| Time |
S |
Nick |
Message |
| 03:08 |
|
|
copumpkin joined #darcs |
| 04:25 |
|
|
SamB joined #darcs |
| 08:59 |
|
|
kowey joined #darcs |
| 08:59 |
|
kowey |
good morning! |
| 09:00 |
|
twb |
Is it? |
| 09:00 |
|
lambdabot |
twb: You have 1 new message. '/msg lambdabot @messages' to read it. |
| 09:00 |
|
twb |
I haven't seen any evidence |
| 09:01 |
|
kowey |
are we talking morning or good? |
| 09:01 |
|
twb |
The latter |
| 09:01 |
|
twb |
I am not a morning person. |
| 09:02 |
|
kowey |
nod... apparently non-morning people are somehow more effective than morning people |
| 09:02 |
|
kowey |
or at least, can stay alert for longer than morning people can, or something like that |
| 09:03 |
|
|
lelit joined #darcs |
| 09:03 |
|
twb |
I don't know if I've *ever* been alert. |
| 09:20 |
|
kowey |
twb: I'm pleased to report that your moin 2009-05-30 import patch applies to my rst branch |
| 09:20 |
|
kowey |
*phew* |
| 09:20 |
|
kowey |
my conflict resolution is going to consist of just clobbering by redoing the conversion from clean moin |
| 09:21 |
|
kowey |
this means we likely lose changes made to test.darcs.net |
| 09:21 |
|
twb |
Stinky. |
| 09:21 |
|
kowey |
agreed |
| 09:22 |
|
kowey |
we can retrieve those with darcs pull and diff though, so I'm not too concerned |
| 09:46 |
|
kowey |
mornfall: I think my hashed-storage version of darcs behaves strangely when I do darcs whatsnew some_file_in_repo (it seems to show all changes) |
| 09:46 |
|
kowey |
if you're not around, I'll try to follow up on this later |
| 09:49 |
|
Heffalump |
does redoing the conversion make different patches, then |
| 09:49 |
|
Heffalump |
? |
| 09:50 |
|
kowey |
it makes rather clean patches, actually... which is potentially surprising |
| 09:51 |
|
Heffalump |
what do you mean by rather clean? |
| 09:51 |
|
Heffalump |
I was asking if two successive conversions make the same patch for the same change |
| 09:51 |
|
Heffalump |
(trying to understand what you meant about conflict resolution above) |
| 09:51 |
|
kowey |
OK, let me see if I can unfold what's in my head without tripping over myself |
| 09:52 |
|
kowey |
I've got on the one hand: a repo from 2009-03 PLUS rst conversion patches |
| 09:52 |
|
kowey |
and on the other hand: a repo from 2009-03 PLUS one giant 2009-05 content patch |
| 09:52 |
|
kowey |
so the rst conversion and the giant 2009-05 content patches conflict |
| 09:53 |
|
kowey |
and the conflict resolution is just to re-run the conversion from the giant 2009-05 branch and copy the working dir contents over |
| 09:53 |
|
kowey |
so by rather clean patches, what I meant is that that when I do a darcs record; it's as if I just added some new content in the RST format |
| 09:54 |
|
kowey |
you know, a couple of hunks here and there with new content from May |
| 09:54 |
|
kowey |
and this is potentially surprising if you think that the effect of a conflict is to nullify both the RST conversion and the new moin content |
| 09:55 |
|
kowey |
but I guess it's not SO surprising if you think that changes that touch independent hunks don't interfere with each other |
| 09:55 |
|
kowey |
does that sort of make sense? |
| 10:11 |
|
Heffalump |
(sorry, was elsewhere. /me tries to digest that) |
| 10:11 |
|
Heffalump |
ok, I sort of see. Why not just obliterate the old rst conversion? |
| 10:13 |
|
kowey |
because it's already live |
| 10:14 |
|
Heffalump |
in what sense,that people have pulled from the repo? |
| 10:14 |
|
kowey |
or that they potentially have pulled from the repo |
| 10:14 |
|
kowey |
it's what's been on test.darcs.net for a while now |
| 10:15 |
|
Heffalump |
hmm, true, but I think people just pulled from that for experimentation |
| 10:15 |
|
kowey |
the annoying thing is that I now have some kind of corruption (maybe): my repo has files that darcs whatsnew insists I removed, but I can't re-add, revert, and they're not in pending |
| 10:15 |
|
kowey |
and if I darcs get my repo, it's perfectly clean |
| 10:16 |
|
kowey |
hmm |
| 10:16 |
|
kowey |
it's true that we are still officially in a testing phase |
| 10:19 |
|
kowey |
twb: any preferences? |
| 10:20 |
|
kowey |
I could either give you my incremental conversion patch, or we can start from 2009-05-30 import and re-do the conversion |
| 10:27 |
|
kowey |
@tell gwern HTTP request failed with: Data/FileStore/Utils.hs:131:6-41: Irrefutable pattern failed for pattern (fn : n : res : _) |
| 10:27 |
|
lambdabot |
Consider it noted. |
| 10:28 |
|
kowey |
OK, I'm going to swoop in and redo the conversion from scratch, this time as one big patch |
