#foswiki 2014-10-03,Fri

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

WhoWhatWhen
***gac410 has left [04:06]
......................... (idle for 2h3mn)
ChanServ sets mode: +o CDot [06:09]
........................................................................ (idle for 5h57mn)
GuilainChello everybody !
on my new installation on debian box, filter plugin crash
my topic use the makeindex macro of filterplugin, and i get the following errors
| 2014-10-03T12:03:13Z warning | Can't locate Text/Unidecode.pm in @INC (@INC contains: /etc/foswiki /var/lib/foswiki/lib . /etc/perl /usr/local/lib/perl/5.14.2 /usr/local/share/perl/5.14.2 /usr/lib/perl5 /usr/share/perl5 /usr/lib/perl/5.14 /usr/share/perl/5.14 /usr/local/lib/site_perl) at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin/Core.pm line 26.
at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin/Core.pm line 26
[.... lot of lines]
BEGIN failed--compilation aborted at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin/Core.pm line 26.
at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin/Core.pm line 26
[... lot of lines again]
Compilation failed in require at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin.pm line 74.
at /var/lib/foswiki/lib/Foswiki/Plugins/FilterPlugin.pm line 74
if anyone knows how to correct this, I will appreciate, and if not, I will investigating during the week-end.
[12:06]
..... (idle for 21mn)
GithubBot[distro] cdot pushed 2 new commits to master: http://git.io/tb44-A
distro/master 64d0d7f crawford: Item12952: package installer fixes
distro/master a252352 crawford: Item12952: clean up reports in email configuration
[12:31]
***GithubBot has left [12:31]
.... (idle for 16mn)
GithubBot[distro] cdot pushed 1 new commit to master: http://git.io/1pbzsw
distro/master c3cd337 crawford: Item12952: bring back onbeforeunload handler for navigation feedback
[12:47]
***GithubBot has left [12:47]
ChanServ sets mode: +o Lynnwood [12:58]
........ (idle for 35mn)
ChanServ sets mode: +o gac410 [13:33]
gac410Hi CDot .. saw your checkin, I didn't try it yet. But I note you've mixed html/css into the Package code. That will be a problem when using it from the shell.
Package is supposed to not have any markup, otherwise shell use it a problem.
ExtensionInstallerTests verifies that is all valid and will fail. :(
I for one use the shell installer more than the web gui.
[13:33]
hm. magic. I ran a shell install, and it sort of worked. ... no embedded html. Cool. Still failed with a perl error though
Can't use string ("1") as a HASH ref while "strict refs" in use at /var/www/foswiki/distro/core/tools/extender.pl line 473.
Ah... I see you've also change the shell extender to use $reporter. cool.
Hey CDot ... Unrelated to configure. You were asking about workflows and why --rebase or not --rebase. There is a good tutorial on workflows here https://www.atlassian.com/git/tutorials/comparing-workflows/gitflow-workflow/
I think we are currently using the "centralized" workflow ... the first one in the tutorial. Since github is our centralized repository.
[13:43]
......... (idle for 42mn)
CDot ... yikes, are we going to need to release every extension for 1.2? I installed FortunePlugin from foswiki.org while testing the shell installer, and now configure won't run due to invalid .spec [14:30]
..... (idle for 24mn)
hm... I wonder how many other .Spec files I've corrupted. PATH O** does not mean explicit optional, vs. PATH M** means Mandatory. ... Looks like there was never an explicit O*ptional flag, just the absence of M** [14:54]
CDotgac410: no - if there's an existing spec, the new configure has to deal with it, end of.
even those that had mayfly existences with the old "new" configure
I have not mixed html/css. The Package code is entirely TML.
Unless I missed something.
[15:02]
gac410Nope.. I am unable to read Diff's esp. before first cup of coffee.
And clearly the spec issue is my fault. There is no option "O" ... I need to fix and re-release FortunePlugin . Hopefully there are no others.
