#foswiki 2015-07-16,Thu

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

WhoWhatWhen
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/vmRqW
distro/master 539b03d George Clark: Item13526: Bump TipsContrib version
[01:03]
***GithubBot has left [01:03]
FoswikiBothttp://foswiki.org/Tasks/Item13526 [ Item13526: TipsContrib needs optimizations ] [01:03]
GithubBot[distro] FoswikiBot pushed 2 new commits to master: http://git.io/vmRqV
distro/master 83b4ce6 CH yang: Item13505: Translations updated using Weblate (Traditional Chinese)...
distro/master 699b3ad CH yang: Item13505: Translations updated using Weblate (Traditional Chinese)...
[01:05]
***GithubBot has left [01:05]
FoswikiBothttp://foswiki.org/Tasks/Item13505 [ Item13505: Translations for Foswiki 2.0.1 / 2.1.0 ] [01:05]
................. (idle for 1h22mn)
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/vmRuK
distro/master 7e1f547 George Clark: Item13510: Package installer needs $FALSE $TRUE defined...
[02:27]
***GithubBot has left [02:27]
FoswikiBothttp://foswiki.org/Tasks/Item13510 [ Item13510: Configure is incorrectly applying =$FALSE= and =$TRUE= defaults from Config.spec files ] [02:27]
GithubBot[distro] gac410 pushed 1 new commit to master: http://git.io/vmRav
distro/master 36aca0d George Clark: Item13513: Don't overlay existing config with extension install...
[02:40]
***GithubBot has left [02:40]
FoswikiBothttp://foswiki.org/Tasks/Item13513 [ Item13513: Configure extension installer overlays extension settings with defaults when upgrading an extension. ] [02:40]
gac410phew. Fixed the two configure issues related to installing extensions.
So if we can get to a resolution on bulk_copy, we could consider a 2.0.1 release.
[02:46]
.......................................... (idle for 3h29mn)
***ChanServ sets mode: +o CDot
ChanServ sets mode: +o MichaelDaum
[06:17]
............................................................................... (idle for 6h33mn)
ChanServ sets mode: +o gac410 [12:51]
gac410Hiya CDot - good morning. I fixed the two configure ext. installer bugs last night. Easy once I figured out where to look. [12:54]
***ChanServ sets mode: +o Lynnwood [13:08]
........ (idle for 39mn)
CDotgac410: well done! I have had a couple of looks at bulk_copy, but it's quite complex to solve correctly.
Your hack solution is a good starting point, though not a full fix.
[13:47]
................................... (idle for 2h54mn)
gac410CDot_: Just remember, yesterday, JulianLevens said he is also digging into bulk_copy. I'd still like to understand what's wrong with the old convert_store. For at least my tests, it seemed to work well. [16:42]
CDot_gac410: nothing - except it only work on RCS stores.
bulk_copy is needed for moving between store implementations.
[16:44]
gac410Yes, very restricted. bulk_copy is much more thorough. But it works ;) [16:44]
CDot_so does bulk_copy, if you haven't abused the attachment histories [16:45]
gac410hm. I'm pretty sure my attachments had valid histories. I'll have to look [16:45]
CDot_then with your patch, it should work
your complaint, though, was that "hidden" revisions of the attachments get "lost"
[16:46]
gac410I'm just concerned that hidden revisions are normal ... if an attach operation falls into the "force-new-revision" event horizon [16:47]
CDot_note that attaching a new rev of an attachment should cycle up the topic as well - unless it is edited-within, in which case the rev no in the topic needs to upcycle even though the topic rev number stays the same
(or, of fcourse, the attachment rev no can be frozen in that case)
[16:47]
gac410Right. Just looking back in history ... store has been very inconsistent at times. [16:48]
CDot_if this is *not* the case, then Web/Topic/attachment?version=X may well be (currently) giving you Y (or F, or U, or C....) [16:49]
gac410Anyway. My test topic, has a "big attachment" a big zipfile. How do I validate that the history is in sync. Topic has 4 revisions, Attachment has 3. [16:49]
CDot_there are several cases to consider. First, topic cycles without attachment change. That's pretty well covered.
Second, attachment cycles after long delay. In that case, topic should cycle when attachment does (I think)
third, attachment cycles after a short delay. In that case, either topic stays solid and attrev upcycles, or attrev stays solid and data only changes.
[16:50]
gac410My TestBigAttach.txt,v topic ... grep for the attachment shows only one META in the history. [16:51]
CDot_Fourth, att cycles outside of Foswiki [16:51]
gac410%META:FILEATTACHMENT{name="CADEMIA_3.0-RC2.zip" attachment="CADEMIA_3.0-RC2.zip" attr="" comment="" date="1339776810" path="CADEMIA_3.0-RC2.zip" size="15191747" user="BaseUserMapping_333" version="3"}%
So I don't see in history a Rev 1 or 2.