| 10:32 |
|
twb |
kowey: I abstain from having an opinion on that. |
| 10:41 |
|
kowey |
wiki.darcs.net now up to date |
| 10:53 |
|
Heffalump |
so now people that pulled from it previously might have a patch that shouldn't be in mainline, right? |
| 10:53 |
|
Heffalump |
(which is fine, just checking on what cleanup is needed) |
| 10:54 |
|
kowey |
they might have lots of patches that aren't in mainline, since the last time I did it, I broke the conversion down into a bunch of little patches |
| 10:55 |
|
kowey |
plus the test patches from various folks |
| 10:55 |
|
kowey |
darcs obliterate --from-tag . should do the trick |
| 10:55 |
|
Heffalump |
. ? |
| 10:55 |
|
kowey |
handy trick I learned from #darcs (maybe Max?), since it matches anything, it effectively works as a "from last tag" in this context |
| 10:56 |
|
Heffalump |
nice |
| 10:56 |
|
Heffalump |
umm, does the search page work for you? |
| 10:56 |
|
kowey |
it's empty! |
| 10:56 |
|
Heffalump |
it's empty if you search for anything containing spaces |
| 10:57 |
|
kowey |
oh, I searched for "reordering" and it was empty... not even a no results found |
| 10:57 |
|
kowey |
but searching for "darcs" worked |
| 10:57 |
|
Heffalump |
oh yeah |
| 10:57 |
|
Heffalump |
so it's not spaces |
| 10:57 |
|
kowey |
huh, but searching for "womble" correctly returns "no results" |
| 11:06 |
|
Heffalump |
hmm, interestingly I get a merge blowing up when I pull into the repo I had before |
| 11:06 |
|
* Heffalump |
wonders why |
| 11:07 |
|
kowey |
this is after getting rid of the first batch of RST conversion patches? |
| 11:07 |
|
Heffalump |
nope |
| 11:07 |
|
Heffalump |
I was planning on pulling first then removing stuff |
| 11:08 |
|
Heffalump |
it just doesn't seem like the kind of scenario where a darcs 2 merge should blow up |
| 11:08 |
|
kowey |
could be a good case study for camp |
| 11:08 |
|
kowey |
hmm that gives us 2 real world cases of darcs-2 blowing up :-( |
| 11:10 |
|
kowey |
what could be interesting actually, is to watch it blow up with one RST patch, 2 RST patches, etc (since I had broken them up so) |
| 11:10 |
|
Heffalump |
I'll try that |
| 11:13 |
|
twb |
kowey: Darcs 2 definitely blows up when trying to merge in patches against something "a long way back" in the current repo's history |
| 11:13 |
|
twb |
That is, the further back the ancestor is, the more likely blowups are. |
| 11:14 |
|
Heffalump |
it seems to be blowing up when I pull in the third RST patch. |
| 11:14 |
|
kowey |
do we need to add something to the Conflicts FAQ? |
| 11:15 |
|
kowey |
it'd be good if we could have a useful characterisation of what makes darcs blow up in practice |
| 11:15 |
|
Heffalump |
Wed Apr 15 18:41:20 BST 2009 Eric Kow <E.Y.Kow brighton.ac.uk> * Convert Darcs* pages from MoinMoin syntax to RST. |
| 11:15 |
|
Heffalump |
actually, just that patch on its own causes the blow up. |
| 11:15 |
|
kowey |
e.g. we knew that darcs 1 can blow up in doppelganger situations |
| 11:15 |
|
kowey |
hmm... just that patch + the convert everything patch |
| 11:15 |
|
twb |
I successfully pull all the patches that didn't introduce conflicts, one at a time |
| 11:16 |
|
kowey |
so I'll bet that if we boiled down the convert everything patch to just the Darcs* patches, it should still blow up |
| 11:18 |
|
Heffalump |
oh, actually it did finish in the end, just quite slowly |
| 11:19 |
|
kowey |
small explosion |
| 11:19 |
|
* Heffalump |
reads the patch file |
| 11:21 |
|
Heffalump |
lots of duplicate, and conflictor |
| 11:21 |
|
Heffalump |
but no nested conflictor |
| 11:25 |
|
|
kpreid_ joined #darcs |
| 11:40 |
|
Heffalump |
is darcs-users the right target for wiki patches? |
| 11:40 |
|
Heffalump |
given that people can just edit the wiki directly too without such a notification |
| 11:42 |
|
ronny |
does darcs have a way to group sets of patches to one virtual patch? |
| 11:42 |
|
ronny |
so i can trac kfine-grained changes till a feature is done, then concat it to virtually one |
| 11:44 |
|
kpreid_ |
ronny: out of curiosity, do you know any other vcs with that feature? |
| 11:45 |
|
ronny |
kpreid_: hg as some extension to do that, and i belive bzr has something semilar |
| 11:46 |
|
ronny |
i bleive its helpfull, to have n people collaborate with n patchs on a feature, then view that feature as one change later when the details wont matter |
| 11:46 |
|
ronny |
however only view |
| 11:46 |
|
ronny |
i wouldnt want to loose the details |
| 11:46 |
|
Heffalump |
I agree, but darcs doesn't do it |