It's probably not used widely if at all. :(
[15:09]
CDotMANDATORY isn't supported by anything AFAICT
it's legal; it just doesn't do anything
[15:14]
gac410Probably good intentions that got forgotten along the way. [15:15]
CDotyes, on extender.pl, the goal is to try and use reporters everywhere so that code sharing is maximised [15:16]
gac410yup. If you want, I'll add a TestReporter to ExtensionInstallerTests and try to fix the unit tests. [15:16]
CDotyou mentioned in the last release meeting minutes that a lot remains to be done on =configure=. Do you have a list? [15:16]
gac410No, TBH, I think extension installer was the big thing. I think of that rewrite as "A lot" :) [15:17]
CDotcos while i think it needs a lot more testing, from a code persepctive I think it's all there now [15:17]
gac410Needs flags to support simulate install / dependency report only, [15:17]
CDotone thing to note is the ">" at the start of a NOTE that I added [15:17]
gac410Yes I noted that note
:)
[15:18]
CDot:-)
i wondered about the simulated install. I'd really like to provide more immediate feedbacl during an install, but it's hard
I was considering opening the wizard dialog before invioking the wizard
[15:18]
gac410gac410 finds that a big issue with extension installer is you can install what appears to be a small extension, and discover 15 minutes later that an innocent extension called out a dependency that was HUGE [15:19]
CDotsomething liek "Running the Expelliarmus Wizard"
right
but splitting an install up is..... tricky
[15:19]
gac410I've shot myself in the foot that way. When web server timed out an install due to a huge dependency.
What I wanted to do, but it was beyond me. Resolve the extensions BEFORE the install and giving the user a list of what will be installed before clicking the go button.
[15:19]
CDothmmm
CDot reads code
[15:20]
gac410Some way to auto-select the to be installed extensions. Maybe as you say, a wizard dialog, but says "This is what you will install" and shows the resolved dependencies. [15:21]
CDotyeah, i get your drift
I did conside a "F**k uff, I'm not installing that until....." option
forcing you to resolve dependencies befeore the install starts
[15:21]
gac410There is another bug lurking. in the old installer. If you check/select Ext A and Ext B, and A installs B as a dependency, B gets installed twice. Though by changing it to a pushbutton, you've resolved that. [15:22]
CDotbit more refactoring in order to get there, I think
what about if A depends on B and C, and B depends on C?
is it smart enough for that?
CDot didn't really read the logic of Foswiki::Configure::Package
[15:22]
gac410It's the NatSkin family of extensions where this can really gut ugly. Yes. Each package.pm recursively installs it's dependencies [15:23]
CDotrecursively? so it installs C twice? meh [15:24]
gac410Well maybe not. If C is there, then it won't be installed again. It's the GUI checkboxes that don'
don't re-check presence.
[15:24]
CDotok
I also made it auto-enable installed plugins. Not sure that's strictly desireable - but it can always be switched off before a save
[15:25]
gac410Yes I noticed that . Very nice. Does it auto enable and save from the shell too. That would probalby be too much, since Config.spec often needs reviewl. [15:26]
CDotno [15:27]
gac410From the GUI install, by deferring enable until [Save] That is safe behavior. [15:27]
CDotI thought so. Just wanted to sanity check. [15:27]
gac410That all sounds good. Maybe someday from shell a -e "Enable" switch might be nice, but that's lipstick [15:28]
CDoty [15:28]
gac410I guess from shell, recursive dependency installation is reasonable. You see what is happening along the way. But from web, since feedback is deferred, a suddenly long-running install (did it hang?) is less desirable. [15:29]
CDotagreed. I'm looking at this dependency issue....
a two step install - a wizard to start the installation, report the deps, offer to install the deps it can, refuse to install if CPAN deps
are missing.
