Camelia, the Perl 6 bug

IRC log for #darcs, 2013-06-28

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

All times shown according to UTC.

Time Nick Message
01:13 konundra joined #darcs
01:32 edwardk joined #darcs
01:38 mizu_no_oto joined #darcs
01:55 intripoon joined #darcs
03:50 edwardk joined #darcs
04:09 mizu_no_oto joined #darcs
05:02 carter joined #darcs
05:34 carter joined #darcs
06:03 carter joined #darcs
06:44 preflex joined #darcs
07:11 Heffalump why do darcs patches have to be invertible?
07:12 raichoo joined #darcs
07:12 Heffalump I'm trying to work out if it's fundamental to cherry-picking, or just a convenience that allows us to unpull etc without replaying the whole history from an old snapshot
07:14 alexei___ joined #darcs
07:24 lelit joined #darcs
07:40 carter joined #darcs
07:41 carter joined #darcs
07:42 amgarching joined #darcs
07:45 amgarching Dear developers of a third largest Haskell application, do I need to worry if I see a "one-time" warning "WARNING: Doing a one-time conversion of pristine format." more than once? I happen to use the same repo (on NFS) from Lenny and Wheezy having Darcs 2.8.1 and 2.0.2, respectively.
10:04 whaletechno joined #darcs
10:15 owst joined #darcs
10:49 donri joined #darcs
11:03 favonia joined #darcs
11:36 gh_ joined #darcs
11:49 Heffalump amgarching: that's a bug in darcs, you don't really need to worry about it
11:51 owst Heffalump: did you come to a conclusion about invertability of patches being required?
12:18 mizu_no_oto joined #darcs
12:56 konundra joined #darcs
13:03 edwardk joined #darcs
13:11 uniquenick joined #darcs
13:16 raichoo joined #darcs
13:59 mizu_no_oto joined #darcs
15:19 dolio joined #darcs
15:55 Kindaro joined #darcs
15:59 raichoo joined #darcs
16:43 bsrk joined #darcs
16:48 sm g'day all
17:00 mizu_no_oto joined #darcs
17:05 * Heffalump appears
17:05 Heffalump bsrk, sm: shall we start with the review?
17:06 sm sounds good
17:06 Heffalump bsrk: also, where's your blog post? :-)
17:06 sm hi bsrk
17:06 Heffalump I guess the first issue is the -threaded problem, right?
17:07 Heffalump My understanding is that simply changing darcsden to use -threaded with no other changes causes random failures to find the CSS files etc, perhaps because libdarcs changes CWD from under the rest of darcsden
17:07 Heffalump is that right?
17:07 sm that's right, and also occasional worse failures
17:07 Heffalump what were the worse failures, if you remember
17:07 Heffalump ?
17:08 sm just errors in random user operations IIRC
17:09 sm whenever requests come in concurrently there's a chance of the working directory being messed up. We talked about other workarounds such as forcing absolute file paths and that work out
17:09 Kindaro_ joined #darcs
17:09 sm that didn't work out
17:09 Heffalump fixing libdarcs is still on my radar but in the "quite difficult" bucket :-/
17:10 bsrk hi
17:10 sm Heffalump: I don't see a great alternative.. and this seems like maybe the right time ?
17:10 sm what makes it hard ?
17:11 Heffalump (1) there's a lot of darcs code to change (2) things darcs calls like hashed-storage may need to be changed
17:12 sm is it a large quantity of relatively simple changes ? I guess it might be a little tricky to be sure we've fixed it everywhere
17:12 bsrk Can we use threaded selectivily?
17:12 bsrk There are only a few functions in 2 files that use it,
17:13 sm hmm, like for specific modules ?
17:13 bsrk Yes
17:13 sm I'm guessing it's a property of the RTS so no
17:14 Heffalump no, we definitely can't use it selectively
17:14 Heffalump can you explain why HsOpenSSL needs it?
17:14 Heffalump what specifcailly does it do?
17:14 Heffalump <going underground>...
17:17 bsrk I think HsOpenSSL needs threaded because openssl requires a threaded runtime;
17:18 bsrk For example, this code https://gist.github.com/bsrkaditya/5886392
17:19 bsrk fails with error: user error (RTS doesn't support multiple OS threads (use ghc -threaded when linking))
17:19 bsrk In any case, I we can make https requests without -threaded, the issue is solved.