| 11:46 |
|
Heffalump |
maybe sometime |
| 11:47 |
|
ronny |
bbl, gotta go for some time |
| 11:48 |
|
eivuokko |
Isn't *implementation* of (checkpoint?) tag like that currently? There's just no interface to say which patches are part tag? |
| 11:48 |
|
Heffalump |
tags just mark dependencies |
| 11:48 |
|
eivuokko |
Yes, and that's what "virtual patch" would do as well? |
| 11:48 |
|
Heffalump |
in theory you could use the diff between two tags to record a composite patch |
| 11:49 |
|
Heffalump |
but it wouldn't compose |
| 11:49 |
|
Heffalump |
a tag marks all the patches in the history when it was made, not a subset |
| 11:49 |
|
eivuokko |
But it is possible to have a patch that doesn't about tag (and depend behind tag), and tag, and combine them? |
| 11:49 |
|
ronny |
bacsially what i want would be a "tag" that gets a patch name and hides its direct dependencies from normal listing of changes |
| 11:50 |
|
eivuokko |
Or atleast was. |
| 11:50 |
|
eivuokko |
Was pretty annoying too, sometimes, near the releases. |
| 11:51 |
|
Heffalump |
eivuokko: I don't understand what you mean |
| 11:52 |
|
eivuokko |
If I have state s of repo, and I make a patch s -> s1. Sametime someone makes tag on s, tag1. It is possible to combine tag1 and s1, is it not? Which means tag could as well just pick a set of changes, instead of all? |
| 11:53 |
|
Heffalump |
but it can't leave out changes earlier in the history if those are depended on |
| 11:53 |
|
eivuokko |
Yeah. |
| 11:54 |
|
eivuokko |
So in interface you'd need to have difference between tag and virtual patch: Tag hides all the old changes, while virtual only hides the ones that are one edge away in graph? |
| 11:55 |
|
eivuokko |
It doesn't seem so bad approximation to me -.- |
| 12:38 |
|
kowey |
we could create something like a darcs-wikidarcs.net address if that would be helpful |
| 12:39 |
|
kowey |
probably a list of committers, people who could spend a few seconds to sanity check (e.g. not spam) |
| 12:41 |
|
kowey |
or actually, just use darcs-devel for that |
| 12:41 |
|
kowey |
but I doubt the traffic generated by the wiki will be high enough to justify this move |
| 12:43 |
|
Heffalump |
but are you proposing for edits straight to the wiki to be reviewed too? |
| 12:44 |
|
kowey |
nah; I'm not even proposing for the emailed changes to be reviewed, just cursory checked |
| 12:44 |
|
kowey |
I figure that we don't want direct application |
| 12:45 |
|
kowey |
because somebody could then submit a patch to (say) remove all the files, which is harder to do on the web |
| 12:45 |
|
kowey |
... not that anybody would do that |
| 12:45 |
|
Heffalump |
fair enough |
| 12:45 |
|
Heffalump |
so I'll just send what I have to darcs-users for now? |
| 12:46 |
|
kowey |
yep... I can go ahead and apply this (while awaiting a policy from gwern/twb) |
| 12:54 |
|
* kowey |
runs darcs optimize to avoid massive wiki bundles |
| 13:04 |
|
|
tux_rocker joined #darcs |
| 13:05 |
|
kowey |
hi tux_rocker, thanks for looking at Darcs.Gorsvet |
| 13:06 |
|
kowey |
(that seemed like a relatively big job, so I'm feeling extra-appreciative) |
| 13:15 |
|
ronny |
re |
| 13:22 |
|
|
balor joined #darcs |
| 13:31 |
|
|
arjanb joined #darcs |
| 14:10 |
|
tux_rocker |
wouldn't it be a good idea to call Darcs.Gorsvet Darcs.HashedStorage? |
| 14:10 |
|
tux_rocker |
the code in it seems to express darcs patch operations in hashed-storage library calls |
| 14:11 |
|
tux_rocker |
thus forming the glue between darcs and hashed-storage |
| 14:11 |
|
tux_rocker |
(or perhaps Darcs.HashedStorageUtils, like ByteStringUtils) |
| 14:18 |
|
kowey |
something like that makes sense |
| 15:07 |
|
mornfall |
kowey: Yes, I think that's broken now. : - ( |
| 15:08 |
|
mornfall |
kowey: Also, with --ignore-times it behaves differently than without. |
| 15:08 |
|
mornfall |
kowey: I'll fix that (more details in my reply to Reinier, when I'm finished with it, anyawy.) |
| 15:08 |
|
mornfall |
anyway* |
| 15:10 |
|
kowey |
OK |
| 15:10 |
|
mornfall |
The whole path restriction business is a little messy. |
| 15:11 |
|
kowey |
I'm still working on my last quarter of the review and hopefully thinking about our plan for the SoC |
| 15:13 |
|
mornfall |
One thing. Can the darcsit thing use a non-proportional font in the big editbox? |
| 15:13 |
|
mornfall |
Editing ReST in proportional font = sucks bigtime. |
| 15:13 |
|
kowey |
gwern: are you around and able to answer that one? |
| 15:13 |
|
gwern |
around, yes |
| 15:14 |
|
gwern |
