#foswiki-release 2018-02-05,Mon

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

WhoWhatWhen
MichaelDaumHi there [13:02]
gac410Hi everyone.
Are we ready to start? How is Fosdem going/
[13:02]
MichaelDaumIt is over by now.
took place Sat+Sun
[13:03]
gac410Ah. Did you make it there? [13:03]
MichaelDaumerm
Didn't you noticed that we had a joint booth with Tiki and Xwiki?
https://foswiki.org/Community/Fosdem2018
[13:04]
gac410Yes I heard. Wasn't sure who was at the booth. Though if not you ... ;) [13:05]
MichaelDaumI was on the booth. The only one from Foswiki. [13:06]
gac410Anyway, how was the it. [13:06]
MichaelDaumI repeatedly blogged and twittered and asked here on IRC for people to join the initiative. Alas no feedback whatsoever. [13:07]
gac410That's too bad nobody else from the community could make it. [13:07]
MichaelDaumIt shows how dysfunctional we are
as a community
anyway
[13:07]
gac410Much traffic / interest at the booth? [13:08]
MichaelDaumit was fun nevertheless, though a bit intense with all those geeks visiting the fosdem
I've never seen so many IT geeks in one place.
absolutely overcrowded ... you constantly bumped into somebody
[13:08]
vrurgHi all [13:09]
gac410Cool. Well it was great that you could make it and staff the booth. [13:09]
MichaelDauminteresting talks. [13:09]
gac410Hi vrurg [13:09]
MichaelDaumthe best of it was meeting the other communities
while I was looking forward to meet the Tiki guys again, I was most exciting meeting the Xwiki people: very nice, very friendly, lots of interesting conversations with em.
[13:09]
gac410nice [13:10]
MichaelDaumLudovic, the CEO, is awesome. As well as Vincent Massol, the CTO.
I've demoed Foswiki to them
[13:11]
gac410CEO/CTO of ... Xwiki? or Fosdem [13:11]
MichaelDaumCEO and CTO of Xwiki
https://twitter.com/vmassol and https://twitter.com/ldubost
[13:12]
gac410Good size organization then. They've made Xwiki into paying software. [13:13]
MichaelDaumI had a very exciting conversation with Anca Luca (https://twitter.com/lucaa) about OpenFoodFacts (built with perl)
well Xwiki is open source, but not the way we do it.
was interesting to talk to them about their business model, i.e. how features are evolving.
next to us were CiviCRM, a crm system for non profit orgs. interesting as well.
https://civicrm.org/
[13:13]
gac410So XWiki seems to be somewhat syntax agnostic? Feed it TML or feed it Confluence or ... .... [13:18]
vrurgMichaelDaum: would you somehow summarize your Xwiki talks? It is definitely not the IRC format. [13:18]
MichaelDaumonly sour point is that I didn't have any giveaways for people visiting the booth. I simply wasn't prepared well enough. [13:18]
gac410Well our marketing team doesn't give you much support. :( [13:19]
MichaelDaumFroscon would be another fair to organize a booth for...but I am not going to make it all alone, nor without any substantial material to garnish the booth.
if we want to grow the community, then we have to do events like these.
but right
[13:21]
gac410y, or pull in some new developers from the existing installed base community. But so far we don't find too many willing to give back to the community. Really vrurg has been the only new blood in a long time.
once in a while we get a github related nibble, but nothing seems to have developed :(
Well anyway ... onto tasks. I discovered a bug in extension installer. Perl has removed "." from the default path for "do" ... so "do tools/configure" in extender.pl does not work. Needs to be either do "./tools/configure" or resolve to the explicit filename
[13:24]
MichaelDaumFoswiki isnt visible enough.
one thing that I noticed while having seen and heard about Tiki and Xwiki is: we are actually quite good in what we do
[13:26]
gac410Also discovered, trying to build a new site. That if Config.spec is empty, the extensions installer fails to save the config, if ASSERTs is enabled in LocalLib.cfg
That is a lot of your extensions Michael, Empty Config.spec stops the install when the "do Config.spec" fails.
It might be best to NOT include any empty Config.spec file in the manifest. Or just not check them in.
[13:27]
MichaelDaumsorry. let's get back to our release meeting.
empty Config.spex
[13:32]
gac410The failure is in Configure/Load.cfg It does a "do Config.spec" to pull in the spec. And do fails for empty files. But most times the failure is trapped and ignored. With debug enabled configure aborts the install.
I was scripting a CLI install for all the extensions for NatSkin, And couldn't figure out why none of them were doing their "save config" to pull in the defaults.
[13:33]
MichaelDaumdoes it only happen in debug mode? [13:36]
gac410Yes. Otherwise the do fails, and we ignore the failure.
