(09:01:13 AM) jomo entered the room.
(09:01:21 AM) JulianLevens entered the room.
(09:01:34 AM) gac410: Hi everyone.
(09:01:43 AM) JulianLevens: HI
(09:01:53 AM) jomo: hi :)
... _some discussion redacted_
(09:04:58 AM) gac410: I've got is a change to %CALC to encode certain characters.
(09:05:35 AM) gac410: But this might be disruptive to existing SSP macros. if someone is using them in legitimate html.
(09:06:57 AM) jomo: or only for WikiGuest… e.g. normal users wont see any change…
(09:07:16 AM) jomo: or configurable bejhavior..
(09:07:24 AM) gac410: no... not a good idea.
...
(09:08:25 AM) gac410: Configurable ... yes. I was making it page configurable ... So a site could enable /disable as desired, and finalized it for more control
(09:09:00 AM) jomo: i don’t know and can’t predict the CALC usege scenarios… - must tell others, who routinely uses it…
(09:09:23 AM) gac410: Y. Hoping Lavr is paying attention. Kenneth is our master of arcane ssp usage :D
(09:09:43 AM) jomo: :)
(09:11:21 AM) gac410: jomo, I'll check in my fix, lacking any further feedback here. ... But give you something to test.
(09:11:42 AM) gac410: Okay ... another urgent task. Regular Expressions and performance.
(09:11:56 AM) jomo: ok
(09:12:59 AM) gac410: The symptoms showed up in EditRowPlugin, I did some profiling for a site that was seeing minute-long response time compared to EditTablePlugin.
(09:13:42 AM) gac410: putBackBlocks was accounting for a lot of time. Turns our Regexes on utf-8 text are many many times slower.
(09:13:49 AM) gac410: http://foswiki.org/Tasks/Item13820
(09:14:53 AM) gac410: Converting from a regex scan to a loop using index and substr gained back 50%. We need to be cautious about using regexes for scanning when there are other solutions.
(09:15:29 AM) gac410: The 2nd half of the performance was ERP re-reading the topic 1000's of times. .. Func::getRevisionInfo re-reads the topic.
(09:16:15 AM) gac410: Changing to Meta::getRevisionInfo on a loaded meta object fixed the rest of the issue for my test cases, but not for the user. :(
(09:16:30 AM) gac410: Anyone here good at strace?
(09:16:48 AM) jomo: not me… :(
(09:17:06 AM) gac410: http://foswiki.org/Tasks/Item13817 The fixes worked for me but not for the user.
(09:17:22 AM) gac410: He seems to have a different issue.
(09:17:49 AM) jomo: and also i can’t repro the users issue - i tried edit with ERP 10k lines table - it tooks 2-3 minutes (but 10k lines!)
(09:18:01 AM) gac410: Is that after my fixes?
(09:18:09 AM) jomo: no - before..
(09:18:47 AM) gac410: Ah... I have not checked in the ERP fix. just the putBackBlock fixes. I was hoping CDot would review the fix. I'll check it in
(09:21:09 AM) gac410: One more urgent http://foswiki.org/Tasks/Item13785 For some reason the ExtensionInstaller occasionally fails to update certain files.
(09:22:26 AM) gac410: Anything anyone else wants to bring up regarding tasks?
(09:23:58 AM) jomo: https://rt.perl.org/Public/Bug/Display.html?id=126182
(09:24:07 AM) gac410: As far as 1.1.10, or a Patch contrib for 1.1.x, I've not made any further progress.
(09:24:27 AM) jomo: this (.(?2))((?<=(?=(?1)).)) regex definitely hangs search…
(09:25:40 AM) gac410: Interesting. I wonder if somehow there is some path through EditRowPlugin that hits that condition.
(09:26:06 AM) jomo: probably not - this is an “new” recursive regex...
(09:26:42 AM) gac410: oh :(
(09:27:39 AM) jomo: but eats all memory - the server needs to be configured do wait for inifinite response from the script - otherwise it eats all swap…
(09:27:41 AM) gac410: I wonder if it's some "optimization" or local fixing that SuSE did for their perl. I could not recreate the issue either even with the same perl version with perlbrew
(09:28:51 AM) jomo: ok - it just for the record - probabluy we can’t to do anything with the issue...
(09:29:20 AM) gac410: Okay ... thanks. yeah once someone is down in strace, it leaves me behind. :)
(09:31:56 AM) gac410: Okay ... any thoughts on branching 2.0 out from master so that 2.1 can proceed. Any feature branches needing merge?
(09:32:16 AM) gac410: I think MichaelDaum has one feature branch ready ... to add in the security headers
(09:41:23 AM) gac410: regarding regex performance. A simple trim of trailing spaces. We do it all over. s/\s*$//g huge performance differences
(09:41:23 AM) gac410: while substr, chop. at least twice as fast.
(09:41:57 AM) gac410: Anyway ... I don't have much else to discuss. Anyone?
(09:42:09 AM) jomo: :) not me ;)
(09:42:12 AM) JulianLevens: Nothing from me
(09:44:47 AM) jomo: so.. okay - thank you and CU..
(09:44:59 AM) gac410: Okay. Lets adjourn then ...
(09:45:23 AM) gac410: jomo, I'll check in my fixes for you to review. for ERP and SSP. thank you.
(09:45:40 AM) jomo: ok, doing to pull - later a bit…
(09:46:09 AM) jomo: need gtg - CU
(09:46:13 AM) jomo left the room (quit: Quit: jomo).
(09:47:53 AM) jast: rats, I missed all the fun while in a conference call
(09:48:11 AM) gac410: Ah... still here. Any comments to add?
(09:49:41 AM) jast: can't think of anything, except as noted in the main channel I noticed that PlainFileStore will sometimes not return author/comment for the most recent revision of an attachment (in the getVersionInfo method)
(09:49:57 AM) jast: so there'll probably be a bug report soonish
(09:50:06 AM) gac410: okay thanks
(09:53:05 AM) JulianLevens left the room.
(10:05:15 AM) Lynnwood [~Lynnwood@foswiki/developer/lynnwood] entered the room.