Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-12-04

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

All times shown according to UTC.

Time Nick Message
00:02 maerwald sm: I found a solution to the post thing
00:02 maerwald https://gist.github.com/hasufell/9d4e5cd46697b2960826
00:08 sm maerwald: clever
00:18 maerwald sm: http://hub.darcs.net/maerwald/darcsden-mail-notify/patch/2c79edf6310e02d67663bafcedfcf1f03cdf2ea0
00:39 sm maerwald: nice, deployed to hub-dev
00:39 maerwald do they share the same db?
00:39 sm yes
00:40 sm which does require a little caution
00:41 maerwald I think we should focus on getting https now
00:41 maerwald probably via front proxy
00:44 sm I just heard about letsencrypt.org, which sounds helpful
00:44 maerwald http://www.cyberciti.biz/tips/using-nginx-as-reverse-proxy.html
00:44 maerwald pretty easy
00:44 sm thanks
00:46 maerwald it's basically just: location / { proxy_pass http://127.0.0.1:8900 ; }
00:46 maerwald inside the server block
00:51 maerwald I think for free certificates most people use https://www.startssl.com/ or https://www.wosign.com/english/
01:24 mizu_no_oto joined #darcs
01:42 Big_G joined #darcs
02:04 mizu_no_oto joined #darcs
03:06 Big_G joined #darcs
12:00 mizu_no_oto joined #darcs
12:46 mizu_no_oto joined #darcs
14:24 siel joined #darcs
15:03 maerwald Heffalump: so... what's the concern of compiling darcs with -threaded? We need -threaded to enable ssl support
15:04 maerwald afais, darcs locks repositories anyway when it's applying stuff on them
15:08 sm concurrent libdarcs threads mess up each others' current working directory, which breaks everything, not just writes
15:08 sm good morning/afternoon
15:08 maerwald so if all darcs operations run in bounded threads, that would be fine?
15:12 sm what's a bounded thread ?
15:12 maerwald "A bound thread is a haskell thread that is bound to an operating system thread. While the bound thread is still scheduled by the Haskell run-time system, the operating system thread takes care of all the foreign calls made by the bound thread."
15:13 sm sound like it might work
15:22 sm maerwald: I made a start on configuring ssl on my reverse proxy - more to do. Maybe there's no need to add it to darcsden ?
15:23 maerwald I don't think it's a good idea to require everyone to run a reverse proxy
15:23 maerwald those are deployment details that should be left to the admin
15:25 sm I wonder how many people run public darcsden instances, and how many of those would choose *not* to use a reverse proxy, and how much effort we should spend on that use case instead of other things
15:25 sm right now there is one known public production instance in the world
15:25 maerwald I don't know. But it's not really good design to not have ssl support in-house
15:25 sm fair enough
15:29 maerwald but all this suggests that it's rather something that needs to be fixed in libdarcs...
15:30 Riastradh joined #darcs
15:31 sm certainly
15:32 maerwald and... we could still merge the ssl patches (once they are reworked), because they don't enable -threaded per-se. You'll just not be able to compile it then
15:32 sm I'm not quite sure how, but I seem to remember there's one file/directory lib on hackage that maintains a haskell thread local working directory
15:32 maerwald that would avoid a lot of rebasing
15:33 sm are there a lot of changes ? could you just postpone this till the darcs issue is worked out ?
15:34 maerwald if it's ever worked out, so I'd much rather have it merged and be a disabled codepath
15:35 maerwald people can compile stuff manually with -threaded then if they want to test it
15:35 maerwald and it can be left undocumented
15:35 sm if it depends on darcs fixes, and we don't know when those might happen, I don't see that it belongs in the darcsden code base yet
15:36 maerwald that depends on how you look at it. I am running darcsden with those things right now
15:36 maerwald so I wouldn't say it "depends" on the darcs fixes
15:36 maerwald it's just unsafe
15:36 maerwald and it's not possible to enable it per-se, because -threaded is disabled for libdarcs
15:37 sm so you want to run a public instance with SSL, without a reverse proxy, and you don't care that it will break if it gets concurrent users
15:37 sm ?
15:37 maerwald that's not what I said.
15:38 maerwald I am saying that it'll make maintaining the ssl codepath more difficult it will be a concurrent branch and I will probably just give up on it then
15:39 sm if there's no actual need for it yet, then I think that would be the right course - rather than maintaining dead code
15:40 maerwald no need for ssl? ugh.
15:40 sm built-in native ssl.
15:40 maerwald pretty weird opinion
15:41 sm I think you're misinterpreting me
15:44 sm look, if it works, by all means add it. If it makes the app break, but you think it's still valuable for some use cases, then add it. If it's not going to be used any time soon, adding it now just to avoid future rebasing effort is not a good reason IMHO
15:47 maerwald well, then I'm probably not going to work on it further
15:49 maerwald I am not doing concurrent branches if the timeframe is not forseeable
15:49 maerwald that always ends up with a lot of wasted time
15:50 sm would you want to take a look at the darcs situation and see how big a task that is ?
15:51 maerwald I don't have the time to hack on darcs right now and starting with next month, I don't have time for pretty much anything
15:52 sm I hear you
18:11 Igloo joined #darcs
18:32 maerwald Heffalump: does darcs use relative paths on filesystem operations anywhere?
18:39 maerwald well yes, "withRepository"
18:40 maerwald that function basically has to be removed from the API and then we are thread-safe. The cli tools can ofc have it
19:01 Igloo joined #darcs
19:40 Igloo joined #darcs
20:02 Igloo joined #darcs
20:29 Heffalump maerwald: it uses them everywhere
20:29 maerwald so we just have to replace it :D
20:29 Heffalump replace what? darcs?
20:29 maerwald we need absolute paths
20:30 maerwald the current directory is a global state in a threaded program
20:30 Heffalump I mean that the entire repository handling code uses relative paths, and the functions that enter the repository handling code like withRepository and withRepositoryDirectory have to set up cwd to accomodate that.
20:30 Heffalump it's a trade-off: it would take effort to implement the change, and it might make the darcs cli slower.
20:30 maerwald we can move the "cwd code" to the cli tools
20:31 Heffalump if we want both relative and absolute path handling, then we need some kind of overloading so the core repository code can switch between the two
20:32 Heffalump if the performance drop isn't too significant (which would need at least a vague attempt at measuring it to prove that), I think we should just switch to absolute paths. But it's still significant implementation effort, AFAIK.

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