Camelia, the Perl 6 bug

IRC log for #darcs, 2013-06-03

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:05 mizu_no_oto joined #darcs
01:36 intripoon_ joined #darcs
01:57 carter joined #darcs
02:22 edwardk joined #darcs
03:38 preflex_ joined #darcs
04:30 edwardk joined #darcs
05:28 carter joined #darcs
05:39 edwardk joined #darcs
07:17 kmels joined #darcs
07:19 lelit joined #darcs
07:38 raichoo joined #darcs
07:47 donri joined #darcs
09:28 alexei___ joined #darcs
09:40 owst joined #darcs
10:06 kmels joined #darcs
10:08 jyyou_ joined #darcs
11:11 * owst idly wonders if we can put ConstraintKinds to use to get rid of Patchy/RepoPatch and their instances
11:11 owst I suppose it requires us to have GHC 7.4 as a minimum version
11:25 favonia joined #darcs
11:29 Heffalump there's loads of stuff I want to do once we can move to GHC 7.4
11:30 owst What GHC are we depending on at the moment?
11:30 Heffalump 7.0
11:30 owst Righto
11:30 Heffalump once 2.10 is out I think we can definitely move. Perhaps we can mvoe anyway now.
11:30 owst What else are you thinking once we have 7.4?
11:31 Heffalump superclass equalities should allow us to get rid of the ApplyState p ~ Tree all over the place
11:31 Heffalump because we can put that constraint into RepoPatc
11:31 Heffalump I can't actually remember what else, but there was some :-)
11:31 owst Ah yeah, that'd be nice
11:31 Heffalump what were you thinking of with ConstraintKinds?
11:32 owst Oh just that we can get rid of the trivial instance definitions, we can use a type synonym for Patchy/RepoPatch
11:32 Heffalump Also, I'd like to anntoate the type fo patches with whether they are inverted or not - I tried before but ended upw ith loads of p ~ Inverse (Inverse p) constraints
11:32 Heffalump oh, I see.
11:32 owst Nothing interesting :-)
11:33 owst Reminds me of my fun with 1+1 = 2 on the type level :-)
11:37 whaletechno joined #darcs
11:52 mizu_no_oto joined #darcs
12:05 mizu_no_oto joined #darcs
13:21 mizu_no_oto joined #darcs
13:29 konundra joined #darcs
13:36 uniquenick joined #darcs
13:46 rarasd joined #darcs
14:11 edwardk joined #darcs
14:27 gh_ joined #darcs
14:41 bsrk joined #darcs
14:59 bsrk_ joined #darcs
15:31 dolio joined #darcs
15:56 MasseR joined #darcs
16:18 konundra joined #darcs
16:22 bsrk joined #darcs
16:40 raichoo joined #darcs
16:41 dolio joined #darcs
16:51 * Heffalump appears
16:57 bsrk Hi
16:57 Heffalump imminent tunnel, so back in a bit :-) I sent you an email about a bit of your code.
16:58 Heffalump in general I was wondering about the how you plan to add OpenID - how pluggable would different authentication mechanisms be?
16:58 alexei___ joined #darcs
17:06 bsrk Heffalump: I am now doubting the worth of supporting OpenID. I think the most painful part of the registration is public keys, and that has to be done manually. The user does get nicer registration and login, but this is only a part of what we want.
17:24 * Heffalump reappears
17:25 Heffalump bsrk: hmm, personally I find not having to remember more passwords useful, and I'd be happy if github supported openid
17:25 Heffalump but I don't know about the population as a whole
17:26 Heffalump I guess also I don't like the idea of subordinating ourselves completely to github :-)
17:30 bsrk Heffalump: If not having to remember password is more useful, then it does make sense to support providers like google
17:30 bsrk (google will be used way more frequently than github)
17:30 Heffalump I keep my github password in a local (encrypted) password store on my laptop, so when my browser looses my cookie i have to look it up again there.
17:30 Heffalump whereas I know my google password off by heart and anyway I only log into that once for google, stackoverflow, etc.
17:32 bsrk Okay, then I will work on OpenID. I actually went ahead and started working on forgot password. :-)
17:33 Heffalump doing things ni whatever order is convenient is fine - and you don't have to take my championing of OpenID as decisive
17:33 Heffalump I'm sure there's plenty of other stuff not in the plan you can find to do too (might be worth thinking about that as you go so you don't run out of things to do anyway)
17:33 konundra joined #darcs
17:33 Heffalump do you have a plan for submitting patches to sm?
17:35 bsrk I am not sure actually. I was assuming integration will happen at the end of gsoc. :-)
17:36 bsrk That said, I do think github support is functional.
17:36 Heffalump definitely don't leave things to the end, that's a recipe for pain
17:38 bsrk There are a few cleanups I want to do(You pointed out one), and then I will ask. :-)
17:38 edwardk joined #darcs
17:38 Heffalump in general the earlier you submit individual finished bits of work the better, though the definition of "finished" may be an entire thing like "github support" or a smaller thing like "support registration with github" depending on sm and your preferences.
17:39 * Heffalump pulls the latest changes
17:40 Heffalump my immediate reaction to "allow blank passwords for github registrations" was "that's pretty dangerous", though I see from the code that it actually doesn't allow any password at all in that case.
17:40 bsrk It does allow passwords
17:41 Heffalump I mean that "allowing a blank password" is a bad idea because then anyone can hack your accuont
17:41 Heffalump but it's not because if the user chooses a blank password then subsequent logins won't work with any password at all
17:41 bsrk You can't login with a blank password
17:42 Heffalump but this topic as a whole does raise the issue of security testing you mentioned a while ago
17:42 Heffalump so I wonder what tests you could write to make sure of this kind of thing
17:45 Heffalump I just tried to do a github sync on your dev server and go "This github account is already linked to another account!"