The test site I was playing with happened to have asserts enabled.
[13:36]
MichaelDaumdo 'empty.pl'; is just fine
do 'empty.pl' && continue(); won't call continue()
[13:37]
gac410right
In 2.1.5 it's lib/Configure/Load.pm line 129
[13:37]
MichaelDaumthe default Config.spec has got a 1; at the end ... which we better should not rely on to always be there [13:38]
gac410Fails here: https://github.com/foswiki/distro/blob/master/core/lib/Foswiki/Configure/Load.pm#L129
Well, the block following it.
[13:40]
MichaelDaumand a bit further down: "Pointless (says CDot), Config.spec does not need 1; at the end" [13:40]
gac410Right, but a bit prior, the WARN gets turned into a failure if ASSERT is enabled. [13:41]
MichaelDaumwhy do we have to warn and fail? [13:42]
gac410well, if the Config.spec has real perl syntax errors, we still want to know the resulting configuration is bogus.
That establishes all the defaults for the extension.
[13:43]
MichaelDauma real perl error, like an exception, should better be catched using Error.pm
try { do ./file; } catch with {...};
[13:44]
gac410Y, so maybe it needs rewriting. :p
For now and for all the older foswikis out there, may be best to not ship empty Config.spec
We could also have BuildContrib fail during packaging, and/or ignore zero length Config.spec files.
Better to catch typos during build than during install.
[13:44]
vrurgZero-length check wouldn't help. Config.spec could be empty with a comment inside. [13:47]
MichaelDaumxactly [13:47]
gac410Well 3 changes. 1) have our "skeleton" extension include a comment and 1; in the prototype spec file. 2) Have buildcontrib "do" and report issues before upload and 3) have configure be smarter about error handling. [13:48]
vrurg+,+,+ [13:49]
gac410I'll open a task later [13:50]
vrurgBTW, BuildContrib has another minor architectural deficiency I stumbled upon recently. But that's, perhaps, a matter beyond release meeting subjects. [13:51]
gac410Other than that .. and the "do " no longer includes "." current path, which breaks tools/configure, I have not heard any feedback +/- regarding 2.1.5
Only one other minor one vrurg?
[13:51]
vrurggac410: I should open a task on 2.1.5 configure issue too. Related to running it on macOS. A checker for storage encoding tries matching agains unitialized config entry. [13:53]
gac410Y, please do. [13:53]
vrurggac410: Sorry, on your question – my English falls short here. :(
Could you, pls, re-phrase it?
[13:53]
gac410It was a joke... you only found one other minor issue, which I figure is an understatement.
/me takes the blame on that one I'm sure. I did not have a mac to test against for the utf encoding test
CDot did comment on our other two blockers.
[13:55]
vrurgIt's not about encoding itself but rather about settings it to something meaningful in bootstrap.
Regarding CDot – it reminds me that my fix for EditRowPlugin being shaky/flikering is done. Works fine for me.
[13:56]
MichaelDaumvrurg, go ahead checking it in then. [13:58]
gac410Item14494 - (ProjectContributor vs UnknownUser) in PFS store, I don't really like his solution but ... [13:58]
vrurgI'd like somebody to test it too beforehand. If possible.
If nobody can do it then will merge it.
[13:58]
gac410And Item14516 - the incomplete %MACRO{} breaking tables, he mostly said that's life... if I read between the lines ;)
I can try it after the meeting vrurg
So that's our blockers, I'm not sure where to go with them
[13:59]
MichaelDaum_saveDamage is a pita
though the workaround CDot proposed in Item14494 is okay, in some sense.
Item14516 is next to WontFix
[14:00]
vrurggac410: Thanks, it in the branch. BTW, I based it on Release, eventually. :) [14:01]
gac410Y I was somewhat seeing your adventure there ;)
Y, In some ways, our macro parser really needs a redesign, but it's so fundamental to foswiki, it's mostly untouchable. The ability to create new macros out of concatenated macros, not just nested macros, is a mess.
At one point I traced the macro expansion, and in some cases it iteratively expands useless macros many may thousands of times looking for a possible closure.
Item14437: Macro expansion runs away, reprocessing some macros many times
Anyway that's it for tasks unless anyone has any to escalate.
Going once ... going twice ...