[16:51]
CDot_in that case, topic should auto-update (if necessary) on next edit/attachment update
Fifth, if some goit deletes attachment.dat,v then history should (I think) reset to 1
[16:51]
gac410The attachment itself is at R3 as well. ... revision 1.3
date: 2012/06/15 16:13:33;  author: BaseUserMapping_333;  state: Exp;  lines: +3366 -2448
[16:52]
CDot_note that I'm describing what *should* be done, not what *is* done
CDot_ strongly suspects that the core code isn't consistent in it's treatment
[16:52]
gac410Interesting thing. When I run my patched bulk_copy, I get NO attachments. [16:52]
CDotis there a META:FILEATTACHMENT for the attachment? [16:53]
gac410It reports the message, and even claims to convert it, but the pub directory is gone. [16:53]
CDotand a .dat? [16:53]
gac410.dat??? [16:53]
CDot.whatever the attachment is e.g. .jpg
.fnord
[16:53]
gac410yes. The topic has 3 META:FILEATTACHMENTS. including the TestBigAttach.
interesting. the 3rd attachment ".lesshist" a "hidden file" is actually not attached, but listed in the topic meta. This is sure a messed up topic.
I wonder if it hits that file and then fails to save the others somehow.
maybe I should not be testing with this one. It's all fouled up. the 2nd attachment has 4 revs, but the cached copy is empty.
These are just crap webs that I've accumulated over the years. good ugly data for trying things on.
But I suspect that sites that migrated back from twiki over the years have stuff like this crap lurking here and there.
Okay. according to the "api" under 1.1.9, the file I'm having trouble with is in "pending" state. So 4 rev's are in the rlog, but a 5th rev is pending.
Anyway. Let's "table" all this, and we can pick it up after you or Julian think you've got a solid version of bulk_copy.
[16:54]
CDotThat means the file date on the .txt (equivalent) is ahead of the ,v
I don't think the script is far off correct. The problems more lie with the underlying store, which *may* make it impossible to copy it cleanly.
The one problem which I know has to be solved is where there are "unrecorded" revisions of the topic between revs in the META:FILEATTACHMENT
for example, topic history says revs 1, 1, 2, 6, 9
which is fine as long as 1..9 all exist in the ,v. If they don't, you are going to end up with 9 copies of the same data :-(
The code i wrote orignially was a (possibly misguided) attempt to clean that up by only copying revs that were actually mentioned in the topic history
[17:04]
gac410okay. hm so part of my problem is a "bad copy" between foswiki versions. The data is much cleaner in the source webs, which I keep in a "LocalDataContrib" private git repo.
I'll rebuild my 1.1.9 testcases, before I test again.
[17:08]
CDotIt's all made more complex by the fact that the very nature of the RCS store means that you have no guarantee that what *claims* to be rev 5 is the rev 5 that the user originally attached.
It could be some other rev 5 that some other bright spark decided to plunge in
Lasciate ogne speranza, voi ch'intrate
[17:11]
gac410Well I had to hit up google translate, but y. good sentiment [17:27]
jmk0What would be the implications as far as public release and compatibility be if I chose to implement my plugin using the perl XS interface?
(i.e. perl interface to c/c++)
[17:32]
gac410Hm. I don't think we have any easy way to upload architecture specific plugins. So you'd need to document the build procedure. [17:34]
jmk0apparently there are already some such things. NativeSearchContrib for one [17:34]
gac410Precedence is the NativeSearch.... yeah exactly what I was going to say
It's the only one I know of off-hand. But configure / package installer would not have any easy way to compile.
[17:34]
jmk0shouldn't be architecture-specific. Hopefully portable C++ code but i'll do a bit more poking to see what NSC had to do. I was thinking of something that could be built using cpan or something like that [17:35]
gac410Hm. maybe. There is a "post-install" exit that can process after an install.
The package installer has 4 exits . PREINSTALL POSTINSTALL PREUNINSTALL and POSTUNINSTALL
[17:35]
jmk0there are instructions for building NSC but I have to wonder if it'd be possible/easier to upload the XS interface to CPAN then just use that to build the XS module and have a separate plugin to install that just requires that perl module [17:36]
gac410So you could possible have preinstall test for available complier and die if unavailable. And postinstall do the compile. If you really want to be thorough
Yeah. Release the module to cpan would work too.
[17:36]
jmk0I haven't put together a package installer yet :) [17:37]
gac410cd YourPlugin/lib/Foswiki/Plugins/YourPlugin
./build.pl release
[17:37]
jmk0thus far i've just been rsyncing or extracting tarballs
oh, that's pretty straightforward
i need to start writing these things down :D
[17:38]
gac410build.pl also has "manifest" target that will build a manifest. ./build.pl manifest > MANIFEST Then edit the file to clean it up.
Foswiki:Extensions/BuildContrib has the docs. Nothing to write down I hope
[17:38]
FoswikiBothttp://foswiki.org/Extensions/BuildContrib [ BuildContrib ] [17:39]
jmk0ah, thanks [17:45]
jomojmk0: if possible consider the Inline:C(PP) also. Because I have some bad experience with BADLY written perl/xs (using OSX). Here are many great XS what builds cleanly, but also many what is terrible. (one example is Image::Magick). But, because the Inline:C is well-tested, i never meet any problems with compiled modules what uses Inline:C. Just my 2 cents... [17:45]
jmk0using foswiki to document the design, build and release plan for my foswiki plugins :-)
yeah imagemagick I could never get to work under windows :-/
[17:45]
gac410I think I got it or graphcmagick to work once ... but I don't use windows except for an occasional release test and my VM is so obsolete, not worth using any more. [17:47]
jomotherefore i recommending Inline::C what hides the implementation/compilation details ... [17:48]
........................ (idle for 1h57mn)
gac410jmk0: Didn't see any split's mentioned here ...