17:21 bsrk I have talked with the maintainer of http-streams about the issue, and he recommended that we should solve the issue with libdarcs. :-/
17:21 sm yeah, I don't blame him :)
17:21 sm what about using something else for https, like http-conduit ?
17:23 bsrk Hmm, it is using tls and not HsOpenSSL, so it may not run into this problem..
17:23 sm could be worth a test
17:24 sm I guess another option, a hacky one, is to build a separate helper executable and talk to it by local socket or something
17:24 alexei___ joined #darcs
17:27 sm http://hub.darcs.net/simon/rss2irc uses http-conduit, or there may be a simpler example to test a non-threaded build
17:27 sm bsrk: shall we start looking at your patches ? I expect Heffalump will reemerge from the mines soon
17:28 bsrk okay
17:29 sm ok.. for anyone else helping: we're at http://hub.darcs.net/simon/darcsden/patches -> Aditya's darcsden-gsoc
17:29 sm let's start at the bottom
17:30 sm show my repos page after login - I like it, though ultimately we might want it to just stay at the current urlo
17:30 sm url
17:31 sm now some packs-related patches..
17:31 bsrk Just want to point out,
17:32 bsrk the backward order is not broken feature wise
17:32 sm yeah I saw.. no worries
17:32 bsrk okay
17:32 sm by the way how did you do the history cleanup ?
17:33 sm unrecorded everything and started over ?
17:33 bsrk yes
17:33 sm good to know
17:35 * Heffalump reappears
17:35 sm these 4 packs patches look just fine, except the last makes it a user ui action and we discussed that users probably shouldn't even know about packs
17:35 sm we could apply it now and use it for testing and maybe streamline ui later
17:36 Heffalump so we'd have a UI we didn't like indefinitely?
17:36 Heffalump I guess the question is whether packs+ugly UI is better than no packs
17:37 sm well it's just a very little UI, very easy to remove. If we can jump straight to automatic packs requiring no UI, even better
17:37 sm when should it automatically do packs, bsrk ?
17:38 sm we can also focus on the other features and discuss packs more later
17:38 Heffalump if you would be fine with living with the UI indefinitely in case bsrk doesn't manage the atuomatic version, then merging now makes sense
17:38 bsrk I'd say when the number of patches is large enough
17:38 bsrk but IMO, the user should still have the manual option
17:39 bsrk it's unobtruisve enough in the settings page
17:39 sm Not a big deal, we can iterate. I'll merge the packs stuff after all the rest, perhaps it will have evolved further by then
17:39 bsrk okay
17:40 sm btw Heffalump do you have darcs.net dns manager access ? it would be nice to have new-hub.darcs.net, with the same ip address, for the dev instance
17:40 sm http://hub.darcs.net:8901/ works for now
17:41 sm it's not running bsrk's stuff yet
17:41 Heffalump i don't know anything about the setup of the dns - do you?
17:41 sm I don't have access, probably kowey then
17:42 sm @tell kowey it would be nice to have new-hub.darcs.net, with the same ip address as hub.darcs.net, for the dev instance
17:42 lambdabot Consider it noted.
17:42 sm moving on ? password reminder patches start at "return the updated session with setUser"
17:44 sm IIRC the mime-mail package is cross platform, or if not you took care of it in another patch
17:45 sm those 6 patches look fine
17:46 sm then there's the two "password" patches further up.. "Change the password fields of the user, so that they are optional"
17:47 sm bsrk: that made me wonder a bit. Is it possible to create or have a user account without a password ?
17:47 sm related, I noticed a hardcoded pw "password1" in the code (not yours). Any idea when that is used ?
17:47 bsrk Yes, you can have a account without a password;
17:48 bsrk Currently, the only way is if you register with oauth
17:48 bsrk If you just take in the password reminder patches
17:48 sm makes sense, what if you try a non-auth login to that user account then ?
17:49 bsrk It is not possible, with the current implementation
17:49 sm ok
17:49 sm so no obvious security issues here
17:49 konundra joined #darcs
17:50 sm overall I must say I'm not expert enough to spot security issues in this stuff, best I can do is ask dumb questions and try things
17:50 sm how do you guys feel, is the new stuff secure ?
17:51 Heffalump I couldn't spot any obvious problems. I'd feel a lot more comfortable with tests.
17:52 sm tests ? like.... ?
17:52 sm testing web logins, access control and so on ?
