#foswiki 2014-08-27,Wed

↑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 [00:14]
..... (idle for 24mn)
gac410Tarbox: Twisties in list: http://foswiki.org/Sandbox/TestTableCalc1 works for me [00:38]
I'm guessing that the issue is that the twisties are multi-line. TheTML bullet is a single line, so yes the twisty breaks the bullet.
or .. the bullet breaks the twisty.
Tarbox: All I can think of off hand is that someone needs to edit those topics, and insert <sticky> ... <sticky> tags surrounding those lists, so the wysiwyg editor won't convert them to TML.
A real fix would need to be a change to how lists are rendered. I'm not sure if it would be possible. Certainly it would need a feature proposal to discuss a major change to render.
Something like "Take out blocks (twisties)" then render the list, then restore the twisty blocks, then render lists again, since the twisties might have list in them.
[00:46]
Though in thinking about that. that would be pretty tricky. I doubt it will fly. TwistyPlugin uses registered macros. So a preRendering handler would have to strip out the TWISTY elements, and then render them separately.
SvenDowideit: I'm a bit with you on forking. I think encouraging people to fork and PR is a good practice. We need to discourage major breakages caused by master experiments.
I'm thinking that the getting started with extension development, getting started, and some other topics could be refactored
Maybe a "GettingStarted" topic with a pull-down. "I'm a "CoreDev / ExtensionDev / CasualContrib" who wants to "CreateAExtensino" "MakeASimpleFix" "RewriteTheWorld" ..
Probably also need to factor in, is the extension "Coordinate with author" which typically be a fork / PR, Or FeelFreeToModify, which is a fix/push for developers.
[01:01]
SvenDowideitpretty much every project on GH that i'm involved in uses the PR+review workflow for everyone
consistency being simplest
and o/ :)
[01:09]
gac410Who are the reviewers? Is there a "benevolent dictator" D
:D
[01:10]
SvenDowideitsometimes, but no
peer review
peer review with a group of core devs that do the merging
that way more people know more code
[01:10]
gac410I wouldn't object, but I suspect that would be a huge culture shock to the project, at least for the core/default exts
To me the "CoordinateWithAuthor" pretty much demands what you suggest.
We also have a social contract to take time to review / apply PR's There are two outstanding on GH that I'm not familiar with the underlying code. But someone needs to pick them up or the submitters will walk away.
The one I could validate I reviewed and applied. (SpreadsheetPlugin)
[01:12]
SvenDowideitmostly yes, its a discipline and collaboration thing that the foswiki (and twiki) core devs have not taken advantage of [01:14]
gac410I think there would be considerable pushback for current core devs to submit themselves to review. Often I would really appreciate that though. I often make changes where I want another set of eyes, [01:15]
SvenDowideitpersonally, i don't like working without review
but there is a big need for devs to consider reviewing code to be more important that writing more
SvenDowideit knows nothing - though I spend a bit of time trying to get master of distro running in a container
i gave up
[01:16]
gac410I agree. There was a recent article about git workflow, gpg signing and a site that had a backdoor inserted. by a core author who merged in code. It was an interesting read. [01:18]
SvenDowideiti'm hoping its because i've not cfg'd a foswiki for >6months [01:18]
gac410Nah. configure is TOAST . Need to have an existing LSC or you are out of luck. [01:18]
SvenDowideitoh. ok, well ping me some time when master works again, and i'll look at making a Docker container
way too many balls in the air to hack stuff where master is bust
[01:19]
gac410CDot started some really important restructuring of configure, ... but checked it all in to trunk. We were all on the verge of branching 1.2 but now that's on hold. [01:20]
SvenDowideitfun :)
tis impressive to see you've moved stuff to GH tho
argh! my OSX is now totally beachballed
[01:21]
gac410We've been sitting on that move for 4-5 years I wanted it done before we branched 1.2.
I was planning to propose the Release01x02 branch right after GH stabilized, but then came the mega-configure commit on svn. :P
[01:22]
SvenDowideityou could just revert it in git?