kowey: tell me more about the request? presumably this was from a search attempt |
| 15:15 |
|
kowey |
the request? you mean mornfall's question or something else? |
| 15:15 |
|
gwern |
the http request |
| 15:15 |
|
kowey |
I still don't follow, I'm afraid |
| 15:15 |
|
gwern |
lb message |
| 15:16 |
|
gwern |
HTTP request failed with: Data/FileStore/Utils.hs:131:6-41: Irrefutable pattern failed for pattern (fn : n : res : _ |
| 15:16 |
|
kowey |
ah yes! hmm senility |
| 15:16 |
|
kowey |
I just noticed it when I did a screen -R on gititdarcs.net |
| 15:17 |
|
kowey |
darcsit seemed to be quite happy and still running despite printing a couple of these |
| 15:17 |
|
gwern |
mornfall: I don't really know what sets the font on gitit. probably the css |
| 15:17 |
|
mornfall |
I guess so. |
| 15:18 |
|
gwern |
are #s comments in css? |
| 15:19 |
|
kowey |
I think they are either for names or classes, and I can never remember which is which |
| 15:19 |
|
eivuokko |
I don't think so. /* */ are comments, # are some sort of selector probably. |
| 15:19 |
|
gwern |
because screen.css has : #editform textarea { height: 25em; width: 98%; font-family: monospace; font-size: 93%; } |
| 15:19 |
|
gwern |
#editform #logMsg { width: 98%; margin-right: 1em; margin-bottom: 0.3em; } |
| 15:19 |
|
gwern |
mornfall: and monospace is nonproportional, no |
| 15:20 |
|
mornfall |
Hm. |
| 15:21 |
|
mornfall |
It doesn't work in Opera. Gitit doesn't work at all in konqueror. |
| 15:21 |
|
|
maltem joined #darcs |
| 15:21 |
|
mornfall |
I wish it'd be a little more browser friendly. : - ( |
| 15:22 |
|
mornfall |
Preview doesn't seem to work, either. |
| 15:22 |
|
mornfall |
It says There is no page '_preview'. You may create the page by clicking here. |
| 15:23 |
|
gwern |
_preview? my memory tells me that there's something odd about that... |
| 15:23 |
|
mornfall |
That's what I get when I click the preview button in the editform. |
| 15:27 |
|
gwern |
in HEAD, it should be ___preview... |
| 15:28 |
|
gwern |
kowey: how old is the gitit on wiki.darcs.net? |
| 15:28 |
|
gwern |
it changed from _preview to ___preview on 19 april |
| 15:29 |
|
kowey |
I don't know (I vaguely remember uploading a gitit-something in Florida) |
| 15:29 |
|
kowey |
but I can update everything and upload a fresh one if that would help |
| 15:29 |
|
gwern |
mornfall: you've hard-refreshed and gotten this preview problem? |
| 15:30 |
|
gwern |
wonder if maybe you have stale js/css |
| 15:31 |
|
kowey |
hmm... I get fatal: read error (Connection reset by peer) when I try to git pull from my gitit dir |
| 15:33 |
|
gwern |
ah, I think I've got it |
| 15:33 |
|
maltem |
Ok folks, I've got a bug report that I think is quite important, but the information isn't very precise, so I'm only telling this here on IRC (seems to be the perfect opportunity, with kowey and gwern around): |
| 15:33 |
|
gwern |
strings tells me that the gitit binary is using ___preview, but when I look in static/ in the js, I see just _preview |
| 15:33 |
|
gwern |
so the css/js in static/ has become outdated |
| 15:33 |
|
gwern |
(this accords with my being able to reproduce it) |
| 15:35 |
|
* kowey |
listens for maltem's bug report with interest |
| 15:39 |
|
|
maltem joined #darcs |
| 15:40 |
|
maltem |
Sorry, got disconnected |
| 15:40 |
|
maltem |
I installed ghc 6.10.3 with extralibs, then compiled darcs |
| 15:40 |
|
maltem |
I got an utterly broken darcs, pulling/getting always crashed (patches not found that were really there, ...). In the meantime I installed and upgraded some cabal packages, then tried compiling darcs again |
| 15:40 |
|
maltem |
Now it works nicely. |
| 15:41 |
|
maltem |
Of course, if I knew what packages caused this, I would have a proper bug report :/ |
| 15:41 |
|
kowey |
ghc 6.10.3 with extralibs without any updates gives you a broken darcs? |
| 15:41 |
|
gwern |
getting pull failed? maybe the autoconf fell back to a no-http-functionality built darcs? |
| 15:42 |
|
kowey |
did you use the cabal build method or autoconf? |
| 15:42 |
|
|
lelit` joined #darcs |
| 15:42 |
|
maltem |
kowey: The ./configure method |
| 15:43 |
|
maltem |
gwern: It did have http functionality, it failed only after some time |
| 15:43 |
|
gwern |
hum |
| 15:43 |
|
maltem |
It's a shame I don't have the old source tree to see the config |
| 15:43 |
|
gwern |
mornfall: ok, preview seems to be working again |
| 15:43 |
|
kowey |
any recollection for the kinds of error messages? |
| 15:45 |
|
maltem |
I fear I don't have any anymore |
| 15:46 |
|
maltem |
I'll see if I'll do a fresh ghc install to reproduce |
| 15:46 |
|
mornfall |
gwern: Thanks! |
| 15:46 |
|
gwern |
kowey: we have any need for DarcsWiki.~1~/ ? |
| 15:46 |
|
gwern |
or DarcsWiki-test.darcs.net/ ? |
| 15:46 |
|
kowey |
gwern: I don't know what that's for |
| 15:47 |
|
kowey |
the latter is the version with my first attempt at RST conversion |
| 15:47 |
|
kowey |
I think we can get rid of that |
| 15:47 |
|
gwern |
ok, then I'll get rid of those two. ~/ is too crowded |
| 15:47 |
|
gwern |
and maybe get rid of one of these git binariezx |
| 15:47 |
|
mornfall |
Hm, tough luck (no preview for me). |
| 15:47 |
|
mornfall |
Maybe it's opera-related? |
| 15:48 |
|
gwern |
mornfall: you hard refreshed? |
| 15:48 |
|
mornfall |
Nah, just hit Edit again. It doesn't work that way? :) |
| 15:48 |
|
kowey |
maltem: that's really generous on your part, if you could |
| 15:48 |
|
gwern |
no |
| 15:49 |
|
mornfall |
Let's try. |
| 15:49 |
|
gwern |
kowey: and kowey-gitit-default.conf is inutile? |
| 15:49 |
|
maltem |
kowey, so far I only found one error message pattern in an IRC log, namely: “Failed to download URL http://...gz: does not exist” (where the URL really does exist) |
| 15:49 |
|
kowey |
gwern: I think so |
| 15:49 |
|
mornfall |
(In konqueror, when the page view does not kill the browser, I get to edit (font is right there) but instead of preview get a blank page with just "mefoo.pw" in it. |
| 15:50 |
|
mornfall |
Oh, font fixed in opera now, too. |
| 15:50 |
|
mornfall |
And preview works. |
| 15:50 |
|
mornfall |
\o/ |
| 15:50 |
|
gwern |
ok, ~/ looks much cleaner now |
| 15:51 |
|
gwern |
kowey: we should probably at some point run gitit for the DWN as well |
| 15:51 |
|
maltem |
kowey, ok I'll try |
| 15:51 |
|
mornfall |
And I suspect `` ... `` doesn't work for multiline bits. |
| 15:51 |
|
gwern |
although since gitit is taking 30% of ram, we might want to hold off on that |
| 15:51 |
|
mornfall |
In ReST (or at least not in pandoc ReST). |
| 15:51 |
|
gwern |
hm |
| 15:52 |
|
gwern |
wish I knew how much of gitit's memory use is due to that ghc bug which won't give ram back to the OS |
| 15:52 |
|
kowey |
gotta run |
| 16:02 |
|
mornfall |
Hm. The pandoc ReST is useless. |
| 16:03 |
|
mornfall |
Can't have verbatim blocks in a list, I guess. |
| 16:03 |
|
mornfall |
If someone feels like it, DeveloperTips is an unreadable jumble from about half on. |
| 16:03 |
|
mornfall |
s/like it/like fixing it/ |
| 16:15 |
|
|
fophillips joined #darcs |
| 17:25 |
|
maltem |
boo, I could not reproduce the bug with a fresh ghc-6.10.3. I guess I had already updated some specific package... |
| 17:26 |
|
maltem |
Anyways, that probably means that the bug originated from some dependency release, not from darcs itself |
| 17:33 |
|
|
balor joined #darcs |
| 17:54 |
|
|
tux_rocker joined #darcs |
| 17:54 |
|
|
olejorgenb joined #darcs |
| 18:05 |
|
Heffalump |
is there some general mechanism to make running the test suite not be influenced by things like ~/.darcs ? |
| 18:14 |
|
tux_rocker |
yes. at least that is the idea. if your testsuite is influenced by your ~/.darcs that's a bug |
| 18:15 |
|
Heffalump |
can't my ~/.darcs do stuff like disabling commands? |
| 18:15 |
|
Heffalump |
or changing options |
| 18:17 |
|
|
zooko joined #darcs |
| 18:17 |
|
zooko |
The front page link with anchor text "Darcs binaries" is a broken link. |
| 18:18 |
|
zooko |
Hm, the whole wiki is gone. |
| 18:18 |
|
zooko |
Oh there it is. |
| 18:18 |
|
zooko |
The link should be changed to http://wiki.darcs.net/Binaries |
| 18:19 |
|
Heffalump |
there should be a redirect from http://wiki.darcs.net/index.html/Foo to http://wiki.darcs.net/Foo |
| 18:19 |
|
zooko |
Who has the ability to update the darcs.net front page? |
| 18:20 |
|
zooko |
I'm trying to get David Abrahams (C++ guru and founder of the Boost library) to use Tahoe. |
| 18:20 |
|
zooko |
He says installing darcs on OpenBSD is "a nightmare". |
| 18:20 |
|
zooko |
I suspect mostly just because the binaries link was broken, or else maybe he didn't see it. |
| 18:20 |
|
zooko |
Because he's struggling to build ghc 6.10.3 on open bsd. |
| 18:20 |
|
zooko |
I sent him the right link just now to the binaries. |
| 18:26 |
|
Heffalump |
kowey does, not sure who else |
| 18:29 |
|
tux_rocker |
don't the OpenBSD guys have a ports system of some kind? |
| 18:30 |
|
gwern |
all the bsds do |
| 18:31 |
|
zooko |
Okay, thanks folks. |
| 18:31 |
|
zooko |
Gotta run get ready for a big party. |
| 18:32 |
|
tux_rocker |
he should be able to build it with http://openports.se/lang/ghc, shouldn't he? |