17:54 Heffalump yes, tests that sequences of actions that don't involve supplying proper credentials don't get you logged in
17:55 Heffalump we discussed scripting that kind of test with a web client, or abstracting the server code and calling it directly
17:55 sm yes that would be nice, and also a great starting point for a test suite in general
17:55 sm even one test would be good
17:55 Heffalump so the question is whether it's considered essential for adding these features or not
17:56 sm I think a test or three is a reasonable thing to ask. bsrk, would that be ok ?
17:57 sm snap may have some nice test helper lib
17:57 bsrk Hmm, I am not sure what the tests should be, but okay I'll think of something
17:58 Heffalump are you not sure how to write the testing infrastructure, or what the tests should actually test?
17:59 bsrk Both, but the second most of all
17:59 sm #snap should help with the first
18:00 sm for the second, we need a test db with a test user, then tests that check access denied without login, and access granted with regular login (for a start)
18:01 sm setting up the test db seems a little interesting
18:01 sm an easier alternative is to make tests that run against hub.darcs.net
18:01 sm #snap would help with testing too
18:02 sm I'll merge these changes, meanwhile
18:03 Heffalump running against hub.darcs sounds bad to me, how would anyone test their local work?
18:03 sm btw, seeing http://hackage.haskell.org/packag​es/archive/pkg-list.html#cat:snap reminds me that snap has its own auth system - which is probably worth looking at
18:04 sm Heffalump: just an easier first step. But yes.
18:05 sm more patches ? The 3 starting at "Create handlers for edit file" are web edit ui related
18:05 bsrk yes
18:06 Heffalump bear in mind that the github integration involves special knowledge of github, so a generic web servicse authentication mechanism couldn't be used directly
18:06 bsrk but I don't recommend pulling them
18:06 bsrk there is still a bug that's pretty bad
18:07 bsrk where when a file is first edited, the patch contains the entire file
18:07 sm bsrk: ok. If it helps, you could fork a child of this repo and keep your not-ready-to-merge patches there
18:08 bsrk I'll do that
18:08 sm the compressed codemirror js is 376k (and hub can't display it) - probably nothing to be done about that
18:08 sm I'm looking forward to trying it out
18:09 bsrk you can try it at http://www.darcsden-gsoc.biz:8900/
18:10 sm ah, I see the threaded error when trying to log in via github there
18:11 bsrk yeah, I have removed the threaded option;
18:11 bsrk so no https requests
18:11 bsrk I will shift to http-conduict
18:11 sm incidentally I don't know if UI polish is in or out of scope here
18:11 bsrk what do you mean?
18:12 sm like, I see the github/google login buttons layout and wording could be a little more integrated with the old login button.. should I give feedback on such things ? there are always a lot of these
18:13 sm if not, it'll just get polished by someone eventually
18:13 bsrk Yes please
18:14 alexei___ joined #darcs
18:14 sm I don't want to soak up all your time on ui stuff. But ok, it might be nice to move the github/google login buttons & pw link inside the box with the other button, and it would be great to use the official logos like most sites do
18:15 Heffalump sm: I think you should give feedback, if only to help with future features.
18:16 sm also consistent wording, the app currently uses "login" and "register". sign in and sign up can be confused
18:16 bsrk joined #darcs
18:17 sm we can change wording, but let's keep it consistent across the app
18:18 sm so the oauth stuff starts at "Add dependencies HSOpenSSL, network, http-streams for OAuth feature"
18:18 sm 9 patches
18:18 bsrk Yes
18:19 sm which will change if you switch to http-conduit. Either way, this stuff will still be cross platform I believe
18:19 Heffalump so how will https work in the future world?
18:19 Heffalump does http-conduit use the tls package?
18:20 sm yes IIRC. It does https
18:21 sm http://hub.darcs.net/Aditya/darcsd​en-gsoc/patch/20130626111553-ae621 is the big oauth patch. Seems to work.
18:21 sm http://hub.darcs.net/Aditya/darcsd​en-gsoc/patch/20130626115338-ae621 shows how the terminology is a little different for github and google ("gid" vs "gsub"). This seems a little obscure
18:22 sm but it works
18:22 sm since it's getting baked into our db though, maybe there's some generic thing we can call these ? I expect there'll be more in future
18:23 sm github id ? google id ? I don't know what a "sub" is
18:23 Guest79013 joined #darcs
18:24 bsrk It's supposed to be an identifier to uniquley identify a google/github account
18:24 sm the oauth identity gets saved in the session (cookie, and redis)
18:24 bsrk No; it's only used when registering
18:24 sm oh
18:24 bsrk the identitiy is saved in the User though
18:25 sm yeah
18:25 sm in this big patch I noticed you put nice docs in the commit message - I think the 1/2/3 items would be even better as function haddocks
18:26 sm eg I can't tell what getSubAndEmail  does at a glance
18:27 sm regarding sync: do other oauth-supporting sites have this feature ? I don't remember seeing it
18:27 sm seems like a useful thing though
18:27 bsrk It's like import settings
18:28 bsrk and also like uniting with your other accounts
18:28 Kindaro joined #darcs
18:28 sm and finally we have a small conflict resolution patch, which shouldn't cause trouble
18:29 sm great work bsrk
18:29 Heffalump did you fix the "sync" language, btw? I found it confusing to know which way the data was flowing.
