| Time |
S |
Nick |
Message |
| 00:45 |
|
|
twb joined #darcs |
| 01:18 |
|
|
JaffaCake1 joined #darcs |
| 06:39 |
|
|
balor joined #darcs |
| 08:02 |
|
|
darcscommitbot joined #darcs |
| 08:03 |
|
|
darcswikibot joined #darcs |
| 09:00 |
|
|
lelit joined #darcs |
| 09:29 |
|
|
gh_ joined #darcs |
| 09:55 |
|
|
gal_bolle joined #darcs |
| 11:09 |
|
|
kowey joined #darcs |
| 11:09 |
|
kowey |
morning |
| 11:43 |
|
gh_ |
hi |
| 11:50 |
|
kowey |
did slurpies just die? |
| 11:54 |
|
|
lelit joined #darcs |
| 11:55 |
|
|
vinc456 joined #darcs |
| 12:00 |
|
|
artemis joined #darcs |
| 12:21 |
|
gh_ |
they are one conflict away from dying |
| 12:21 |
|
gh_ |
(at least) |
| 13:09 |
|
|
kpreid_ joined #darcs |
| 13:16 |
|
|
nwf joined #darcs |
| 13:33 |
|
mornfall |
Hello. |
| 13:33 |
|
mornfall |
Well, the patches depend on the cabal 1.8 patches. |
| 13:33 |
|
mornfall |
(Due to changes in cabal file.) |
| 13:47 |
|
kowey |
we're still waiting on cabal-install to support those, right? |
| 13:49 |
|
* kowey |
tries to find a cabal-install ticket number that we can put in our tracker |
| 13:51 |
|
mornfall |
I could care less about cabal install on HEAD as long as cabal-install gets fixed before 2.5. |
| 13:52 |
|
mornfall |
(runghc Setup configure --user && cabal build && cabal copy works just fine..) |
| 13:53 |
|
kowey |
oh, well maybe that's not so bad |
| 13:54 |
|
kowey |
dcoutts: would you like me to file a ticket for http://irclog.perlgeek.de/darc[…]0-01-20#i_1916008 or is there one already? |
| 13:54 |
|
mornfall |
Well, it's less bad than waiting for the recompiles. |
| 13:55 |
|
mornfall |
And it makes life less bad on train, wrt. battery life. |
| 13:55 |
|
kowey |
hmm, so it's a choice between remembering that you can't use cabal configure (or cabal install; but that's a lesser issue) |
| 13:55 |
|
kowey |
or remembering to comment out the build-the-library bits? |
| 13:56 |
|
mornfall |
Unfortunately only the first one is an option. |
| 13:56 |
|
mornfall |
You have to haul files over to different directories to go back to the old behaviour. |
| 13:56 |
|
mornfall |
Well, maybe you can have multiple hs-source-dir thnigs. |
| 13:57 |
|
mornfall |
things* |
| 13:57 |
|
kowey |
(I meant without using the patch for the second one) |
| 13:57 |
|
mornfall |
So that could work, maybe. |
| 13:57 |
|
kowey |
well, I'm not too bothered either way... I think this can be settled with a quick informal poll of the review team |
| 13:57 |
|
mornfall |
Ah, well. I think that you'd have to twiddle the build-deps which were wrong up to move to 1.8. |
| 13:58 |
|
mornfall |
(Cabal used to do an union over them, and doesn't do anymore.) |
| 13:58 |
|
mornfall |
Maybe I fixed that in the first patch, not sure. |
| 13:59 |
|
kowey |
if this turns into a big issue, which I don't think it will, I'll coin-toss towards taking the patch now (provided dcoutts tells us we're likely to have this working by July) |
| 14:02 |
|
mornfall |
Well, I won't do it again (already did it twice) so the less conflicts it accumulates the better... |
| 14:03 |
|
kowey |
sounds like you need darcs rebase |
| 14:05 |
|
mornfall |
Or a more flexible HEAD. I'm no big fan of rebasing (although it could be useful from time to time... but getting to the point where you need rebase already sucks and it only gets worse afterwards...) |
| 14:05 |
|
|
artemis left #darcs |
| 14:07 |
|
kowey |
gal_bolle: do you care if "cabal configure" breaks for HEAD? (you can still do runghc Setup configure; and hopefully this will be fixed) |
| 14:07 |
|
* kowey |
is doing a little mini poll |
| 14:07 |
|
kowey |
(I think there's two of us in the don't-really-mind camp) |
| 14:07 |
|
gal_bolle |
on a regular basis |
| 14:07 |
|
gal_bolle |
or just once? |
| 14:08 |
|
kowey |
basically you won't be able to do "cabal configure" until cabal-install supports the new functionality that makes darcs build faster (so once, but indefinitely) |
| 14:08 |
|
gal_bolle |
it's not nice to non-core developers, which we want to become core developers eventually |
| 14:08 |
|
gal_bolle |
then i'm ok with it |
| 14:09 |
|
gal_bolle |
ok-ish actually |
| 14:09 |
|
kowey |
yeah, the non-core developers point is true. I guess we'd have to post a heads-up and also update the wiki |
| 14:11 |
|
mornfall |
We didn't do that when we started breaking without libicuuc, and neither when we bumped minimum cabal to 1.8 (both of which probably broke build for quite a few people). |
| 14:11 |
|
mornfall |
(Although I'm not saying we shouldn't start being more careful.) |
| 14:11 |
|
kowey |
oh yeah |
| 14:13 |
|
mornfall |
But then, it may take a couple weeks to get this through review anyway. |
| 14:14 |
|
mornfall |
Oh, hmm. |
| 14:14 |
|
mornfall |
I quite forgot that there are some hashed-storage patches that I have to get through to HEAD to make this work. |
| 14:17 |
|
gal_bolle |
mornfall: the problem i see is that getting libicuuc, or the latest version of cabal are problems where cabal itself will point to the solution, while making cabal configure just not work has no discoverable solution |
| 14:18 |
|
gal_bolle |
hence it needs to be widely announced (can cabal configure be made to yell something constructive at the developer?) |
| 14:18 |
|
mornfall |
Well, the amount of trouble icuuc caused among *core* devs was nontrivial. |
| 14:19 |
|
mornfall |
But yes, I guess this is even more obscure. |
| 14:19 |
|
mornfall |
But, anyway, bassoon. |
| 14:45 |
|
|
nomeata joined #darcs |
| 14:55 |
|
|
kpreid_ joined #darcs |
| 15:16 |
|
|
sshc joined #darcs |
| 15:36 |
|
|
lispy joined #darcs |
| 15:44 |
|
|
lispy joined #darcs |
| 15:48 |
|
lispy |
Good morning |
| 15:51 |
|
kowey |
morning lispy... got a quick developer poll question for you |
| 15:52 |
|
lispy |
Okay |
| 15:52 |
|
lispy |
emacs, BTW |
| 15:52 |
|
kowey |
do you particularly mind if "cabal configure" stops working on HEAD for some time? "runghc Setup configure" should still work |
| 15:52 |
|
kowey |
and building darcs would be a lot faster (because we won't be rebuilding the lib) |
| 15:52 |
|
lispy |
Well, it sounds painful to explain |
| 15:52 |
|
lispy |
huh |
| 15:52 |
|
kowey |
it's just a temporary thing until cabal-install supports the new features in Cabal 1.8 |
| 15:53 |
|
lispy |
actually, we need, runghc Setup configure --user, right? |
| 15:53 |
|
kowey |
yeah |
| 15:53 |
|
lispy |
And will runghc look at my cabal config for my other local options? |
| 15:53 |
|
* lispy |
has non-default cabal configs on all machines |
| 15:53 |
|
kowey |
I don't know. It's all about the Cabal library, right? |
| 15:53 |
|
lispy |
I don't know too! |
| 15:54 |
|
lispy |
Do we have an estimate on how long we'd have to wait for cabal-install to catch up? |
| 15:54 |
|
kowey |
our deadline would be the Darcs 2.5 release |
| 15:54 |
|
kowey |
after which I suppose we'd have to rollback the change |
| 15:55 |
|
lispy |
So, we'd never make a stable release that required this hoop jumping? |
| 15:55 |
|
kowey |
certainly not |
| 15:55 |
|
lispy |
And, BTW, what does this hoop jumping gain us again? |
| 15:55 |
|
kowey |
much faster darcs building without any intervention (like commenting out bits of the cabal file) |
| 15:56 |
|
lispy |
Does cabal-install have a fallback? eg., will it go back to slow builds if you're missing the feature? |
| 15:56 |
|
kowey |
http://bugs.darcs.net/patch147 <-- is what brought this up |
| 15:56 |
|
kowey |
also mornfall's recent slurpy-killing patch has a dependency on the above, so it'd be convenient |
| 15:57 |
|
kowey |
we don't have to make a decision until that passes review, of course, but it's nice to get an idea of how much fuss this will cause |
| 15:58 |
|
lispy |
And I'm fussy? :) |
| 15:58 |
|
lispy |
<shrug> It certainly seems annoying to use features of tools that exist only in the dev version of the tool |
| 15:58 |
|
lispy |
Does the darcs repo version of cabal-install have this? |
| 15:59 |
|
kowey |
unfortunately, it'll just say "you stupid-head!" (precise wording pending my digging it up from the IRC logs) |
| 15:59 |
|
lispy |
I think it's a bit risky to depend on bleeding edge things from our toolchain, but it's equally annoying to work around bugs in the toolchain. |
| 16:00 |
|
lispy |
Oh, that is sad |
| 16:00 |
|
lispy |
I think most of us are fine with the build times at the moment? |
| 16:00 |
|
kowey |
I'm just taking a(n informal) poll |
| 16:00 |
|
lispy |
Has it been frustrating people? |
| 16:00 |
|
lispy |
ya, I see that. |
| 16:01 |
|
lispy |
So, I would say. If people have been fine with the status quo and we don't have a promise from the cabal-install devs about a release date, then we should hold off. |
| 16:01 |
|
lispy |
That's my opinion |
| 16:01 |
|
lispy |
Because, otherwise we start working in non-standard ways again |
| 16:01 |
|
kowey |
I think the *main* issue is that the thought of having to rebase the slurpy-killa may significantly increase mornfall's grumpiness ;-) |
| 16:02 |
|
lispy |
Yeah, that sucks too. |
| 16:02 |
|
kowey |
speaking from experience! |
| 16:02 |
|
kowey |
sorry to dredge that up |
| 16:02 |
|
lispy |
It's really hard to work on things when these dependencies come up |
| 16:03 |
|
lispy |
I know in the past I've been quite frustrated with the hardship that comes when you want to change a patch at the bottom of the stack |
| 16:04 |
|
lispy |
kowey: so, I guess for me. The answer depends a lot on what is going on with cabal-install |
| 16:05 |
|
kowey |
ok, thanks! I guess we should just wait a little while longer until we hear back from the Cabal folks |
| 16:05 |
|
kowey |
maybe this will be solved for us and made moot |
| 16:05 |
|
lispy |
Yeah |
| 16:05 |
|
lispy |
That would be flipp'n sweet |
| 16:05 |
|
lispy |
We could do the happy dance |
| 16:06 |
|
kowey |
2x if the slurpy-killa passes review |
| 16:06 |
|
lispy |
ya |
| 16:06 |
|
lispy |
maybe that would obviate the work I started with HunkHandles |
| 16:06 |
|
kowey |
every bit of homegrown code we kill in favour of a generic library a cause for minor celebration |
| 16:06 |
|
lispy |
Indeed |
| 16:07 |
|
lispy |
kowey: I think our permutations modules are in that same category |
| 16:07 |
|
lispy |
kowey: they are necessarily homegrown, but I strongly suspect they have rotten memory characteristics |
| 16:07 |
|
kowey |
here's the current list of things I know about - http://wiki.darcs.net/DarcsLibraries |
| 16:08 |
|
lispy |
kowey: I've been investigating Data.List.sort lately (to help a co-worker) and it turns out to be rotten. I wouldn't be at all surprised if our internal algorithms have all the same issues. |
| 16:09 |
|
kowey |
hmm, I would've thought that Data.List.sort has already been quite thoroughly inspected/tuned/etc |
| 16:09 |
|
lispy |
Because it needs copies, it takes at least 2x the input memory size |
| 16:10 |
|
lispy |
and it's sufficiently strict |
| 16:10 |
|
lispy |
So sorting big lists (even if they are lazy) uses all your memory |
| 16:10 |
|
|
kpreid_ joined #darcs |
| 16:10 |
|
lispy |
packages like uvector can do this sort of sorting way more efficiently |
| 16:11 |
|
lispy |
But, ultimately, I think we want something like an external sort |
| 16:11 |
|
lispy |
But, doing our permutations externally |
| 17:06 |
|
|
gal_bolle left #darcs |
| 17:44 |
|
|
kpreid_ joined #darcs |
| 17:53 |
|
|
tux_rocker joined #darcs |
| 18:54 |
|
|
tux_rocker joined #darcs |
| 19:01 |
|
|
kpreid_ joined #darcs |
| 19:04 |
|
|
tux_rocker joined #darcs |
| 19:13 |
|
|
tux_rocker_ joined #darcs |
| 20:35 |
|
|
kpreid_ joined #darcs |
| 22:11 |
|
* Heffalump |
sends in a big patch sequence, just to keep up with mornfall |
| 22:22 |
|
lispy |
Heffalump: haha |
| 22:22 |
|
lispy |
Heffalump: I'm really excited to see it though |
| 22:22 |
|
lispy |
Heffalump: BTW, Petr's might need to wait for a cabal feature |
| 22:22 |
|
lispy |
Heffalump: if that is the case, it might be better to deal with conflicts in his patches than yours |
| 22:30 |
|
Heffalump |
his slurpy removal might need a cabal feature? |
| 22:39 |
|
Heffalump |
btw I think gal_bolle could also review the witnesses work so if you're busy or want a second opinion talk to him |
| 22:42 |
|
lispy |
Heffalump: cool, I already assigned it to myself |
| 22:42 |
|
lispy |
Heffalump: Some how Petr's patches depend on a .cabal file change |
| 22:43 |
|
Heffalump |
oh right |
| 22:43 |
|
lispy |
Heffalump: we need transplant :) |
| 22:45 |
|
Heffalump |
I have it :-) |
| 22:45 |
|
Heffalump |
not sure I'd want to inflict it on anyone else just yet, but I could... |
| 22:46 |
|
* Heffalump |
actually developed the witnesses and the rebase patches together, and then used rebase to pull out the witness patches. |
| 22:47 |
|
Heffalump |
kowey: (I assume you'll read this in logs): I'm not too fussed about the cabal 1.8 thing but we do need certainty about cabal-install becoming available in time, and also I'm quite happy with working locally with buildable: False on the library so personally I'm not desperate for the new features. |
| 22:48 |
|
Heffalump |
also if it'd help I could rebase mornfall's patches for him, but obviously that may still cause him pain if he has other local stuff depending on them |
| 22:50 |
|
lispy |
Heffalump: ah cool. So transplant is ready? |
| 22:51 |
|
Heffalump |
err, no |
| 22:51 |
|
Heffalump |
hence "not sure I'd want to inflict it on anyone else just yet" |
| 22:51 |
|
Heffalump |
but it's usable enough that I'm using it |
| 22:52 |
|
Heffalump |
and "rebase" is the definitive name for the feature, unless a collective decision is taken otherwise, because there's no point in not using the existing git name |
| 22:53 |
|
lispy |
Is it actually the same as rebase? |
| 22:53 |
|
Heffalump |
yes |
| 22:53 |
|
lispy |
Also, David has said he would rather not use the name rebase |
| 22:53 |
|
Heffalump |
I thought he expressed opposition to the feature |
| 22:53 |
|
lispy |
So out of respect for him it seems polite? |
| 22:53 |
|
Heffalump |
rather than the name specifically |
| 22:54 |
|
lispy |
I could have sworn there is a ticket in roundup that says he doesn't like the name rebase but transplant is okay with him |
| 22:55 |
|
Heffalump |
http://bugs.darcs.net/issue938 is the only comment from him I'm aware of |
| 22:55 |
|
Heffalump |
anyway, given iolaus, half of what he said there seems to be superceded |
| 22:56 |
|
lispy |
"I would not call anything in darcs rebase. " <-- that's the phrase I was thinking of, perhaps it's not as strong as I remembered it |
| 22:56 |
|
lispy |
I took that to mean, we should never have a 'darcs rebase' |
| 22:57 |
|
lispy |
But other interpretations exist |
| 22:57 |
|
lispy |
Like, that feature doesn't exist yet |
| 22:57 |
|
Heffalump |
I took it as opposition to the feature itself, though what he says in the same comment about commutation makes me think he didn't quite get the use case for it. |
| 23:00 |
|
Heffalump |
anyway, I'm open to discussion on the name, but my current working name for is "rebase", I believe that it fulfills the same use cases as the "transplant" that's been discussed (I think I originally came up with the name "transplant", though it's possible I didn't or that multiple people did so independently) |
| 23:00 |
|
lispy |
I think we independently came up with it |
| 23:01 |
|
Heffalump |
and I would personally be against changing its name just because David had expressed opposition to one name in the past, even if he actually did so which I think is doubtful, but would be happy to accept a collective decision on the subject. |
| 23:01 |
|
lispy |
So my thought on how a 'transplant' should work is that you move the diffs to a new repo as new patches (instead of moving patches as we normally do) |
| 23:01 |
|
Heffalump |
rebase allows that. |
| 23:02 |
|
lispy |
So your rebase could be different than my 'transplant'? |
| 23:02 |
|
Heffalump |
arguably the two names have a slightly different emphasis in how you might intuitively think they work |
| 23:02 |
|
Heffalump |
but I think ultimately they're equivalent in what they enable |
| 23:04 |
|
lispy |
Nice |
| 23:04 |
|
lispy |
Sounds like progress |
| 23:10 |
|
lispy |
Heffalump: when you get a chance, could you archive your patches to implement rebase on the ticket? Not because they are ready for review or application, but so we don't lose them if your computer gets hosed or you disappear |
| 23:19 |
|
Heffalump |
they are reasonably well backed up, FWIW. If I do disappear without making any contact I'll be dead or something and Igloo would know how to track down my stuff. |
| 23:19 |
|
lispy |
Heffalump: um, how morbid :) |
| 23:19 |
|
Heffalump |
which isn't to say I don't agree with the principle/idea, but I don't think it's too critical :-) |
| 23:20 |
|
Heffalump |
well, I would be :-) I wouldn't voluntarily disappear completely from the darcs project without giving some explanation and handover as appropriate. |
| 23:51 |
|
|
kpreid_ joined #darcs |