| 18:34 |
|
zooko |
The darcs wiki says that the openbsd developers recommend against using ports. |
| 18:34 |
|
zooko |
See also a note I just posted to darcs-users about a *different* Tahoe contributor who is trying to build on OpenBSD/sparc64... |
| 18:35 |
|
idnar |
David Abrahams uses OpenBSD? No wonder he's so grumpy ;) |
| 18:35 |
|
zooko |
He's been very polite and helpful so far. |
| 18:35 |
|
zooko |
He's contributed a number of bug reports to the Tahoe project. |
| 18:35 |
|
gwern |
idnar: fear would make anyone grumpy |
| 18:37 |
|
Heffalump |
GHC on OpenBSD on sparc? An obscure language on an obscure OS on an obscure architecture? :-) |
| 18:40 |
|
tux_rocker |
the binaries on the wiki for OpenBSD are pretty old |
| 19:02 |
|
|
kowey joined #darcs |
| 19:04 |
|
* Heffalump |
can help with the redirect but has an annoyed baby to pacify first |
| 19:04 |
|
kowey |
so we've got links from the front page to http://wiki.darcs.net/index.html/Foo and http://wiki.darcs.net/DarcsWiki/Foo |
| 19:05 |
|
kowey |
OK :-) I'll continue working on understanding mornfall's patches |
| 19:06 |
|
kowey |
mornfall: by the way, one of the GADT-enabled modules imports Darcs.Gorsvet (which tux_rocker pointed out could be renamed more helpfully to something like Darcs.HashedStorageUtils), which means that if I apply the invalidateIndex patch, witnesses break |
| 19:12 |
|
Heffalump |
the pacification problem is beyond me and has been passed on to an expert |
| 19:13 |
|
kowey |
linguistic skills will soon be a help, I imagine |
| 19:13 |
|
Heffalump |
RewriteEngine On |
| 19:13 |
|
Heffalump |
RewriteRule ^/index.html/(.*)$ /$1 |
| 19:14 |
|
Heffalump |
that's the right way to get apache to rewrite URLs as wanted |
| 19:14 |
|
Heffalump |
but I'm not 100% sure where to put it - it depends on how the server is set up |
| 19:14 |
|
Heffalump |
is there a config file for wiki.darcs.net in /etc/apache2/sites-enabled ? |
| 19:14 |
|
kowey |
yeah, I'm trying there |
| 19:14 |
|
Heffalump |
(and the obvious other line for DarcsWiki too) |
| 19:14 |
|
kowey |
heh, I'm already using it for RewriteRule ^/(.*) http://darcs.net:5001/$1 [P] |
| 19:15 |
|
Heffalump |
oh, right :-) |
| 19:15 |
|
Heffalump |
ok, in that case short-circuit it |
| 19:15 |
|
Heffalump |
i.e. put the more specific rule first and go direct to the darcs.net:5001 URL |
| 19:15 |
|
Heffalump |
oh, hmm. Perhaps not |
| 19:16 |
|
Heffalump |
what we actually want is a [R] so that the browser gets the right information (i.e. "this thing has moved") |
| 19:16 |
|
kowey |
would this be what we want? RewriteRule ^/index.html/(.*)$ http://darcs.net:5001/$1 [R] |
| 19:17 |
|
kowey |
that's before the [P] rule |
| 19:18 |
|
kowey |
or actually, what I would sort of want is RewriteRule ^/index.html/(.*)$ /$1 [R] to mean "http://wiki.darcs.net/index.html/Foo has moved to http://wiki.darcs.net/Foo" |
| 19:18 |
|
kowey |
and not to http://darcs.net:5001/Foo |
| 19:18 |
|
Heffalump |
right |
| 19:18 |
|
Heffalump |
I think that's the right way to do it |
| 19:19 |
|
Heffalump |
I can't remember whether [R] defaults to permanent or temporary redirects |
| 19:19 |
|
kowey |
yay! that seems to do what we want |
| 19:20 |
|
kowey |
although I don't know if what's going on underneath is really what we want |
| 19:21 |
|
kowey |
ah, when we've committed to using this wiki, we can finally clean up that FAQ |
| 19:27 |
|
Heffalump |
ok, it's returning 302, moved temporarily |
| 19:27 |
|
Heffalump |
I'll go check the syntax for 301 which is what we want |
| 19:27 |
|
Heffalump |
[R=permanent] |
| 19:29 |
|
kowey |
hmm |
| 19:29 |
|
kowey |
so then, the returning 302 should be surprising? |
| 19:30 |
|
Heffalump |
if you've already got R=permanent, then yes |
| 19:30 |
|
kowey |
ok, then I'm surprised :-) |
| 19:30 |
|
Heffalump |
did you reload at the right time? |
| 19:30 |
|
kowey |
what does [P] mean? |
| 19:30 |
|
Heffalump |
[P] is proxy |
| 19:30 |
|
Heffalump |
i.e. keep this URL from the client's point of view, but serve up the page you get from this other URL |
| 19:31 |
|
kowey |
don't show them the 5001 ugliness |
| 19:31 |
|
kowey |
hmm... what do you mean by reloading at the right time? |
| 19:33 |
|
Heffalump |
I just mean after the [R=permanent] went in |
| 19:33 |
|
kowey |
OH |
| 19:34 |
|
kowey |
parsing error... I somehow read the [R=permanent] to mean [R] means permanent |
| 19:34 |
|
Heffalump |
oh :-) |
| 19:34 |
|
kowey |
not so much a parsing error but a general brain malfunction |
| 19:34 |
|
kowey |