17:45 bsrk It is not a "test" per se, but I will look at where the sUser field of Session is getting set, and make sure it only works in the intended way.
17:46 bsrk Ah, 1 github account can be linked to only atmost 1 darcsden account
17:46 Heffalump tests are important for makign sure things don't break after further changes
17:47 bsrk Heffalump: yes, which is why it is not a test. :-)
17:47 Heffalump but I didn't select a github account, I just used "sync" on my existing  darcsden account
17:47 Heffalump :-)
17:49 bsrk Ah, my first bug. It used to work before, but I made a change and it's not working anymore. :-)
17:49 Heffalump :-)
17:51 Heffalump btw, the other point that flows from my reaction to the "allow blank passwords" patch is that it helps to write comments that explain why things are safe/what invariants there are in your code and sometimes in yuor checkin comments.
17:52 mizu_no_oto joined #darcs
17:52 Heffalump "allow the user not to set a password" would have scared me less, and the code would benefit from a comment that if uPaS is Nothing, the user can't login by password
17:56 bsrk I can see how commenting on the patch is helpful. But it it really a good idea to comment on the code? It would be advantageous to the reviewer, but won't it be unhelpful for the future user?
17:56 bsrk I mean future reader/contributor.
17:57 Heffalump the code is tied together by invariants - in this case, the fact that uPaS is Nothing ==> no login by password
17:58 Heffalump for a reader of the code itself, discovering that fact requires searching for the login code  - but a comment on the definition of uPaS would save that time
17:58 Heffalump and it also makes it less likely that a future contributor will break things
17:58 Heffalump in general data structures are the most important thing to comment - explain what particular values *mean*
17:59 bsrk I get it.
18:01 Heffalump and I think I would also have written a comment in the bit that gets the blank password in the first place, saying "if the password is blank then we disable password login for the account by setting uPaS to Nothing", or whatever precisely is going on there.
18:04 Heffalump anything else to discuss today?
18:06 bsrk Ah, that is it.
18:06 Heffalump if you want to meet next on Thurs I'll be shifted half an hour earlier - i.e. available 1730 -> 1850 UK time with the usual tunnel gap.
18:07 Heffalump it's all looking nice so far, anyway!
18:07 bsrk Hmm, Thurday seem fine.
18:10 bsrk Heffalump: Try the sync again. It should work now.
18:11 Heffalump seemed to work
18:12 bsrk Ah, good.
18:12 Heffalump I think it would be useful to be clearer about what "sync" does - e.g. which way the transfer goes
18:12 Heffalump and what gets updated
18:13 bsrk By sync, I mean your darcsden account will import your github keys, and from now on, you will be able to login with your github account.
18:13 bsrk Syncing again will reimport keys.
18:14 bsrk Perhaps, it should be "Link with your github account" (first use) and "Reimport keys with with your github account" (subsequent uses)
18:15 carter joined #darcs
18:15 Heffalump "Link" implies a live link to me, i.e. it'll update if github does.
18:16 Heffalump I think when you register via github, it's obvious that it would import state from there and that's quite visible.
18:16 Heffalump I like the "Reimport" terminology for what's now sync.
18:17 Heffalump I wonder if it would be nicer in web design terms to have that link right next to the keys
18:17 Heffalump e.g. the label "pubkeys" and under that a "reimport from github" link
18:17 edwardk joined #darcs
18:17 Heffalump I think actions are generally supposed to be buttons rather than links btw.
18:18 Heffalump But my web design knowledge and sense of aesthetics suck :-)
18:18 Heffalump anyway, train arriving - see you Thursday (or I'll be around in a while, but it's late for you)
18:18 bsrk Ah, that makes sense.
18:19 bsrk See you!
18:22 gh_ joined #darcs
18:24 gh_ Heffalump, owst, about your conversation on using features from GHC 7.4 and 7.6, we have this piece of policy which sort of enables us to do the switch now: http://darcs.net/Development/Policy#ghc-versions
18:32 Heffalump gh_: yeah, I was wondering if being slow with 2.10 would lead to that :-)
18:33 gh_ looks like the last two HP use ghc >= 7.4 http://www.haskell.org/platform/changelog.html
18:36 gh_ the one from  June 2012 seems to introduce ghc 7.4
18:43 Heffalump the one just released has GHC 7.
18:43 Heffalump 6
18:43 gh_ yes
18:43 gh_ for the older versions of HP, it's surprisingly complicated to find which version of ghc is provided
18:58 dolio joined #darcs
18:59 uniquenick joined #darcs
19:33 whaletechno joined #darcs
19:43 lelix joined #darcs
19:50 uniquenick joined #darcs
19:52 lelix joined #darcs
19:52 edwardk joined #darcs
19:53 mizu_no_oto joined #darcs
20:01 mizu_no_oto joined #darcs
20:08 edwardk joined #darcs
20:18 konundra joined #darcs
20:35 edwardk joined #darcs
20:47 Heffalump joined #darcs
20:54 alexei___ joined #darcs
21:04 konundra joined #darcs
21:20 carter joined #darcs
22:39 edwardk joined #darcs
23:06 dolio joined #darcs
23:08 mizu_no_oto joined #darcs
23:13 preflex joined #darcs
23:23 edwardk joined #darcs
23:38 preflex joined #darcs
23:55 edwardk joined #darcs
23:57 mizu_no_oto joined #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary