| Time |
S |
Nick |
Message |
| 02:20 |
|
|
ManateeLazyCat joined #darcs |
| 02:40 |
|
gwern |
hm. darcs record can be pretty slow |
| 02:40 |
|
gwern |
I have one change: ./docs/gwern.xml -48896 +32296; it's been runnign for 80 minutes now |
| 02:41 |
|
gwern |
pegs a core, too |
| 02:42 |
|
twb |
What version of Darcs? |
| 02:42 |
|
gwern |
2.3.1 release |
| 02:47 |
|
twb |
IMO if you're gonna try to add 32kloc in a single commit, you should use a better VCS |
| 02:48 |
|
gwern |
:( |
| 02:49 |
|
gwern |
presumably a single commit compresses better than a bunch of small ones |
| 02:49 |
|
twb |
You have less CPU than disk space? |
| 02:50 |
|
twb |
Er, other way around |
| 02:50 |
|
gwern |
cpu is free; disk space is tight for me |
| 02:50 |
|
twb |
The fact that you have 32kloc in a single file alone will stress darcs |
| 02:51 |
|
twb |
It probably copies at least one version of that file into memory before it even starts the diff |
| 02:52 |
|
gwern |
it's at 1.7% of ram, and constant |
| 02:52 |
|
twb |
Hm. |
| 02:52 |
|
twb |
strace it and find out what it's doing |
| 02:53 |
|
gwern |
2.4 uses .2% more ram |
| 02:53 |
|
gwern |
wonder why |
| 02:55 |
|
|
alexsuraci joined #darcs |
| 03:53 |
|
twb |
I hate how running "darcs record" to add a NEW FILE will show a line being deleted if the file has no trailing newline. |
| 04:08 |
|
|
sm joined #darcs |
| 05:24 |
|
|
alpounet joined #darcs |
| 05:30 |
|
|
gh_ joined #darcs |
| 06:09 |
|
|
kmc joined #darcs |
| 06:34 |
|
|
ManateeLazyCat joined #darcs |
| 07:00 |
|
|
alpounet joined #darcs |
| 07:02 |
|
|
darcscommitbot joined #darcs |
| 07:03 |
|
|
darcswikibot joined #darcs |
| 07:12 |
|
|
Korusef joined #darcs |
| 07:15 |
|
|
ManateeLazyCat joined #darcs |
| 07:16 |
|
kmc |
is there a darcs/git command cross reference somewhere? or a "darcs for git users" document? |
| 07:18 |
|
twb |
wiki.darcs.net/RosettaStone? |
| 07:25 |
|
|
sxs joined #darcs |
| 07:38 |
|
|
gh_ joined #darcs |
| 07:57 |
|
|
lelit joined #darcs |
| 08:03 |
|
|
alpounet joined #darcs |
| 08:37 |
|
|
FunctorSalad joined #darcs |
| 08:37 |
|
FunctorSalad |
is something going to break if you work around the boringfile and add _darcs/prefs/defaults to the repo? |
| 08:40 |
|
twb |
Probably |
| 08:40 |
|
twb |
It'd be better to solve the problem |
| 08:41 |
|
twb |
i.e. get consensus on how to (safely) include defaults in the patch set |
| 08:42 |
|
FunctorSalad |
I just made _darcs/prefs/defaults a symlink to ./darcs-defaults |
| 08:42 |
|
FunctorSalad |
oh, not a public one (at this time) |
| 08:42 |
|
twb |
Suppose I send you a patch that adds "ALL post-hook rm -rf /" and "ALL run-posthook" to that file |
| 08:43 |
|
FunctorSalad |
I see |
| 08:47 |
|
FunctorSalad |
btw I noticed one can work around symlinks not being supported with unionfs and the like... not sure what the consequences are |
| 08:47 |
|
FunctorSalad |
bind-mounting a superdir as a subdir and adding that seemed to work fine |
| 08:48 |
|
FunctorSalad |
(as long as you don't do recursive stuff on it) |
| 09:10 |
|
|
florent joined #darcs |
| 09:10 |
|
twb |
Whatever |
| 09:11 |
|
florent |
hi all |
| 09:11 |
|
|
alpounet joined #darcs |
| 09:12 |
|
exlevan |
hi |
| 09:21 |
|
|
gal_bolle joined #darcs |
| 09:32 |
|
|
kowey joined #darcs |
| 09:32 |
|
kowey |
morning :-) |
| 09:33 |
|
* exlevan |
waves |
| 09:33 |
|
|
alpounet joined #darcs |
| 09:38 |
|
kowey |
exlevan: I think to finish up your proposal, a couple more details about where you've used Haskell may be nice |
| 09:38 |
|
kowey |
there are also some questions on http://hackage.haskell.org/tra[…]iki/StudApply2010 that may be worth thinking about |
| 09:41 |
|
kowey |
I would also consider pushing the transfer-mode fusion to the very end of your project |
| 09:45 |
|
exlevan |
does unfinished application count? I've got semi-finished haskell->php compiler using GHC API, but it hardly counts as proof of my ability of getting things done :-( |
| 09:49 |
|
kowey |
sure! it's just to get an idea of what kind of work you've done |
| 09:49 |
|
kowey |
and I think it's a very nice example even if unfinished |
| 09:50 |
|
exlevan |
kowey: well, transfer-mode fusion is already the last feature I implement, and pushing it after "stop coding, continue debugging" phase doesn't look like a good idea for me |
| 09:51 |
|
kowey |
you may be right about that (and also it may change the documentation) |
| 09:52 |
|
kowey |
one trivial point is that the smart server should probably just be darcs transfer-mode --http |
| 09:52 |
|
kowey |
(and also outside of the application, you may also want to think about how people will actually run this thing) |
| 09:53 |
|
kowey |
(eg. are they expected to fiddle with their current webserver config to do some sort proxying? anyway, that's probably irrelevant) |
| 09:55 |
|
kowey |
don't feel like you have to rename 'darcs serve' to 'darcs transfer-mode --http' in your application by the way; it may make the story more confusing... I was just pointing it out for if you actually started the work |
| 10:01 |
|
exlevan |
in context of darcs serve, darcs serve --local seems to better name for transfer-mode :-) and darcs transfer-mode --http sounds confusing... it's serve in Mercurial and SVN, btw |
| 10:02 |
|
kowey |
hmm! I never thought of it that way before |
| 10:02 |
|
kowey |
ok, and another point (not really for the application)... I wonder how this would fit into the caching scheme? |
| 10:03 |
|
kowey |
if you darcs get the GHC repo a second time (this happens surprisingly often), it's actually fast because you're using the global cache |
| 10:04 |
|
kowey |
I suppose one evil thing we could do is try both simultaneously; if copying from global cache finishes first, we abort the transfer |
| 10:06 |
|
exlevan |
I didn't think of cache before, but that trick sounds cool (unless cache and server connection using the same line) |
| 10:07 |
|
kowey |
if we don't do this, the worst that will happen is that the second darcs get will be slower than before... but fast enough to be bearable still (and predictable) |
| 10:08 |
|
kowey |
so it's not so important to worry about it, I suppose. The smart server makes the global cache unnecessary, but we may still need the global cache to deal with (i) no smart server (ii) situations like copying across fs boundaries |
| 10:10 |
|
kowey |
anyway, I'd better get back to work. exciting! three interesting applications this year... |
| 10:32 |
|
gh_ |
http://blog.prelode.com/2010/0[…]wmicrobenchmarks/ ← a simple (vaguely relevant?) benchmark involving darcs, git, svn |
| 10:33 |
|
gh_ |
i've reproduced it, what i get is: |
| 10:33 |
|
gh_ |
2.3.1 cold cache 56s |
| 10:33 |
|
gh_ |
2.3.1 again 52s |
| 10:33 |
|
gh_ |
2.4 + 162 patches cold cache 25s |
| 10:33 |
|
gh_ |
2.4 + 162 patches again 25s |
| 10:33 |
|
gh_ |
(user time) |
| 10:35 |
|
gh_ |
2.3.1 is the static from wiki, compiled with ghc 6.10, 2.4 is compiled by myself with ghc 6.12 |
| 12:24 |
|
|
alpounet joined #darcs |
| 12:45 |
|
|
kpreid_ joined #darcs |
| 12:49 |
|
Purdy |
arg - my repo upgrade is taking forever and locking up my machine :( |
| 12:49 |
|
Purdy |
i left it overnight and it's only halfway done this morning |
| 12:50 |
|
Purdy |
i had to kill it (prolly corrupted the repo in the process) ... will start it before i leave for the weekend |
| 12:50 |
|
Purdy |
not keen on doing that on our production server ... |
