| 01:17.28 | *** join/#storm oubiwann (n=oubiwann@c-71-196-236-126.hsd1.co.comcast.net) |
| 01:31.37 | *** join/#storm fcorrea (n=fcorrea@189-54-145-62-nd.cpe.vivax.com.br) |
| 02:38.56 | *** join/#storm stub (n=stub@ppp-58-8-3-219.revip2.asianet.co.th) |
| 05:17.31 | *** join/#storm stub (n=stub@ppp-58-8-3-219.revip2.asianet.co.th) |
| 06:40.12 | *** join/#storm elmom_ (n=elmom@hoasnet-fe29dd00-137.dhcp.inet.fi) |
| 07:50.43 | *** join/#storm fcorrea (n=fcorrea@189-54-145-62-nd.cpe.vivax.com.br) |
| 08:12.39 | *** join/#storm jukart (i=lovely@81.189.156.94) |
| 09:34.34 | *** join/#storm Mr_Shark (n=jens@90-229-228-173-no153.tbcn.telia.com) |
| 11:52.22 | *** join/#storm jamesh_ (n=james@canonical/launchpad/jamesh) |
| 12:09.28 | *** join/#storm fcorrea (n=fcorrea@187.3.137.148) |
| 13:00.53 | *** join/#storm fcorrea (n=fcorrea@187.3.137.148) |
| 13:50.42 | *** join/#storm keppla1 (n=keppla@i577B331B.versanet.de) |
| 16:18.48 | *** join/#storm jtv (n=jtv@ppp-124-122-79-208.revip2.asianet.co.th) |
| 16:20.24 | jtv | Is this true? If so, is it intentional? ISTM if a connection fails during commit, it needs to be rolled back before it can become usable again. |
| 16:22.23 | jtv | jkakar, hny btw! |
| 16:24.15 | therve | jtv: like which kind of failure? |
| 16:24.34 | jtv | therve: hny to you as well. :-) A connection failure, so e.g. a killed backend. |
| 16:25.03 | jtv | As I read Connection.commit, it puts the connection in STATE_DISCONNECTED. |
| 16:25.03 | therve | happy new year :) |
| 16:25.07 | jtv | :-) |
| 16:25.14 | jtv | Only a rollback will get it back to STATE_RECONNECT. |
| 16:25.38 | jtv | So you can't do "if ok: txn.commit() ; else: txn.rollback()" in a web app like LP |
| 16:25.39 | therve | jtv: that's probably a good idea anyway |
| 16:26.41 | therve | in web app like LP, transaction.abort should be called after an exception |
| 16:27.23 | jtv | sure, but commit() is a very special time at the end of the year^H^H^H^Hrequest, a bit different from all the rest |
| 16:27.44 | jtv | It happens during a callback that's outside the normal exception-handling scope of the request body. |
| 16:28.04 | jtv | doesn't actually like christmas all that much, for the record |
| 16:30.47 | jtv | therve: also, there's normally a kind of symmetry between commit and abort. You do either the one, or the other. |
| 16:31.08 | jtv | If a commit fails, there's just no other place to go. |
| 16:31.23 | jtv | So might as well have a failed commit act as an abort anyway. |
| 16:33.23 | jtv | It seems very obvious to perform the transaction in a "try" block ending in "commit," and then having an "except" clause that aborts. But then you get another kind of subtle bug creeping in: |
| 16:34.09 | jtv | the "except" clause may assume that after the rollback, the transaction was rolled back. But it's also possible that the backend completed the commit but was unable to tell youâso you do a rollback but your transaction was committed anyway. |
| 16:34.57 | jtv | So lumping the commit in with the rest of the "try" block isn't strictly safe. |
| 16:36.02 | jtv | OTOH you place the commit outside the "try" block, and it breaks, you get a horrible error which is exactly appropriate in that horror-of-horrors case. |
| 16:37.09 | *** join/#storm danilo_ (n=danilo@static-213-198-236-193.adsl.eunet.rs) |
| 16:37.22 | jtv | therve: am I making sense? You're very quiet. :-) |
| 16:38.17 | therve | jtv: sorry meeting started |
| 17:39.42 | *** join/#storm fcorrea_ (n=fcorrea@187.3.137.148) |
| 17:56.35 | *** join/#storm jtv1 (n=jtv@125.25.115.248.adsl.dynamic.totbb.net) |
| 18:03.18 | therve | jtv: so, I looked at Landscape code |
| 18:03.36 | therve | jtv: and we're actually rollback'ing every stores when the request starts |
| 18:03.46 | therve | probably to prevent problems like that |
| 18:04.07 | jtv | therve: that makes sense now that I know about this... |
| 18:16.22 | jtv | therve: I have to go off now, but our bug is here: https://bugs.edge.launchpad.net/launchpad-foundations/+bug/504291 |
| 18:16.36 | mup | Bug #504291: DisconnectionErrors (already disconnected) happening again <oops> <Launchpad Foundations:New> <https://launchpad.net/bugs/504291> |
| 20:43.44 | *** part/#storm brianmac1 (n=Brian@75-151-32-30-Naples.hfc.comcastbusiness.net) |
| 23:37.48 | *** join/#storm fcorrea (n=fcorrea@189-54-145-62-nd.cpe.vivax.com.br) |
| 23:45.46 | *** join/#storm danilo_ (n=danilo@static-213-198-236-193.adsl.eunet.rs) |