CDot knows way more than enough git to re-do it without losing any work
as i'm one who's long drifted off, yup
[01:23]
gac410I wouldn't revert anyway, I'd probably consider going ahead with the Release01x02 branch. Maybe branch prior to his changes, and then cherry-pick the remainder of master changes into r12 [01:27]
SvenDowideitsame thing, yup
though having a broken master is a big no-no
turns off lots of people
[01:27]
gac410I can't argue with that. [01:28]
SvenDowideitSvenDowideit thankss and congratulates gac410 on the work you're doing to keep the lights on [01:30]
gac410thanks SvenDowideit [01:30]
SvenDowideitits one of those increadibly hard things [01:31]
gac410I wonder. I'll see how crawford thinks things are going with his restructuring. If he doesn't think he can get a working configure in master shortly, we should branch his work out. [01:32]
SvenDowideitmostly, y, talking :)
and i find PR's are also good for that
[01:32]
gac410What do you think the best way to recover is. Branch master -> configwork and then revert the checkins from master? [01:33]
SvenDowideitprobably [01:34]
gac410It should be pretty painless for cdot. git stash save; git pull; checkout configwork; git stash pop [01:35]
SvenDowideiteasier than that
`git checkout -b configwork`
makes a new branch from where you are now
then revert from master
[01:35]
gac410Yeah. I would do that. [01:36]
SvenDowideitmost projects would work in the user's repo and then make a PR [01:36]
gac410I'm saying he comes back and encounters that I've whacked master. So stash any dirty index files, switch to the configwork and then reapply.
Yes indeed. I agree. That's another culture change though. Micha shot that down quickly when you suggested it. :)
[01:37]
SvenDowideiti can only really point out what is the norm outside foswiki - and thus what will be expected for others to join up [01:38]
gac410yup
forking with github is goodness, but fork has a really bad connotation
[01:38]
SvenDowideitnot in the real world it doesn't :) [01:39]
gac410well post GH anyway. Fork to the old guard suggests someone got mad and are on their own. aka tw -> fw [01:40]
SvenDowideitSvenDowideit is new here :p [01:42]
gac410For the most part, all of us core devs are very used to just ready, fire, aim. at svn Branches & forks & PR's is huge change.
When I joined fw 4-5 years ago, I was shocked when I asked and was given core commit with a "have fun" and not hardly a lick of guidance
What's your thoughts on the "distro" vs. the per-extension organization. That was another area I didn't get much feedback on so just went for it.
When I saw that Store had a history of ONE commit, I knew something had to change.
[01:42]
SvenDowideiti liked the idea that core could work without extensions
but but really, first step is to make _anything_ work :)
[01:44]
gac410That's a release / packaging question, not a repository organization.
I agree a standalone core is nice.
[01:45]
SvenDowideitie, i git clone .... core
and then ./pseudo-conf it and have something
[01:45]
gac410We were approaching 2 weeks of read-only SVN, and I couldn't figure out any other way to keep history of files that moved between extensions & core.
git bisect was useless too. bisect core didn't bisect Store which is rather integral :)
[01:47]
SvenDowideitSvenDowideit knows nothing :) [01:48]
gac410Anyway. It took over 40 hours of laptop time to generate the "distro" repository so I wasn't about to do it again. Someone else can figure it out. [01:49]
SvenDowideitgiggle [01:50]
gac410I know you weren't happy with it, but tbh I kinda liked timothe's configure. I'm sorry to see it go, but indeed getting to a RESTful configure with CLI capabilities will be nice. Configure was becoming a product in and of itself. And that's not what we are here for :D [01:51]
SvenDowideitits gone?
i only didn't like that it failed to fall back to simpler UI's
[01:52]
gac410Hm. I think so let me go look at the irclog
http://irclogs.foswiki.org/bin/irclogger_log/foswiki?date=2014-08-24,Sun&sel=12#l8 pretty much abandoned trying to save the configure UI now
[01:52]
SvenDowideitso its not just me that had issues with it huh [01:55]
gac410Different devs, different approaches. You got mad, Jast worked on it an fixed some speed issues, CDot got out the machete
To me though, such a major change will probably set us back another year. And Micha wanted 1.2 to get out quickly.
