PMC QA Meeting 20100324

From ADempiere
Jump to: navigation, search
This Wiki is read-only for reference purposes to avoid broken links.

Date: 2010-03-24
Time: 8AM GMT
Venue: irc #adempiere-team
Support Spreadsheet: Adempiere PMC QA
Chat times in GMT-5


(03:13:06) CarlosRuiz: ok, so if you want we can start QA meeting - sorry for the delay
(03:13:08) hengsin: hi Ivan
(03:13:22) interopen: Hi Heng Sin
(03:13:23) viola [~viola@p4FE2F058.dip.t-dialin.net] entered the channel.
(03:13:38) interopen: yes we can start, not worry i was reading your interesting talk
(03:13:57) interopen: We can talk today:
(03:14:26) interopen: 1. Sahi
(03:14:27) interopen: 2. Security testing
(03:14:27) interopen: 3. Review
(03:14:57) CarlosRuiz: great
(03:15:23) hengsin: I saw you have make progress on the sahi test, is it working now with adempiere webui ?
(03:15:31) interopen: About Sahi i see we have help from the developer http://www.adempiere.com/index.php/Talk:PMC:QA:Sahi_with_ZK_Webui_POC
(03:15:51) interopen: it works with some issues
(03:16:02) hengsin: great!
(03:16:18) interopen: we are now working on documenting a POC that creates a product
(03:17:24) interopen: Also as we document with selenium we have the problem of integrating an IDgenerator
(03:17:34) interopen: so we are focusing for now more on Sahi
(03:19:13) CarlosRuiz: the id problem in selenium is not present with sahi?
(03:20:46) interopen: no is not, that is why they recomend Sahi for ZK applications
(03:22:57) interopen: the advantage is that you can record test in any server and later run them in the CI one
(03:23:33) interopen: without any code change, and it could be cross-version also
(03:24:12) CarlosRuiz: meanwhile you don't add fields in the middle? or it can cope with that?
(03:25:25) interopen: not sure if can do it, we haven't test it
(03:25:53) hengsin: i think it is path or position base, so adding field in middle could break it
(03:26:54) interopen: yes, in logic it could
(03:27:12) interopen: but not sure if Sahi have address it, as it seems quite common for any application
(03:27:14) hengsin: so fixing the id is better but will take quite a bit of time to do that ...
(03:27:49) hengsin: probably can be done as we develop the test cases ...
(03:28:29) hengsin: note that using IDGenerator is not a good solution either, we need to named the component and able to do a partial match in the test case
(03:29:20) interopen: i think, as we are having the help of Sahi developer ( without asking it), we can move faster on solving those issues
(03:31:06) interopen: Also we have the problem with IDGenerator, that we only can run the tests in the server who have it deployed
(03:33:11) interopen: and will be harder for users to create functional tests
(03:34:04) interopen: my proposal, is now we focus all our efforts with Sahi, as it can be faster to have it running
(03:34:51) interopen: we always can come back to selenium if later we see a bigger issue with Sahi
(03:35:48) CarlosRuiz: it's perfect, I agree
(03:37:38) interopen: ok, then we focus on it, and see if we can have a real POC soon that let other users to test it
(03:38:58) interopen: 2. About security testing, we can add just add some of this tools to the CI testing
(03:39:28) interopen: have you ever try a tool like this on the Adempiere code? http://findbugs.sourceforge.net/demo.html
(03:40:43) CarlosRuiz: I've seen people using other tools - Michael Judd and Teo if I don't remember wrong
(03:40:43) CarlosRuiz: but first time I see about findbugs
(03:42:13) interopen: also for me, i have not test it yet
(03:42:21) interopen: i come out with it base on your email
(03:43:06) CarlosRuiz: sounds good to pass on our code
(03:44:26) interopen: yes, i see there are lots of this tools we can add to test before and after CI
(03:45:26) interopen: we go researching and adding documentation about them to the wiki
(03:45:41) hengsin: the biggest outstanding security issue is the way swing get db password from server - will be good if someone can work on that ...
(03:47:21) CarlosRuiz: hengsin: I think that's a flaw from client/server architecture (really really hard to solve, you just can workaround it)
(03:48:52) interopen: in our case we use swing client only for system administration, and we restrict postgresql of the client who can use it
(03:51:05) trifon [~trifon@pD95F0CC3.dip0.t-ipconnect.de] entered the channel.
(03:51:26) CarlosRuiz: that's why I say (but people don't like to hear this) - that zkwebui will deprecate swing - because I think those security issues on swing are very hard to solve, and they're very serious.
(03:51:26) CarlosRuiz: Well, at least they're easy to workaround, but most people don't set it up
