gac410MichaelDaum: you asked why demoting cache? Because from my perspective refresh= is a core pagecache parameter. The user asked to refresh the pagecache, and all is denied, and I can't reject it, so i demoted it.
Yes you've used it in other extensions, but I forgot about them. refresh= from my perspective is currently a core parameter so that's what I was looking at it from.
Not justifying that it's right, just that was my thinking. I forgot
[14:03]
MichaelDaumI agree
refresh= is a core parameter and thats fine
also fine to use refresh=foobar in other plugins
no problem here
[14:05]
gac410no it collides.
It's a problem IMO
[14:05]
MichaelDaumwhy? [14:06]
gac410it's too early to argue. Same issue where rev comment & comment both used "comment=" and were double-posting everything. It's unpredictable. [14:06]
MichaelDaumthe core will only take actions on refresh=cache(refresh the current page only) and refresh=all (purge all of the cache) [14:06]
gac410And two others ... refresh=on and refresh=fire [14:07]
LavrHi this weekend I did something MAD!
For the hell of it - I installed Foswiki....
.. on a Raspberry Pi
[14:10]
gac410cool.
Slow, but it should work I hope. The new higher performance Pi 2, or the original?
[14:11]
LavrIt runs in a Raspbian Linux
The 2
It is too slow for real use. It takes about 4-5 seconds for a virgin new Foswiki to display normal plain topics.
[14:11]
gac410When I pay my electric bill at times I think it might be worth converting my server to a Pi. ;D [14:12]
LavrI tried to enable mod_perl and that gave 3-4 seconds.
I may be able to cut more if I kill all logging because I am convinced the writing to the flash card is a bottle neck.
[14:12]
gac410y. On one I was running (not foswiki though) I logged to a nfs mount.
More to prevent the overuse of the flash card.
[14:13]
LavrIf I have have a computer with an NFS share the point of running Foswiki on the Pi is a bit silly :-)
It is a fun little device to play with.
[14:14]
gac410I was running some home automation on it for a while, but it could not keep up with the insteon / x10 messages at time. [14:15]
LavrI have two now. The first now runs KODI (OpenElec). The second I bought this weekend because I ... I don't know. [14:16]
gac410Not exactly a big ticket item :D [14:16]
LavrI made a big mistake when I concluded that Foswiki 2.1 had altered the bullet rendering. It has not changed.
I must have missed that there was a space after the bullet line with no text when I compared.
[14:17]
gac410There are definitiely some issues there though, especially with transition between bullent and : indent format. A task has been open in that area for a while now. [14:18]
Lavrand the spec has always said that a bullet requires Nx3 space * space something
The need of empty bullets is not that important. But the need to be able to have an empty indented line is real. The Wysiwyg translation EATS any empty line between two sections of bullets of indents
[14:18]
gac410MichaelDaum: The other reason I "demoted" the refresh, was because I was not certain that there was some other code down in the PageCache or it's subclasses that might look for the refresh. [14:20]
LavrAnd that prevents having paragraphs whitespace between multiple indented chapters. it works find in the NatEdit. it does not eat the empty lines. It is a Wysiwyg problem [14:20]
gac410So nothing else might see it and react assuming that the delete() was performed?
y. We REALLY need to do a refresh on TinyMCE, which will end up needing a lot of Wysiwyg work too ;(
[14:21]
LavrThe problem with eating the empty lines I am sure is in Perl land. [14:22]
gac410I suppose I ought to open a task and to the update in a feature branch, so that others can help with figuring out how to fix the exits.
Convincing TinyMCE to actually preserve white space is a rather complex mess.
special spans are inserted, and then have to be reversed back out.
Very very sensitive to minor changes. The last time I actually tried to make a wysiwyg change, in the 1.1.4 timeframe, I think it took 3-4 special plugin releases to recover from my disaster.
[14:22]
LavrI just played a little. White space between bullet sections are preserved. It is the white space before and in between indented lines that get eaten and it happens when it translates from TML to HTML loading the editor. [14:26]
gac410gac410 thinks our Development web needs some major TLC. 1) We really need to Park proposals that have sat for 100's or 1000's of days. Sooooo much clutter 2) we need a tagging facility. What are all the Store proposals, what are all the Rendering proposals, etc. [14:26]
MichaelDaumgac410, okay well. the current approach still doesnt make much sense. suddenly converting a refresh=all to a refresh=cache may come as a surprise for a user as these two are quite different functions being triggered. [14:26]
gac410MichaelDaum: right. I'll scan the code then. but are you pretty sure that if I just leave it alone and don't fire the delete, there wont be some other effect ? [14:27]
MichaelDaumbasically only admins should be allowed to interact with the PageCache's refresh functions
yea pretty sure.
the refresh url param is just for maintenance
refresh=on triggering multiple caches installed in different sub-systems might be just fine. like hit the hammer and nuck em all.