[01:56]
SvenDowideitthats another reason why so many projects use a monthly (ie short) release cadence [01:59]
gac410At least what we had worked [01:59]
SvenDowideitmaster has to work, and stuff only gets released when its done [01:59]
gac410Exactly.
I was thrilled. We started bi-weekly release meetings - 5-10 attendees, And had only a few blockers before branching 1.2 and Wham...
[01:59]
Anyway. this coming monday should be our next release meeting. I completely forgot about the last one, due to the git migration mountain. [02:08]
SvenDowideitgrin [02:08]
gac410I'll chat with cdot tomorrow if he's around. see what his thoughts are on getting a working master quickly. and if he would object to me creating a configure branch and restoring master to a usable state.
I agree with gh move, a non-working master really is horrible pr.
I should add a README to distro that tells users to checkout Release01x01 until we get master fixed.
[02:09]
SvenDowideityes, you need a README.md anyway
more people will find it via GH than any other way at some point
[02:16]
gac410yup. In the long run I think that this is a great move.
gac410 wonders what to put in the README
[02:17]
SvenDowideiteverything :) [02:19]
gac410sigh [02:19]
SvenDowideitisn't there a readme in core? [02:20]
gac410yeah. I was just looking for that. Though it's not in .md format ... whatever that is.
hmpf.. nope. I think that the release process generates it ... maybe.
We generate an INSTALL.html in our release package. No readme though.
[02:21]
SvenDowideitdelete all the things! :) [02:23]
gac410We probably need a README customized for the GH source. Maybe something like our GettingStarted doc on f.o/Development
Hm. If I edit a README directly on the github ui, will internally it do a push and drive the push webhook
Hm. probably start with http://foswiki.org/Development/GitBasedInstall
README.md "Markdown" format. Do we have a TML -> MD converter somewhere ?
[02:24]
SvenDowideitiirc, you can make it a README.txt too
but then you don't get as pretty a thing
dare you to use TML :)
[02:28]
gac410well md would be nice, especially if it's similar to TML. NO idea though. [02:28]
SvenDowideitits not
# h1
[02:28]
gac410:( :P [02:29]
SvenDowideit##h2
- bullet
1 numbered
[url links](http://stupid.stuff)
I have some very rude things to say about md
[02:29]
gac410Well that looks like it could be somewhat converted s/^---+++/###/g ... [02:30]
SvenDowideityup [02:31]
gac410Does it support any inline html <pre> ... [02:31]
SvenDowideiti think yes, but i'm not 100% sure [02:32]
gac410okay.... Digging through wikipedia now. I'll try to convert our GitBasedInstall doc just for jollies. [02:32]
SvenDowideitwatch out
GH has their own variation
md is a loose non-parseable standard
[02:32]
gac410yes. I can see. WP says H1 is line with ========= line below it. [02:33]
SvenDowideitwhereas TML has a very ridgid non-paresable standard (goverened by its 2 implementations :)
### is the more modern version
the underline one is old and horrid
[02:33]
gac410y. that's for sure. Might as well just type it in ascii
okay verbatim becomes ```
[02:35]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/gA7U7g
distro/master 9380e6d George Clark: Item11267: Add a README.md
[02:44]
***GithubBot has left [02:44]
gac410Hm... first pass doesn't look too bad. [02:45]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/PCM7CQ
distro/master 38893ba George Clark: Item11267: More README changes
[02:50]
***GithubBot has left [02:50]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/nBrdJg
distro/master 1d5298a George Clark: Item11267: More updates to README.md
[02:55]
***GithubBot has left [02:55]
gac410Okay. That wasn't all that hard. The readme looks okay I guess. [02:56]
...... (idle for 25mn)
SvenDowideitoh wow, dense text
does the ./pseudo-install auto conf still work?