(03:52:13) hengsin: we only use it for development, client is on web ui
(03:52:34) hengsin: much better for development though :)
(03:52:50) interopen: agree carlos, the swing client access directly the database, and all in the user hand... is risky
(03:53:07) CarlosRuiz: Ivan - about 2. Security testing
(03:53:07) CarlosRuiz: I appreciate your help, I guess we would need to conduct also a research about the lot of tools out there
(03:53:40) interopen: yes, that is what we are doing, but as tell you in the email step by step
(03:54:04) CarlosRuiz: yep, we're not in a hurry - and I like the way you're doing
(03:54:07) interopen: first is CI server working with some tests
(03:54:19) interopen: as security testing, i just see it as add more tests
(03:54:48) interopen: most of the security testing tools, just load a bunch of tests predefined
(03:55:17) interopen: not as functional or code test that need to be created base on each program
(03:55:56) CarlosRuiz: yep - that's good - we can add several - and rerun periodically
(03:56:33) CarlosRuiz: what is 3. Review  ?
(03:56:40) interopen: yes, that is the idea adding them to the CI server some before build, some after test, some every month...
(03:57:11) interopen: 3. is a small thing, i create before this mindmap http://www.adempiere.com/index.php/PMC:QA:CI_base_infraestructure
(03:57:48) interopen: as we research i am going to complete it and add sub-mindmaps
(03:58:36) CarlosRuiz: looks great
(03:58:46) interopen: Reading your discussions in Architecture group, where you talk about a lot of concepts and ideas, much more complex than QA, this kind of resumes can help you keep focus
(03:59:20) CarlosRuiz: good tip
(03:59:23) interopen: and help others understand it easier
(03:59:48) interopen: welcome :-)
(04:00:04) CarlosRuiz: I understand load testing as having lots of concurrent users
(04:00:04) CarlosRuiz: and different is performance testings that can be lots of records
(04:00:35) CarlosRuiz: just adding ideas to the mindmap  ;-)
(04:01:04) interopen: thanks, yes but both are loads one in users and other in records
(04:01:18) trifon: Carlos i think both are equal.
(04:01:22) interopen: i can differentiate them
(04:01:28) trifon: load testing and perfomrance testing.
(04:01:36) CarlosRuiz: indeed it must be one mixed - lots of users with lots of data
(04:01:40) trifon: you can make load testing without creating records.
(04:01:58) viola has left the channel (quit: Quit: viola).
(04:02:23) CarlosRuiz: I've seen servers becoming unusable because several users launch financial reports at the same time with a big fact_acct table
(04:02:50) hengsin: carlos, I've that in the architecture wish list
(04:02:51) interopen: yes, both at the same time, is the ideal
(04:03:16) interopen: but also we can get some data from running them separately to fix problems
(04:03:16) hengsin: i.e, to have the ability for user to schedule heavy report instead of run it immediately
(04:04:07) CarlosRuiz: I recently fixed the scheduler to work in clients - tested with aging report and it worked pretty well
(04:04:09) interopen: Hengsin, yes that is a usual functionality in other ERPs
(04:04:10) hengsin: and we can goes further to say how many of those jobs can run concurrently and during what time ...
(04:04:39) CarlosRuiz: and BTW tested your cron4j contribution and it worked very well also
(04:04:43) hengsin: carlos, yes, it can work but to make it usable, we have to integrate that with the current process/report dialog
(04:05:18) CarlosRuiz: good idea
(04:05:33) hengsin: so when user click the start button, the system will tell him/her that the job have been created, will run at what time and notify by note/email
(04:06:04) CarlosRuiz: a process queue - great
(04:06:28) hengsin: actually not a lot of work since most of the code is already in the current scheduler code and it is easy to schedule a one of process using cron job pattern
(04:07:07) CarlosRuiz: interopen: next week is holidays here - can we skip next meeting on 31? so we meet again on april 7?
(04:07:38) interopen: is ok, for me no problem
(04:07:52) interopen: we have a lot of work to finish....
(04:08:10) interopen: so we move it to 7
(04:08:39) CarlosRuiz: ok - hengsin - can we also skip next architectural?
(04:09:13) interopen: totally forget it was eastern
(04:09:16) hengsin: ok. I'm guilty of not having much time for architecture so far, still very busy with the zk work
(04:11:10) CarlosRuiz: me too - totally coped with triage+trunk_integration+average_costing
(04:11:35) interopen: ok, so we move it to 7th, thanks all for the help
(04:11:43) CarlosRuiz: thanks Ivan - c u then
(04:12:24) hengsin: bye all
(04:12:29) hengsin has left the channel ("Konversation terminated!").