ok, so now we should be handing out 302s |
| 19:34 |
|
maltem |
gwern, in re FrontPage on reddit: Yes, I guess so. Or maybe special-case it to print "Darcs", "The Darcs Wiki", something like that |
| 19:35 |
|
Heffalump |
you mean 301, which is what we are doing. |
| 19:35 |
|
kowey |
I mean 301 indeed |
| 19:35 |
|
kowey |
cool, thanks for spotting this and for the help! |
| 19:36 |
|
Heffalump |
zooko spotted it |
| 19:37 |
|
kowey |
@tell zooko thanks for spotting the URL issue with the wiki :-) Heffalump helped to add rewriting thingies |
| 19:37 |
|
lambdabot |
Consider it noted. |
| 19:37 |
|
* Heffalump |
now has a baby in need of further pacifying so is typing one-handed again |
| 19:38 |
|
kowey |
expert has delegated back |
| 19:38 |
|
kowey |
or new pacification needed rather, more likely |
| 19:38 |
|
* Heffalump |
repeats a question from above: is there a standard way to stop a test being influenced by ~/.darcs ? |
| 19:39 |
|
kowey |
err |
| 19:39 |
|
kowey |
by rights, the test harness should be setting HOME |
| 19:39 |
|
kowey |
so that ~/.darcs gets ignored |
| 19:39 |
|
Heffalump |
oh, ruight |
| 19:39 |
|
kowey |
although the test has its own ~/.darcs |
| 19:39 |
|
kowey |
for global things like --ignore-times |
| 19:39 |
|
Heffalump |
so i should just do that myself when running a test manually |
| 19:40 |
|
kowey |
yeah... I wonder if it's as easy as HOME=/tmp/blah tests/foo.sh |
| 19:40 |
|
Heffalump |
it is |
| 19:40 |
|
Heffalump |
ta |
| 19:40 |
|
kowey |
but then why not cabal test tests/foo.sh? |
| 19:40 |
|
Heffalump |
i didn't think of that ;-) |
| 19:40 |
|
kowey |
:-) |
| 19:50 |
|
Igloo |
Doesn't each test source something? If so, why doesn't it set $PATH? |
| 19:51 |
|
|
zooko joined #darcs |
| 19:52 |
|
kowey |
the test harness does add the darcs directory to the PATH |
| 19:52 |
|
Igloo |
Sorry, $HOME |
| 19:54 |
|
kowey |
the test harness does set $HOME |
| 19:55 |
|
kowey |
my use of "should" was because I was concerned that the question meant that somehow it wasn't doing what I normally expect |
| 20:02 |
|
Igloo |
Why do you need to do HOME=/tmp/blah tests/foo.sh then? |
| 20:02 |
|
Igloo |
Or am I just failing to follow what's going on? |
| 20:03 |
|
Heffalump |
the test harness does, the individual test scripts don't |
| 20:03 |
|
Heffalump |
I'm not sure that every test includes lib |
| 20:03 |
|
Heffalump |
or that it makes sense for them to be constrained so heavily |
| 20:03 |
|
|
zooko left #darcs |
| 20:03 |
|
Igloo |
Ah, OK |
| 20:03 |
|
Heffalump |
if they set HOME themselves you couldn't tweak the env yourself |
| 20:07 |
|
gwern |
maltem: don't tell me here, on reddit |
| 20:08 |
|
maltem |
gwern, oo, if you so whish :) |
| 20:40 |
|
dcoutts |
woo! |
| 20:40 |
|
dcoutts |
30% reduction in size of the darcs.cabal file |
| 20:40 |
|
Heffalump |
woo? |
| 20:40 |
|
Heffalump |
nice. |
| 20:42 |
|
* kowey |
arches an eyebrow |
| 20:43 |
|
tux_rocker |
dcoutts: can the smaller file still be read by older versions of Cabal? |
| 20:43 |
|
dcoutts |
tux_rocker: yes but not used |
| 20:43 |
|
Heffalump |
and more importantly the size change is getting rid of the redundancy |
| 20:43 |
|
dcoutts |
.cabal files using this feature are required to specify cabal-version: >=1.8 |
| 20:45 |
|
Heffalump |
when will that be out? |
| 20:46 |
|
mornfall |
kowey: HashedStorageUtils sounds quite silly to me, and moreover it's not really true either. |
| 20:46 |
|
dcoutts |
Heffalump: some time before ghc-6.12 |
| 20:46 |
|
kowey |
OK... btw, in case I mentioned a repo where darcs+hashed-storage was mistakenly thinking that files were removed |
| 20:46 |
|
mornfall |
As for witnesses, I have been asking to have them part of the normal build long time ago. Having them separate makes it pretty easy to break, and it's hard to care, really. |
| 20:47 |
|
kowey |
touching _darcs/index_invalid fixed it for me - I think it's because I had a darcs with the whatsnew patch, but not the invalidate patch |
| 20:47 |
|
mornfall |
Oh, ok, that's a problem indeed. You really want invalidates to happen, otherwise you'll get wonky results. |
| 20:47 |
|
kowey |
is there some mechanism to automatically remember the -ftype-witnesses argument to cabal configure |
| 20:47 |
|
tux_rocker |
mornfall: in what sense is it not really true? |
| 20:48 |
|
tux_rocker |
kowey: shell alias? |
| 20:48 |
|
Heffalump |
what's the argument against making type-witnesses build by default? Speed? |
| 20:48 |
|
mornfall |