[03:21]
gac410hm. It works for unit tests, it's never worked for me for live apache. [03:22]
SvenDowideitmmm, i guess thats like asking if plackup foswiki works [03:22]
gac410I use ./pseudo-install -A for unit tests. But it doesn't set the url paths correctly for me to use it with apache. [03:23]
SvenDowideiti thought that was auto-detected [03:23]
gac410I'm trying to clean up some of the text. [03:23]
SvenDowideitmmm, it used to use whatever was incoming, unless configured to be more retentive [03:23]
gac410No idea. I use a lot of vhosts though Never has worked for me. [03:24]
SvenDowideitah, plackup will rule
can't recal what didn't work, most of it was good tho
[03:24]
gac410I usually forget, and after a minute or so realize the web is killing itself trying to find all the images / js / css from a non-functional path. [03:25]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/IGunlQ
distro/master afde7d5 George Clark: Item11267: More README.md changes
[03:30]
***GithubBot has left [03:30]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/LX7K7A
distro/master 3d2a437 George Clark: Item11267: Messed up the numbered lists
[03:33]
***GithubBot has left [03:33]
gac410Ah geeze... Every time I follow a link I find another topic with subversion instructions. http://foswiki.org/Development/UnitTests#SettingUpATestEnvironment
It's a wiki. If it's worth saying once, it's worth saying it 100 times in 100 different places :D
These Wysiwyg eats whitespace bugs are really annoying.
[03:34]
................ (idle for 1h16mn)
SvenDowideit: Started a new document http://foswiki.org/Development/WhenToBranchWhenToFork
Maybe a place to start to discuss workflow on github. That's not something we've ever really talked about.
Anyway... 1am here .. time to sleep. g'night all
[04:55]
***gac410 has left [04:56]
SvenDowideitnite :) [05:04]
......... (idle for 40mn)
***ChanServ sets mode: +o CDot [05:44]
.... (idle for 16mn)
ChanServ sets mode: +o MichaelDaum [06:00]
.................................................. (idle for 4h9mn)
ChanServ sets mode: +o pharvey [10:09]
......................... (idle for 2h4mn)
ChanServ sets mode: +o gac410 [12:13]
...... (idle for 28mn)
gac410Hi MichaelDaum ... added some comments on your comments. :)
CDot: How are you making out. ... I guess Sven tried distro master for a while yesterday with docker, and gave up not realizing that configure was busted. I added a README.md to distro recommending not using master for a while.
[12:41]
MichaelDaumgac410, yes, I like branches. [12:43]
gac410I'll try to gradually bubble the discussion back up into a "suggested" workflow. [12:44]
CDotgac410: pretty good. After a fight with JQueryPlugin I have the UI looking good. Clean and simple. [12:44]
gac410excellent. [12:44]
CDotjust making sure that all the wizards are callable (though they may not do the right things)
then I have to make sure 'save' is well behaved
[12:44]
gac410I hate that we have to ship a 2nd copy of jquery with configure .. have you smashed that one :) [12:44]
CDotoh, and there's an outstanding question. What happens when a web-configurer loses the configure password? [12:45]
gac410vi LocalSite.cfg ? [12:45]
CDotCDot is using ConfigurePlugin - it's a fully-fledged Foswiki plugin, so no duplication
vi isn't available if you are on a hosted site
[12:45]
gac410Ah... hm... well back doors are probably a really bad idea. :D
hosted sites ftp get / fix / ftp put ?
[12:46]
CDotI agree, but the altervatives are hairy
e.g. email password reset ... and if your email config is snafu?
[12:46]
MichaelDaumfactory reset [12:47]
CDotCDot thinks you will understand the 'new' configure arhcitecture much more easily, and be able to do smart stuff with wizards [12:47]
gac410I think if config password is lost on a site without local editing, then the only solution is to use hostiung tools to reset or edit externally. Absolutely in no way should we ever insert a "secret reset" [12:48]
CDotok, good, I can live with that. the ultimate "reset" is to simply delete localsite.cfg
which will kick fw back into bootstrap mode
i wondered about restoring backups... but it's all to hairy
[12:48]
gac410yeah... I think that the backups are good / necessary, but restore is a task left for experts. No reason to make it in the ui. Hmmm though maybe a "rollback to last good config" might be a nice toucn. [12:50]
MichaelDaumnice feature ... too late for 1.2.0
we should keep in mind to better strip down on features rather than pile up more in order to follow a SPR (shortest path to release)
[12:51]
gac410yeah. I'm really concerned that this config restructure is too much for 1.2. As much as configure is too slow / bloated. Another 6 month cycle of trying to get a stable 1.2 is just too too much. [12:53]
MichaelDaumya. [12:55]
gac410I was really close to pushing the "branch 1.2" button so we could have a stable place to work on 1.2 alpha. And now I'm not sure what to do.
This is a good story about poor review processes ... http://mikegerwitz.com/papers/git-horror-story
[12:55]
CDotyou can always revert my checkins. But iof you do so, you will create a *horrendous* problem for later versions that will probably never be fixed.
it is my opinion that configure (as it was in 1.2.0) is unreleasable, unmaintainable, a horror show
if you were to revert to the 1.1.9 version of configure that would at least be somewhat maintainable
[12:57]
gac410how close do you think you are to a releasable configure. [12:59]
CDotvery close to checkin now. To release? Hard to say
but the new structure is so much simpler than the old, that it doesn't need vast expertise to get it ready
[12:59]
gac410Reverting to 1.1.9 configure would probably be way too difficult. Way too many checkers and config.spec files would break I suspect. It would certainly be very painful. [12:59]
CDotit has been very, very painful getting to where I am. i appreciate you need to see code; I'm working flat out, [13:00]
gac410okay that's fine. We did the github migration just a bit too late for you. This would have been a perfect use for a configure branch / merge strategy. [13:01]
MichaelDaumCDot, this sounds good.
I am never a friend of reverting ... even less so close to done-job.
[13:01]
TarboxWhy is it too late to create a branch? [13:02]
gac410Right I'd never revert directly. I don't want to cause cdot pain. I was thinking of two possible solutions Either branch 1.2 at the commit prior to CDot big breakage, and cherry pick the important part. [13:03]
MichaelDaumnot sure cherry picking will be possible
even less whether that creates more work than finishing current progress
[13:04]
gac410or maybe branch master -> configWork with coordination with cdot, and then revert master. So cdot has a uninterrupted stream and can keep plugging.
But yes. ... probably best to just have cdot finish and us get the f out of the way. :D
[13:04]
MichaelDaumget people test it out as easily as possible [13:05]
CDotso close.... there will be bugs, for sure, but your help in finding and fixing them is essential [13:06]
gac410Tarbox ... The issue is that 3 or 4 of us have been interleaving commits into trunk. If we were going to have cdot work from a branch it needed to be done before he started. And that was back on svn. So water under the bridge. [13:07]
CDotcherry picking is where I started. I just couldn't manage to do it [13:07]
gac410yeah. with your personal deadline looming ... I'll stay out of your way. Just point out where / when you want help [13:07]
CDotI need MichaelDaum to look at the load time of JQueryPlugin. but that can wait until checkin. [13:08]
gac410And owe you a virtual beer for tackling a huge and badly needed task
btw. another interesting foswiki site discovered: http://tig.csail.mit.edu/wiki/TIG/WebHome Someone pointed it out on IRC week or so ago.
[13:08]
CDotthe "brokenness" is in terms of it's future. The whole AUDIT business .... mind-bogglingly complex, total overkill, and badly broken
however I have been able to rescue most of the useful code from it
the UI I just had to bin. Imagine a tin of spaghetti. No, a bucket. No, a skip, with tomato sauce.
[13:12]
MichaelDaumgac410, ssl error on tig.csail.mit.edi :( [13:13]
jastI remember the UI fairly well :) [13:14]
gac410I think the audit was mostly timothe's attempt as speeding configure up by deferring all the checkers that run for every touch. To me the one very useful part was the checking of all dependencies including extensions.
deeper dependency checking was needed. I hated asking people to re-install a plugin to get a dependency report :)
[13:14]
TarboxMichaelDaum, it worked for me. Looked slick too. [13:15]
gac410worked for me too. No ssl on my view. [13:15]
MichaelDauml&f breaks on System.webHome [13:15]
TarboxAh, yes. No ssl.
L&F?