(02:15:33 PM) jmk0 left the room (quit: Ping timeout: 246 seconds). and (03:08:51 PM) foswiki_irc3 ... entered the room
[19:45]
.... (idle for 15mn)
jmk0... down side of using cpan is having two places from which to get updates to the extension [20:01]
gac410yes. off & on there has been ideas offered of distributing all of foswiki as a cpan module. But (thankfully?) no takers.
Anyone heard of sourceforge discontinuing services?
[20:02]
jmk0google code definitely has discontinued. I have no firm memories of sourceforge. It sounds vaguely familiar but only vaguely [20:03]
jomojmk0 Inline::C solves this problem to. The user need install from CPAN only Inline::C and the Foswiki module will compile automatically at 1st use... after every Foswiki Extension update... [20:04]
gac410could not get connected to the sourceforge.net list server today. and that's where all the foswiki-* lists are hosted. [20:04]
jmk0jomo, I'm not sure how useful an inline strategy would be - this is no small amount of code I'm working with [20:05]
jomoI don't really understand how the amount of code affect this - but ok - i just recommending... decision is yours.. :) [20:12]
gac410Reading all the very upset blogs about sourceforge now bundling some malware / crapware in downloads. ... maybe we ought to look for alternate hosting for our email lists and code. [20:15]
jmk0ew
I think I ran into that myself some months ago
[20:23]
gac410Seems to be filezilla that has encountered the most issues. [20:23]
jmk0sourceforge is affiliated with slashdot, isn't it? [20:26]
gac410Used to be. Not sure if it still is. freshmeat, slashdot and sourceforge used to be all from the same organization. freshmeat became freecode became shutdown [20:26]
jmk0I had some stuff for my blog on google code but since I haven't been keeping the blog up to date, I haven't bothered to migrate the code to github either... [20:27]
.... (idle for 18mn)
jomogac410: when get a free minute, tell me please :) [20:45]
gac410tell you what? [20:46]
jomoa fee minute
free
need help with something
[20:46]
gac410Okay go for it.
what's up?
[20:46]
jomowant include this topic http://trunk.foswiki.org/Sandbox/IncludedTopic into this http://trunk.foswiki.org/Sandbox/IncludingTopic topic... what is the correct solution? [20:47]
gac410Ah... SETs are *never* applied from an include. they are strictly set in the base topic [20:48]
jomoas always - i could solve complicated things with foswiki, but have problems with the simple things... ;( [20:48]
gac410you would need to set the %MACRO in the including topic
hm I wonder about %SET macro new with foswiki .. .that might be different
[20:48]
jomobut it wont work in the IncludedTopic - e.g.. need duplicate
tried also - and get nowhere
and the MACRO has different content for the IncludedTopic1 and IncludedTopic2
[20:49]
gac410oh. hm... Not macros, but the other way to set persistent variables at runtime is using %CALC{$SET(...)}%
but if you SET the same variable in multiple places, I have no idea what will happen.
[20:51]
jomook, just wondered how such simple thing is solvable with FW... - will change the logic - thanx for your time ;) [20:54]
gac410np. maybe others might have brilliant ideas - I'm fresh out. [20:55]
jomo:D [20:56]
............. (idle for 1h2mn)
gac410SourceForge is all messed up. Just noticed they've lost Foswiki 2.0, "latest version" has reset to 1.1.9 [21:58]
GuilainCgac410... about sourceforge, take a look about the gimp story...
:(
GuilainC loves sourceforge, before...
[21:59]
gac410Yeah. gimp / filezilla ... both picked up crapware, but in their windows installers, which we do not have.
Our downloads are gpg signed as well. ... though we host the key on sourceforge so I suppose anything is possible.
not that anyone ever checks signatures ;)
[22:00]
GuilainCyes... but the approach is to my point of view unfair... and I'm not against any advertising & so on [22:01]
gac410no I agree with you 100% ... embedding unwanted malware / adware is really wrong.
There are not a lot of options. Self host? email lists hosts? and that's a LOT of users to figure out how to resubscribe
[22:01]
GuilainCIMHO, more set up a new association for opensource software
I'm sure that some buisness model is possible, making the support of development (aka sourceforge & co) free for the projects, and supporting by sponsor/company, but don't want to debate about that... it's like that, and you right, in a short point of view, only selfhosting is possible and is painfull it's true
[22:04]
jmk0i check signatures :)
yahoo groups!
no telling how much longer yahoo will be around, though i do genuinely like yahoo groups
probably not the best choice in this context, though :)
[22:06]
....................... (idle for 1h51mn)
***gac410 has left [23:59]

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