18:30 bsrk It's called link now, I think
18:30 Kindaro joined #darcs
18:31 Heffalump does that mean it's continuously linked, or a one-shot update?
18:32 bsrk one-shot update
18:32 edwardk joined #darcs
18:32 bsrk it is link, in the sense that once you linked the account you can login with it
18:33 sm so when you move the not-ready-to-merge patches out of this repo, I'll deploy the rest on dev here and we can start using them more intensively
18:33 sm (bsrk)
18:34 bsrk okay, I'll move out file edit, and use http-conduit instead
18:34 sm cool
18:34 bsrk so that oauth can stay
18:34 sm well let's see if http-conduit works without threaded
18:34 bsrk it does; I just tested it
18:35 Heffalump bsrk: oh, I see. So where's the "update keys from github" thing gone?
18:35 bsrk Link with github also update keys
18:41 Kindaro_ joined #darcs
18:45 lelit joined #darcs
18:48 sm by the way Heffalump: it seems I rejected nand's "Switch to attoparsec for parsing auto-closing" as obfuscatory, feel free to delete or detach your darcsden-attoparsec branch
18:49 * sm is trying to get back to a clean branches page, but it's probably going to require programming
18:50 Heffalump obfuscatory in what sense?
18:52 sm just that it replaces a clear, familiar and easy to tweak regexp with something that's less so
18:52 Kindaro joined #darcs
18:54 Heffalump the main attraction for me was that it increased portability
18:54 sm isn't pcre-light portable ?
18:55 Heffalump installing C libraries on Windows is a pain, however portable they are
18:55 Heffalump I shall go and look at the actual code and see if I can defend its readability, however.
18:55 sm ack. Ok I misunderstood that
18:55 sm http://hub.darcs.net/nand/darcs​den/patch/20120829130330-cd4f6
18:56 Heffalump thanks
18:56 Heffalump ok, so I know oth regexps and combinator parsing fairly well, and I think the combinator code is quite clear and readable
18:56 sm I think every other system that hooks in to commit messages configures that with regexps
18:57 Heffalump obviously combinator parsers are a niche compared to regexps, but they are common in the Haskell world
18:57 sm and if I need to tweak it, I only have one line to look at
18:57 sm and I don't have to learn unfamiliar parser combinators
18:57 sm (I'm a parsec man :)
18:57 haasn you have two lines to look at in my patch =P
18:58 sm oh hi haasn.. = nand ?
18:58 haasn it's possible some of the boilerplate could be reduced, neither then nor now was I a huge expert in attoparsec
18:58 haasn sm: yes
18:58 Heffalump it's the same number of lines of code before and after (ignoring import statements)
18:59 sm it's totally less practical for a server admin :)
18:59 sm aside from the installability issue, which I agree is more important
18:59 Heffalump I see your point about regexps, if we wanted to expose the patterns to users in future regexps are much better
18:59 haasn “resolves = resolves' <|> anyChar *> resolves” <- must be some built-in function for that
19:00 Heffalump presumably there are native Haskell regex libraries, perhaps that's the right compromise here.