For features, there is one with the 2-week clock expiring.
https://foswiki.org/Development/EnableLowerCaseTopicNames
The last time this was tried, ages ago, Sven also changed the WikiWord definitioin to wikiWord and that broke everything. My proposal does not change wikiWord,
MichaelDaum: On another note ... I like your new proposal pages. Very nice
[14:02]
MichaelDaumgac410, thanks
we have a bit of a mixed terminology here: FeatureProposal = FeatureRequest = ChangeRequest
so I merged the first two into one
a FeatureProposal still has got a ChangeRequest form attached to it ... which kind of is confusing
[14:12]
gac410Yes. Also singular vs. plural. FeaturePropsal was where you create new ones. FeatureProposals (plural) was the mega search that I used constantly [14:14]
MichaelDaumdid you notice that I've added direct links to the Development.WebHome?
direct links to FPs in a certain stage
[14:14]
gac410Yes [14:14]
MichaelDaumI also introduced the notion of "active" vs "passive" FPs . Most of the time, as a RM, you might be more interested in active FPs, part of the ReleasePlan [14:15]
gac410gac410 just liked a catagorized single page with everything that I could scan / and use in-browser search to try to sniff out stuff.
y, the new pages are easy to work with.
[14:15]
MichaelDaumokay. good. [14:16]
gac410I used to do a lot of "ctrl-F" trying to pick out words from the search. We have so many old similar requests. They are impossible to find.
Anyway... Lower case topic names. Seems to be a very easy thing to enable, at least as experimental. I've not found any negative side effects, other than possible collisions with directory names.
[14:17]
vrurgCould collisions be avoided at creation time by simply checking webname against directories? A kind of 'reserved words' check. [14:19]
gac410I have not experimented with that yet. If the cgi script has priority over the existence of a directory, then it just works. The issue would be if apache or nginx attempts to access the directory
The *only* directory that we web server should ever access would be bin and pub. I really don't like our approach of selectively blocking directories in the web server. If you add a new directory and forget ... oops.
[14:21]
vrurgRight. [14:23]
gac410Anyway, more onto our "process"... I am very uncomfortable with our feature proposal process, where things seem to just get approved by default without any obvious review.
Any thoughs on 2.2 Michael, you still have a pretty big list of stuff you wanted in the next release. Should we set another freeze date?
I'm just puttering along. Any features that don't make the current train, I have no problem deferring.
[14:23]
MichaelDaumI was afraid of asking :] [14:27]
vrurggac410: I'm afraid, not much could be done about the process anyway. At least for now, when it's really only three of us mostly. [14:27]
gac410Well we just released 2.1.5 so holding off another 3 months is not a huge hit. Maybe May 1st freeze? Release in June/July?
2.1.0 was released Feb 2016, so we are 2-years out from our last minor release
Okay ... how about 2.2.0 feature freeze May 1st, Beta June 1st. this time no more slips. Any features not merged by May 1st 2018 will be lost to the sands of time ... or Foswiki 3.0 whichever comes first :D
The lower case topic names points out how BAD our sorting is. We really need to implement Julian's changes to dump Locales and move to a better unicode based sort, etc.
[14:28]
vrurgYou have my voice for the strict plan. It makes you all availabe to work oin 3.0 sooner. ;) [14:34]
gac410That our sort is ABCDE.........abcde is not really obvious until aTopic sorts to the end of the topic list. I debugged that sort issue thinking that store was not reading lower case names.
wrapping our sort in lc() cmp lc() fixes that somewhat, although you get mixed results of aA and Aa so a real sort implementation would be better.
[14:34]
vrurgWeird fix for mixed result: lc($a) . $a cmp lc($b) . $b
Dirty, but could serve as a quick fix until better algo is implemented.
[14:37]
gac410hm what would performance be like. I was uncomfortable even using lc() [14:38]
vrurgNo slower than a complex utf staff, I would guess. [14:38]
gac410Anyway, that's it for me. Anything else for the meeting? [14:41]
vrurgNot on my side. [14:42]
gac410MichaelDaum: ping May 1st feature freeze Last chance for features before moving onto 3.0 Are you good with that? [14:43]
MichaelDaumgac410, yes. thanks for the extended reprieve. [14:44]
gac410YW.... though I'm a bit selfish. I have a few features I want to finish up too :D [14:44]
MichaelDaumhehe
all good then
[14:45]
gac410Okay. I'm flipping the LowerCase names to approved unless there are any last minute objections.
We should press JulianLevens a bit more. I'd love to see him get the Locales mess fixed. backlinks would be nice, but really the Locales is more important IMO
but these for 3.0 ... at the earliest :(
thanks everyone. Back to the main channel
[14:46]
vrurgTotally agree. utf support is what made me to move from t* to foswiki a few years ago. Locales are no less important to have this complete.
thanks!
[14:48]
gac410They (t*) are making no movement in that direction at all that I can see.
Focus is more on trying to keep t* running on new distros. They are getting hit by all the perl changes. CPAN they are dealin with by bundling old code yeech.
[14:49]

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