| 12:51 |
|
Purdy |
can i just tarball the _darcs and upload it to production? |
| 13:09 |
|
FunctorSalad |
btw is there some combination of commands to make sure that a given file equals the version in the remote repo? or are you supposed to darcs get a fresh repo copy and cp the file over? |
| 13:10 |
|
FunctorSalad |
assuming you have local-only patches in the history, so it isn't just a matter of "if the patch histories are equal, the results should be equal too" |
| 13:14 |
|
|
exlevan joined #darcs |
| 13:44 |
|
|
drk-sd joined #darcs |
| 13:56 |
|
|
kpreid_ joined #darcs |
| 14:12 |
|
|
ManateeLazyCat joined #darcs |
| 14:26 |
|
|
alexsuraci joined #darcs |
| 14:41 |
|
|
alpounet left #darcs |
| 14:59 |
|
mornfall |
Purdy: Yes, you can tarball. |
| 15:04 |
|
mornfall |
Purdy: How much memory do you have? Maybe you could upload your repo somewhere and I can convert it on one of our big machines? |
| 15:21 |
|
|
idnar joined #darcs |
| 15:29 |
|
|
int-e joined #darcs |
| 15:30 |
|
int-e |
hehe. "Writing inventory 2 done, -1 queued. 0000001029-77d67730713375f0c8e99e958f08c6a0" (using darcs 2.4 to fetch http://darcs.haskell.org/ghc-6.12/ghc-tarballs) -- the fetch seems to be successful though. |
| 15:35 |
|
|
sxs left #darcs |
| 15:39 |
|
|
zooko joined #darcs |
| 16:00 |
|
|
zooko` joined #darcs |
| 16:01 |
|
Purdy |
mornfall: i appreciate it, but i'm wary of handing out our repo ;) ... i have 2 gb of ram and darcs was running 700-800 MB |
| 16:02 |
|
Purdy |
i suppose it could have used more if i had more, tho ... my mem consumption was in the 95% range |
| 16:02 |
|
Purdy |
while it was running |
| 16:02 |
|
mornfall |
That shouldn't be locking up though. |
| 16:02 |
|
Purdy |
it's odd b/c i have a dual core and my cpu is definitely over 90% and the rest of my system is slooow |
| 16:03 |
|
Purdy |
you'd think darcs would just monopolize one core |
| 16:03 |
|
mornfall |
And are you sure it's only 700-800MB? Maybe the system went into swapping... or something else is using a lot of memory at the same time. |
| 16:04 |
|
mornfall |
Also, darcs does a lot of IO during conversion, so it may be that other things are waiting for IO to get through. |
| 16:04 |
|
mornfall |
Anyhow, gotta run, see you later. :) |
| 16:04 |
|
mornfall |
Good luck converting. |
| 16:04 |
|
Purdy |
thanks :) |
| 16:09 |
|
lispy |
Good morning |
| 16:11 |
|
Purdy |
gargh ... this isn't what you want to see when doing a push: 12284 jason 18 0 790m 625m 6104 D 31 17.6 0:17.68 darcs |
| 16:11 |
|
Purdy |
just a simple patch w/ 2 text files |
| 16:11 |
|
Purdy |
that's on the remote production server |
| 16:12 |
|
Purdy |
i think the bottom line is that my repo has outgrown darcs (1.0?) |
| 16:12 |
|
Purdy |
and i can't get to 2.4 quick enough ;) |
| 16:22 |
|
|
alexsuraci joined #darcs |
| 17:00 |
|
* sm |
wonders how Purdy can count conflicts in his repo |
| 17:33 |
|
|
abuiles joined #darcs |
| 17:41 |
|
Purdy |
wha? |
| 17:42 |
|
|
kowey joined #darcs |
| 17:42 |
|
kowey |
good evening |
| 17:44 |
|
kowey |
abuiles, drk-sd, exlevan : have you all submitted your applications? |
| 17:44 |
|
abuiles |
kowey: I'm on that right now.. |
| 17:44 |
|
drk-sd |
yep, just did it |
| 17:45 |
|
drk-sd |
(damn, 500c for the abstract, that's really short) |
| 17:52 |
|
exlevan |
yep, submitted |
| 18:25 |
|
abuiles |
submitted |
| 18:36 |
|
|
abuiles left #darcs |
| 18:47 |
|
|
JaffaCake joined #darcs |
| 18:54 |
|
|
zooko` joined #darcs |
| 19:08 |
|
|
ManateeLazyCat joined #darcs |
| 20:24 |
|
|
zooko` joined #darcs |
| 20:52 |
|
|
exlevan joined #darcs |
| 22:31 |
|
|
dino- joined #darcs |
| 22:53 |
|
|
FunctorSalad_ joined #darcs |
| 23:11 |
|
|
nwf left #darcs |