19:00 haasn I like how the regex already has duplication of the #([0-9]+) bits
19:00 haasn that could be improved, even if you decide to stick with regex
19:01 haasn something like “(closes|resolves|fixes) #([0-9]+)” with some extra bit added to the first () to make it not count as a match
19:01 * haasn isn't a great expert on PCRE
19:01 sm yes a portable (ie native haskell) regex lib would be good. I've always used regexpr for this, it has some performance issues but might be fine here
19:02 sm haasn: +1
19:03 sm here's http://hackage.haskell.org/package/regexpr . Most of the time it works well and it has better ERE support than most
19:04 Heffalump re forks in general, I am still of the strong opinion that "patch submissions" rather than entire repositories are the right way to go
19:04 alexei___ joined #darcs
19:05 haasn personally I just don't like the uncleanliness of the regex-matching functions, all of the crap they return and how you have to piece together which match means what / parse it again
19:05 sm there's also http://hackage.haskell.org/package/regexchar , I haven't tried this one
19:05 haasn with the attoparsec approach, the numerical parsing is directly taken care of, and it returns its data in a type safe way
19:06 haasn irrelevant here since it's a small regex
19:06 haasn but still
19:06 Heffalump for me customization is the killer argument for regexps
19:07 sm haasn: I feel this is normally something site admins would be tweaking, not haskell programmers, (even to currently it requires a recompile) so it should suit them
19:08 haasn yes, I believe that's a fair point; if only for eg. localization purposes
19:08 sm it's not a huge deal, but it just didn't feel like a step forward, and I didn't understand that it helped windows installtion
19:08 haasn (though changing ["foo", "bar", "bat"] is not much harder than changing (foo|bar|bat), mind)
19:09 haasn that must have been my fault
19:09 haasn 2012-08-29 17:16:10 sm nand`: thanks for the patch, this is a step towards easier installation on windows right ?
19:09 sm regexps are just another parser syntax, and the most concise one there is IMHO
19:09 haasn 2012-08-29 17:16:56 nand` no idea
19:10 sm and it's a standard
19:10 sm oh ha
19:10 Heffalump sm: only if the language is regular!
19:10 edwardk joined #darcs
19:10 haasn PCRE is far from regular these days
19:11 sm sure sure, I realize they're not good for all parsing
19:11 Heffalump ok, but then we're well into "abuse of regexps" territory :-)
19:11 mtp regexmucil, with refreshing fiber for regularity
19:11 Heffalump anyway, I'm sold on keeping regexps.
19:11 sm we're just picking out #NNN for the love of roundy :)
19:11 * sm loves his regexps
19:11 haasn PCRE can parse CSGs right?
19:11 Heffalump I can live with the Windows pain for now, and I might get round to submitting a patch to switch to a native library if someone else doesn't get there first
19:12 Kindaro joined #darcs
19:12 sm Heffalump: if it reduces pain, I'll merge this now. I've been resisting because I feel noone will restore the regexps. :) But so be it
19:13 Heffalump I've been happily building on Windows with -f-closing for ages
19:13 Heffalump it's not like anyone we know is actually running a prod server on Windows
19:13 haasn ^
19:13 sm right, or ever will.....
19:13 Heffalump and I don't personally care about the feature on my laptop
19:13 sm it does mean you had to figure out to add that flag though
19:13 haasn I don't think this issue is our #1 priority =P
19:14 mizu_no_oto joined #darcs
19:14 sm another installation roadbump
19:14 Heffalump well, that's less obvious: currently internally at my company a major negative point for a commercial tool that is a bit darcsden-ish is that it only runs on Linux
19:14 carter joined #darcs
19:14 Heffalump (less obvious relating to "or ever will")
19:14 sm ok, fair enough I exaggerated
19:15 sm well.. that decides it: Patches for fork "nand/darcsden" could not be applied cleanly and have been skipped.
19:15 sm sorry haasn
19:15 haasn I have no idea how old that patch even is
19:15 haasn seems like a year old
19:16 sm yup
19:16 Heffalump I fixed up the conflicts and made a branch with that - I'll delete it now
19:16 sm oh, right, you did
19:16 sm let's have a new conflict free patch. I'll apply it, whatever it is
19:18 sm and haasn if you can obliterate the old one, or detach that branch, that'd be great
19:20 haasn Il
19:20 haasn I'll detach it, seeing as I'm not working on darcsden right now*
19:20 Heffalump is there UI to detach branches? I didn't realise.
19:20 sm haasn: and why not :(
19:20 sm we miss you
19:21 sm to detach, I would fork the branch again, then delete the parent
19:21 Heffalump ah :-)
19:21 sm "UI" :)
19:21 Kindaro_ joined #darcs
19:21 Heffalump VCUI : only for very commited users
19:21 sm UI for random attaching/detaching might be nice
19:22 Heffalump yes, just expose the "parent of" field in the UI
19:22 Heffalump IM "child of"
19:22 sm and by random I mean, arbitrary user-selected. Not guess a random repo on hub.
19:22 Heffalump :-)
19:22 sm fun as that would be
19:23 Heffalump btw I was thinking about what an appropriate way of identifying repositories would be if we can't rely on an id stored in couch
19:23 haasn there, detached successfully ;)
19:23 Heffalump any thoughts?