[15:30]
gac410Yes, Or at least STRONGLY warn. Sometimes CPAN deps or other non-perl deps are difficult to detect. I'm thinking of the Image Magick / htmldoc / graphviz / ghostscript related extensions. [15:32]
CDottrue. A "force install" mode. [15:33]
gac410And actually if you are presenting a wizard, It would be nice to search instead of listing all 350 extensions ... fastreport is ... er ... sluggishreport at times :) [15:33]
CDotone thing at a time
fastreport *is* a search, so doing another search the same way is no biggie
[15:35]
gac410y, I was playing around with a new version that output json so we could use the jquery render templates, but I'm useless at this javascript stuff ... [15:41]
CDotno JS required for that. You can do it all using wizards.
in perl :-)
[15:41]
gac410CDot ... The old extension installer used to write the install log to working/logs/configure/SomePlugin-20141003-155122-Install.log That becomes a bit more important esp. trying to go back in time and look at applied patches from the PatchFoswikiContrib [15:46]
CDotstill does (search for LoggingReporter) [15:47]
gac410Ah... Does the ShellReporter call it? [15:48]
CDotno, it logs a different way
I struggled with that a bit
[15:48]
gac410Yeah I saw the package override in the extender [15:48]
CDotI tried not to reinvent the wheel..... but all the wheels were different :-( [15:48]
gac410Good idea ... gave me an easy way to intercept the logs in the ExtensionInstallerTests
I'll ponder the logging a bit. It would still be nice if all installations were consistently logged.
[15:49]
CDotindeed. The patterns are there, should you want to share code. [15:51]
gac410gac410 finds as he gets older trying to remember / find / reverse wtf he did or was thinking gets more and more difficult.
Same with adding a comment when saving from configure.
[15:51]
CDotkeep this ideas, that's what featurerequests are for
can't get to them all; priority just now is working code
[15:52]
gac410gac410 discovered he was to blame for a config change that broke stuff. 6 months later, not a clue as to why I change it. [15:52]
CDot:-) [15:53]
gac410Right now the logging of changes is still missing. I was digging into that a bit. when you rewrote my creative use of Algorithm::Diff to find all the true changes to the configuration file.
gac410 wants forensics
[15:54]
CDotsorry about that; you were using a rocket-propelled grenade launcher to hunt mosquitoes.... [15:54]
gac410y, but My grenade launcher really did find EVERY change. Your code only reports what the UI believes it changed. [15:55]
CDotindeed. [15:56]
gac410There are subtle changes that are not reported. For ex when Load re-maps values. [15:56]
CDotif the two are not the same, the answer is to find out why and fix that
I understand why you did what you did, but IMHO it was total overkill
and I was trying to cut some of the code bloat
[15:56]
gac410Right, But old configs are not sorted. That is easier said than done. I've spent hours. That's exactly why I added Algorithm::diff. It was easier to write it than continue a line-by-line unsorted comparison.
I was debugging a failed migration from 1.1 to 1.2 when I added that compare.
[15:57]
CDotCDot thinks we worry too much about old configs [15:58]
gac410Y, It's so easy to say "Re-install" ... Core is not all that difficult. But once all the extensions are added in, LDAP passwords, DB passwords, arcane extensions, file paths, etc. That is much easier said that ndone. [15:59]
CDotsure, i know. But again, the tradeoff is between supporting every possible angle versus code bloat
the code I replaced it with is much simpler, and does an unsorted compare (IIRC)
[15:59]
gac410Reconfigure from scratch can be extremely painful. [16:00]
CDotI *think* it does everything the old code did, *except* comments
oh, and there's a known problem with embedding code in LocalSite.cfg
not something we should support, IMHO
[16:00]
gac410Hm. Is that the PerlPlugin or something like that? [16:01]
CDotnah
it was for something in the core.... I forget what
[16:01]
gac410The Base64 password encoding frufur
er... foofoo
That seems to work when loading an old 1.2 config, but is not preserved on save, which is fine for me. That was a rather silly change $Foswiki::cfg{Password} = MIME::Base64::decode_base64('....');
[16:01]
CDotthat was it [16:05]
gac410It just meant someone had to do 5 minutes work to break the password.
That code was never "in the wild" except for the lunatic fringe running 1.2 in production.
And since the old LSC includes a "require MIME::Base64" it actually successfully loads when migrating, and the passwords are then saved without encoding. So no issues there.
[16:05]
CDotplease find a better way to do that, that doesn't involve active code in LSC. If we let the thin end of that wedge in, it will become more and more complex.x...x.x.x..x [16:10]
gac410I 100% agree with you. I don't intend to add that back. It was really dumb IMO.
It accomplished nothing other than giving someone knowledgeable a false sense of security.
er... Un-knownledgable
We need to take a pass of http://foswiki.org/Tasks/Configure I think that a lot of these are addressed.
[16:14]
CDotI did that yesterday [16:16]
gac410:) [16:16]
CDotlooked OK to me, but it was only a quick pass [16:16]
gac410Hm. Someone shut down trac.foswiki.org
You don't have permission to access /changeset/13737 on this server.
I thought I left trac running. I wonder if Babar disable it in his recent updates to Apache 2.4
[16:21]
CDotCDot smells elephant [16:22]
gac410I have a way to resolve. Install FoswikirefsPlugin and change the Foswikirev: link to a macro to resolve to github. ... I think.
Bah... no that won't work any more. InterwikiPlugin returns html. Because a svn ref will expand into *multiple* links, ... it's no longer 1:1 I can't be creative in the wikilink definition.
Which means a handler to capture / rewrite Foswikiref:nnnn links before InterwikiPlugin gets to of it.
:(
[16:22]
gac410 wonders ... InterwikiPlugin supports a custom FORMAT setting that defines the returned HTML. If we extended the InterWikis definition with an optional 4th column: [16:38]
| External site alias | URL | Tooltip help text | Format | If it was present, would override the default INTERLINKFORMAT from '<a class="interwikiLink" href="$url" title="$tooltip"><noautolink>$label</noautolink></a>'
| Foswikirev | %REV2REF{$page} | revision $page of 'Foswiki' | $url |
So Foswikiref:12345 would expand to %REV2REF{12345}% which in turn expands to multiple URLs
[16:43]
........ (idle for 39mn)
Okay... One more enhancement to InterwikiPlugin. Extend format for optional 4th column.
Works fine, and allows us to change Foswikirev:123 to %REV2REF{123}%
-| Foswikirev | http://trac.foswiki.org/changeset/ | revision $page of 'Foswiki' |
+| Foswikirev | %REV2REF{$page}% | revision $page of 'Foswiki' on Github | $url |
I guess I need a feature proposal :(
Hm... maybe I'll just piggyback on http://foswiki.org/Tasks/Item12819
[17:24]
..... (idle for 22mn)
Okay, to at least maintain some pretense of following our development process :D
http://foswiki.org/Development/InterwikiPluginMacroExpansionCustomURLs
It's a feature request 50% in arrears for the two features I need to get Foswikirev: links working for the github migration.
I'm going to create a task and check this in, and install it on Foswiki.org so that we can live without trac and not have broken links.
And deal with the fallout if anyone raises objections to the enhancement. I think it's fairly innocuous
[17:49]
.... (idle for 17mn)
GithubBot[distro] cdot pushed 1 new commit to master: http://git.io/QklU7g
distro/master 9829d74 crawford: Item12952: add a depreport stage to the install process to support checking dependencies. Fix dependenices in EditRowPlugin and add a test to BuildContrib for correct construction
[18:08]
***GithubBot has left [18:08]
CDotgac410: OK, you nw have a dependency report before the install (two-step process) [18:08]
gac410Ah cool. [18:09]
CDotCDot senses a need for more explanatory doc in the dialog, but hey..... [18:09]
gac410I'll take a look ... once I move beyond a #($@ regex that *should* be simple [18:10]
...................... (idle for 1h48mn)
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/pXiggQ
distro/master 42f9b94 George Clark: Item13045: Override INTERLINKFORMAT per interlink...
[19:58]
***GithubBot has left [19:58]
gac410Damn... The github hook is broken. I think the Apache 2.4 upgrade has cratered REST on trunk.
Yup. trunk.foswiki.org is completely down. 403 for everything :(
Babar ... BABAR .... what did you do ?
[20:08]
Okay ... trunk.foswiki.org is back up. [20:19]
................ (idle for 1h15mn)
***Slion has left [21:34]
.................. (idle for 1h28mn)
ChanServ sets mode: +o Babar
ChanServ sets mode: +o Babar
[23:02]

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