[14:27]
gac410That was my concern :D "pretty sure" isn't certain. And a partial purge could be worse than the full purge. deleting the param seemed too draconian, ignoring it might miss something, so demoting it seemed to be a safer middle ground
gac410 really doesn't like sharing parameter in multiple ways. it just feels fraught with side effects. unpredictable at that.
[14:29]
MichaelDaumno not really a safe middle ground
refresh=cache is only rarely useful
in real life refresh=all is what you'd probably do after a plugin update or so
[14:31]
gac410I use it all the time. Some of our tasks web would get out of sync. iirc. [14:32]
MichaelDaumthen somethign else is wrong [14:32]
gac410I can't explain why, but at least on f.o I've had to refresh pages
or maybe t.f.o
[14:32]
MichaelDaumthe idea is that the cache & content should be configured in a way that refreshing is performed automatically based on sufficient dependency information
ah t.f.o ... that makes sense now
the content is changing underneath its radar
via f.o
even then a refresh=all is safer as you don't know what t.f.o has cached that f.o changed without t.f.o knowning about it
[14:32]
gac410see there are reasons that I've been pulling hair after seeing pages not update, [14:34]
MichaelDaumy
but thats _only_ due to the t.f.o / f.o setup
not a conceptual problem of the page cache
[14:34]
gac410Anyway. the 15-minute git-update does clear the cache on t.f.o. [14:35]
MichaelDaumusing refresh=all I presume. [14:35]
gac410yes. I think it does a ./view refresh=all [14:35]
MichaelDaumfrom there on we are just fine
for 15 minutes
[14:35]
gac410welll It only fires the update if git has had commits.
So 15 minutes is the worst case.
[14:36]
MichaelDaumwhen an f.o user changes something t.f.o has cached ... within these time spans ... then you are stuffed again ;)
running as guest
[14:36]
gac410And the other way as well. Update on t.f.o in a cached web, and f.o is stuffed. [14:37]
MichaelDaumas fas as I can see disabling the page cache on t.f.o is probably most reasonable
^yes^
[14:37]
gac410so the "refresh it now" is really handy on f.o / t.f.o
which is why I liked my javascript solution there.
[14:37]
MichaelDaumthis all tells us: one content base & two independent page caches --> this is broken [14:38]
gac410yes. [14:38]
MichaelDaumeither they share the same page cache ... or one of them must be switched of ... which is not the one on the production site [14:39]
gac410With a "sql" database option it might be possible to use a common index, But given the rendering of the two codebases need separate page images. [14:40]
Lavrgac410 now that we have a 2.1 and master - if I have a core update (I do not right now) do I simply change branch and apply the update to both and push or is there a smarter way? [14:40]
gac410The only way is to "cherry pick" on one or the other, I think.
"merge" is too big. Can't merge single commits.
[14:40]
LavrIf I alter core it will be very minor things that can easily be manually applied twice [14:41]
gac410MichaelDaum: really need to disable cache on BOTH when shared like f.o and t.f.o Because updates on t.f.o breaks the cache on f.o just as readily as the other direction.
Lavr: yeah. I usually do a quick scan on occasion of 'git log --oneline master and git log --oneline Release02x01 to look for fixes that should have been synced.
But it's important that everyone keeps that in mind.
;)
[14:42]
MichaelDaumgac410, in general you are right. however the page cache on f.o is what saves us from bots and other bad crawlers
updates via t.f.o are pretty rare arent they
[14:45]
LavrFor HolidaylistPlugin I now run with the hack I put in the bug item where I declare the Comment Counter var as local. That isolates it and seems to work [14:45]
gac410I tend to use t.f.o on occasions. especially when getting closer to a new release.
tbh at times I don't remember which one I'm on. Maybe t.f.o ought to have a banner :D Whatever one the autocomplete in my browser url happened to find first.
So if I've been using t.f.o, or the last time I reverenced ItemNNNN in the url bar was on t.f.o, that one wins. for my next update.
Lavr: yeah. Would be nice if when calling "expand common variables" *ALL* variables could somehow be localized. Maybe that's something vrurg to think about with his new object model.
You are potentially doing all sorts of out-of-order processing by focing the macro expansions to happen out of order
[14:46]
LavrYes. And that could be made backwards compatible in the Func by adding a new function parameter which defaults to none local if ommitted.
There is probably at least one plugin that expects the global context to be applied.
[14:50]
gac410But really the worst part is HolidayList expanding the main BASEWEB.BASETOPIC while being called by core to expand a macro in BASEWEB.BASETOPIC Really ugly. [14:51]
LavrYes. I checked the examples we had with INCLUDEs and none of the included topics have a Comment tag on them. They were added by a couple of geeks that felt they wanted their own little list of vacation. [14:53]
gac410It *really* ought to use include sections, to limit the holiday settings expansions to only those applicable lines. Say an included topic had a huge search, or some other very long running macro.
Or a Image macro that updates image caches, or directed graph macro that renders a huge digraph ...