19:23 sm haasn: thanks. bendoyle is next. Interesting, he shows up on branches but has no public repos..
19:23 sm Heffalump: eh...
19:24 haasn sm: you mean his fork is private?
19:24 haasn also why do you want to clean up the branches page?
19:24 sm haasn: could be..
19:25 haasn if so then private forks should probably not be publicly displayed ;)
19:25 sm just decluttering. It's a distraction to keep filtering out the same old patches when I'm looking for things to merge
19:26 sm yes that would be a bug. More likely the db has a stale record
19:26 sm ohh.. it's because I clicked on bendoyle, which is another account
19:27 haasn sm: see, I think the branches problem you are describing is a UI shortcoming
19:27 sm yes, it is
19:27 sm Heffalump: where does that arise ? repos have stable urls
19:27 haasn first of all, ‘boring’ branches with no changes should be moved somewhere out of sight, to the very bottom, consolidated into a single line or not displayed at all
19:28 sm yes
19:29 Heffalump sm: in trying to remove the property that every repository has a record in CouchDB, I need to do something about the "parent" field
19:29 sm ahh
19:29 haasn next, the other repositories should be sorted according to how recent their last change was or something
19:29 haasn so you get interesting, new patches at the top of the page
19:29 sm Heffalump: well we can assume every storage will have some kind of id
19:30 haasn a completely alternate idea I had would be to consolidate all patches into a single ‘patches’ page, instead of separating them based on repository, and have them sorted by date. Which branch that patch comes from could be listed somewhere at the side (consolidated for patches which are shared among multiple branches), for example
19:30 alexei___ joined #darcs
19:30 sm haasn: I'd love to see it
19:31 sm I also want an ignore button that hides patches
19:31 haasn perhaps taking a leaf from bulletin boards, keep track of which patches the user has ‘read’
19:31 haasn and differentiate visually between ‘read’ and ‘unread’ patches
19:32 Moggle_ joined #darcs
19:32 haasn the page could include checkmarks and so forth for ‘bulk actions’ like ‘mark read’, ‘hide’ or ‘merge’
19:32 sm Heffalump: if the storage were just the filesystem, then.. might have to generate an id and store it as part of darcsden's metadata under _darcs ?
19:33 sm haasn: you know there's checkboxes and merge button when you have write access to the repo ?
19:34 haasn oh right
19:34 haasn I thought I remembered those
19:34 Heffalump sm: yeah, perhaps. I'd quite like just to be able to drop a new repo into the right folder and have it managed by darcsden, which implies that darcsden would have to generate the id the first time it noticed the repo.
19:34 haasn fair enough =)
19:34 sm but yes, more please
19:37 sm we also need someone to fix http://hub.darcs.net/simon/darcsden/issue/59 , it's causing restarts several times a day
19:37 Heffalump oops!
19:38 Heffalump any idea where the segfault emanates from?
19:38 sm which hopefully unnoticeable - but embarrassing :)
19:38 sm not really, some quirk with snap route parsing or our configuration
19:40 sm http://hub.darcs.net/simon/darcsden​/browse/src/DarcsDen/Handler.hs#30
19:42 Heffalump a segfault is a pretty bad thing for a Haskell program to do, though. It was more understandable for the ssh server under Windows because that clearly does rely on external libs.
19:42 sm yes it is
19:42 sm and funny, I can't reproduce it now
19:43 sm fixed by rebuilding with latest snap possibly ?
19:44 sm it hasn't happened since 4 days ago, when the current binaries were built. nice
19:48 sm that just leaves the oom-killer restarts, which are a bit less frequent
19:49 mtp the oom-killer is the worst
19:50 sm it's helping out here.. resetting the app when darcs does something huge
19:50 sm but I probably should limit just the darcshub process with cgroups or something
19:54 sm ah, https://github.com/snapfram​ework/snap-core/issues/179
20:03 edwardk joined #darcs
20:13 favonia joined #darcs
20:31 sm http://hub.darcs.net/simon/darcsden/issue/58 is the next serious bug, maybe in highlighting-kate ?
20:31 sm it would be good to set up each heap profiling of a dev instance
20:31 sm s/each/easy/
20:39 lelit joined #darcs
21:49 favonia joined #darcs
22:41 Kindaro_ joined #darcs
22:56 Kindaro__ joined #darcs
23:01 Kindaro joined #darcs

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