[13:15]
gac410Anyway rather than dissecting their web acumen, I was just pointing out another foswiki site. Ah... indeed they have some issues
Oh... and for audit, deferring the deep file system permission checks was useful too. 1) They were slow, 2) they were wrong on AFS / SELinux / other alternative FS, and 3) did I mention they were slow :D
[13:18]
MichaelDaumand pointless all too often [13:24]
gac410Anyway ... I think we are over the hump on github migration We need:
- review / cleanup of Development docs - create a github mirror process on f.o - Get "Weblate" running on translate.f.o ... Cleanup of the dead repositories
[13:25]
CDotgac410: the idea was good, but the implementation was a bear. Passing the entire config hash for every REST request... home-brewed dispatcher... checkers with guessers interleaved in the code..... mutually recursive checks..... complex an inconsistent language for specifying checks (mostly undocumented) [13:27]
gac410yeah. indeed. The "passing entire hash" was on his list to remove. He tried too hard to "remodel inflight" rather than landing for an overhaul I avoided touching configure even before he started. I always got totally lost in the UI/Checker/Visitor code.
Sometimes it's better to just start over and pick the good bits.
The shopping cart was way overkill IMO. Cute .. but really?
Can anyone think of any reason NOT to just delete the old github repos CommentPlugin PatternSkin ... etc. The stuff that is now included in distro ?
[13:29]
TarboxIf there is more than one commit to them you want th ehistory. [13:33]
gac410They are in distro with history. Their problem was that when they were split out from the original monolithic foswiki repo, they lost their history.
They should not have any commits that are not distro If they do, then we have a problem.
[13:34]
TarboxWell the question is if there are any commits made before moving it to distro. But I think you just said no to that two sentences ago. [13:36]
.... (idle for 19mn)
gac410Tarbox, there certainly should not be.
The only other repository to consider is the original foswiki/foswiki That's a svn clone that stopped getting updated with the last svn checkin. So we should probably rename it to "svnArchive" or just delete it and keep an archived copy on foswiki.org server.
After reading MichaelDaum's comments about bitrot on gh, we should probably just delete it.
The main downside there is that is the repo that people probably forked for local work.
[13:55]
***ChanServ sets mode: +o Lynnwood [13:58]
GithubBot[SyntaxHighlightingPlugin] fschlich pushed 1 new commit to master: http://git.io/fhKPbg
SyntaxHighlightingPlugin/master fe277c5 Florian Schlichting: Item13000: Fix "Use of uninitialized value $params{"num"} in lc..."
[14:03]
***GithubBot has left [14:03]
...... (idle for 29mn)
gac410Is there an advantage of running tests by using cd YourPlugin/lib/Foswiki/Plugins/YourPlugin && build.pl test I've always found it much easier to cd core/test/unit && ../bin/TestRunner.pl YourPlugin
That way I'm already in the directory with the tests and can just edit / test / edit / test
Other than reviewing the docs, I've never used the "test" target of buildcontirb
[14:32]
........ (idle for 35mn)
MichaelDaum: I think you were one of the few using the _allFoswiki repo I created. Do you still have a need for it, or can I send it off to the bitbucket in the sky [15:07]
MichaelDaumno I never used _allFoswiki
I created a super project of my own
and linked in repos as needed
alas that doesn't work out anymore with the new structure. so I have a bunch of repos in subdirs without a supre project to coordinate them.
[15:18]
***hardcore has quit IRC (*.net *.split) [15:20]
..... (idle for 23mn)
gac410MichaelDaum: Probably a simple shell script to look for a .git dir in each subdir could do similar to the submodule foreach [15:43]
MichaelDaumthats what I did [15:48]
....... (idle for 30mn)
Tarbox1Can someone answer quickly which protocol you're using for check ins to github? ssh? Trying to future proof the possibility I make a contribution into our firewall. [16:18]
gac410ssh if you set up keys.
https but you have to login with user/pass to push, though there are ways to cache it.
It's relatively easy to change by just setting the remote.
[16:18]
........ (idle for 36mn)
***Tarbox1 has left [16:55]
.......... (idle for 46mn)
Tarbox has left [17:41]
.................................................................. (idle for 5h29mn)
ChanServ sets mode: +o Babar
ChanServ sets mode: +o Babar
[23:10]

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