Or a SQL database macro that updated some databases
Lavr, it sounds like though that your 2.1 upgrade was pretty smooth ... so far :D
[14:53]
LavrYes. But that is like that with simple INCLUDES. The typical usage is a couple of departments that use a topic to track their vacations. And then you can make a house calendar page that includes all the others. The plugin actually has a topic parameter. I need to check if it accepts multiple topics in a list. That is the right way to do it. [14:57]
***ChanServ sets mode: +o Lynnwood [14:58]
LavrIf you use INCLUDE then you will use sections because otherwise you also include the HOLIDAYLIST tags from the other pages and that would look bad [14:58]
gac410No... this is an internal INCLUDE of other topics to get the * bullet list of holidays. It is not rendered, It is thrown away [14:59]
LavrYes. The plugin takes a list of topics. [14:59]
gac410So it wont look bad. The macros just do a lot of work that is mostly not needed [14:59]
LavrOn the topic I had there were actual INCLUDE macros on the page [14:59]
gac410right. But the holidaylist plugin internally generates a list of INCLUDES without any section limits. Including an include for the base topic. [15:00]
LavrOn the smoothness to 2.1. Yes. But that is also because I had spent months preparing it on a test site. [15:00]
gac410right. indeed it reflects a lot of effort you put in for the migration [15:01]
LavrThe biggest issue I have had was a LOT of view and edit templates used in applications that were not compatible with the NatEdit way of working (and could not easily be so that is not a critique) [15:01]
gac410I suggested that was not a good idea, he didn't seem to care. [15:02]
LavrI for sure could not allow that on mine :-)
Though I have deleted a lot of attachment ,v files in selected areas so cut down the backup size
[15:03]
gac410I suggested his users might lean toward a lynching, he said he'll check again, but he really didn't seem to care. History not needed. [15:04]
LavrI have had problems with people making a COPY of a weekly report topic once a week.
And each week they upload new metric images with a date code in the filename.
A copy also contains the old attachments so the whole thing grows exponentially.
[15:04]
gac410:) Ah the things users do [15:05]
LavrSo I spent some hours just deleting all the ,v files for all topics with a certain name sceme. And later i went on just deleting attachments. People never look at weekly reports from 3-4 years ago.
And if they do they do not care about the history of a jpeg file attached to it
Going forward I changed the app that created the new topic based on last weeks topic so instead of creating with last weeks topic as templatetopic, I create a new topic with the content of last weeks topic - without attachments at all. That stopped the nonsense
[15:06]
.... (idle for 17mn)
***TitanOfOld has quit IRC (Ping timeout: 248 seconds) [15:25]
......................... (idle for 2h1mn)
vrurg has quit IRC (Quit: vrurg) [17:26]
............. (idle for 1h0mn)
GuilainC_away is now known as GuilainC [18:26]
................. (idle for 1h23mn)
GuilainC is now known as GuilainC_away
GuilainC_away is now known as GuilainC
[19:49]
....... (idle for 30mn)
JaniHIs there a way to add an attachment to a form so that when the form is submitted the attachment is added to the topic? [20:21]
gac410JaniH: I don't think so. Foswiki:Extensions.TopicInteractionPlugin has pretty advanced upload capabilities, but I don [20:25]
FoswikiBothttp://foswiki.org/Extensions.TopicInteractionPlugin [ TopicInteractionPlugin ] [20:25]
gac410don't think it can do that. [20:25]
JaniHgac410: i've read over TopicInteractionPlugin and didn't see it mentioned...
am i hoping for too much if I try to use a Twiki plugin on Foswiki?
[20:26]
gac410Is there a twiki plugin that can do that? [20:27]
JaniHFileUploadPlugin [20:27]
gac410The TWikiCompatibilityPlugin does allow some compatibility, but it hasn't had the API updated since the fork that I'm aware of. [20:27]
JaniHI'll give it a try later.
Just a bit curious though because maybe I'm missing something...
I'd like to create a small database for materials. So my plan is to create an initial topic that contains a form that includes certain material properties as well as an ability to attach the detailed material datasheet. When the user "Submit" then an autoincrementing topic with a form attached is created and the datasheet is attached to the newly
created topic.
[20:28]
gac410Foswiki:Development/StatusOfPortingExtensionsToFoswiki shows it as "Obsoleted" but I have no idea by what. [20:31]
FoswikiBothttp://foswiki.org/Development/StatusOfPortingExtensionsToFoswiki [ StatusOfPortingExtensionsToFoswiki ] [20:31]
JaniHif it's "Obsoleted" could that mean that the function already exists? [20:32]
gac410yeah .. I really don't know. Or another extension does it. MichaelDaum might know. but he's offline at the moment. [20:33]
JaniHThanks - I'll try to catch him and ask. [20:34]
gac410He's usually online "business hours" in Germany. [20:36]
..... (idle for 22mn)
***EisNerd has quit IRC (Ping timeout: 248 seconds) [20:58]
.... (idle for 17mn)
GuilainC is now known as GuilainC_away
GuilainC_away is now known as GuilainC
[21:15]
GuilainC is now known as GuilainC_away [21:27]

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