tux_rocker: In that it is not utilities for hashed storage... it's not even utilities. It's parts of Repository.Internal rewritten to take advantage of hashed storage. |
| 20:48 |
|
dcoutts |
kowey: there are open tickets for that feature, lemme find it so you can add your vote |
| 20:48 |
|
mornfall |
But Repository.Internal is already an awfully bad module in itself. |
| 20:49 |
|
kowey |
we could enable it by default, I think speed is the only issue |
| 20:49 |
|
kowey |
I was thinking we didn't want a witnesses executable to be installed, but then we could just rename that to something else, so no problem there |
| 20:49 |
|
mornfall |
Heffalump: Well, I more like meant that the production binary should just compile with witnesses enabled. |
| 20:49 |
|
Heffalump |
that doesn't work yet |
| 20:49 |
|
kowey |
that's going to take work, not advocacy |
| 20:50 |
|
mornfall |
Heffalump: That doesn't work because we never applied those patches that make all of darcs compile with those enabled, IIUIC. |
| 20:50 |
|
Heffalump |
and that's because there were significant question marks over their correctness |
| 20:50 |
|
dcoutts |
kowey: ah, now what behaviour do you want from this? |
| 20:50 |
|
kowey |
but enabling it for the building could be useful, although it increases our build time |
| 20:50 |
|
mornfall |
Yeah, which is already over the roof. |
| 20:51 |
|
dcoutts |
kowey: there's a couple things I'm thinking of, one is essentially a local cabal (cabal-install) config file, like the global ~/.cabal/config |
| 20:51 |
|
mornfall |
Unfortunately, I don't have time to fix everything... : - | |
| 20:51 |
|
kowey |
dcoutts: I think I'm thinking of a local config file for a particular directory |
| 20:51 |
|
kowey |
in this particular branch of darcs, I'd like to always configure with witnesses (or whatever) |
| 20:52 |
|
dcoutts |
kowey: ok, so you could keep such a file in a repo but it'd never be allowed in a distributed package |
| 20:52 |
|
kowey |
right |
| 20:52 |
|
dcoutts |
kowey: it's a bit like a local .ghci file |
| 20:53 |
|
dcoutts |
kowey: ok, so that ticket doesn't exist yet, it's only in our heads. Would you mind filing it, give the motivation, use case, restrictions etc. |
| 20:53 |
|
kowey |
on my way! |
| 20:54 |
|
dcoutts |
kowey: note, this would be a patch to cabal-install, runghc Setup would ignore it |
| 20:54 |
|
kowey |
(although I think what I really want is for darcs executable and library to build without double-compilation, which would reduce the time anxiety) |
| 20:54 |
|
kowey |
;-P |
| 20:54 |
|
dcoutts |
kowey: that's nearly done |
| 20:54 |
|
dcoutts |
kowey: I tested that a few min ago, it's not pushed yet |
| 20:54 |
|
kowey |
that's very nice news! |
| 20:55 |
|
dcoutts |
kowey: it'll need a slight reorganisation of the darcs sources however, to move the bits needed for the exe into a separate dir, otherwise it'll just pick up all the local sources rather than getting the modules from the darcs lib |
| 20:55 |
|
dcoutts |
kowey: I tried it just now, moving src/darcs.hs and a couple others into a new main/ dir |
| 20:55 |
|
kowey |
that sounds like it would be good to have anyway |
| 20:55 |
|
dcoutts |
similarly for the unit test prog |
| 20:56 |
|
kowey |
dcoutts; have you seen how camp is divided into a bunch of little packages? |
| 20:56 |
|
dcoutts |
kowey: I've not |
| 20:56 |
|
dcoutts |
however I'm not sure it'll work for the type witness because that seems to use -D cpp flags |
| 20:56 |
|
dcoutts |
that sounds like it cannot go via the library but needs to recompile the source files |
| 20:57 |
|
kowey |
yeah, I guess we can live with darcs lib+exe sharing stuff; and then the witnesses having to be built separately |
| 20:57 |
|
kowey |
2x build time not as bad as 3x build time |
| 20:57 |
|
dcoutts |
does the witness version build everything? or is it just a subset |
| 20:58 |
|
mornfall |
kowey: (Btw., the thing about pending and hunks: these come from darcs remove. |
| 20:58 |
|
kowey |
just a subset (the goal is to keep increasing that subset until it's the whole set) |
| 20:58 |
|
dcoutts |
kowey: I see |
| 20:58 |
|
mornfall |
kowey: Which is quite stupid IMHO, since it breaks darcs move. |
| 20:59 |
|
mornfall |
kowey: A repository with hunks in pending will behave strangely. So we may want to make a case for deprecating "darcs remove" and tell people to move the file out of way if they want to get rid of it...) |
| 21:04 |
|
kowey |
dcoutts: is http://hackage.haskell.org/tra[…]ackage/ticket/556 ok? |
| 21:04 |
|
kowey |
mornfall: pending and hunks? |
| 21:05 |
|
kowey |
it's very scary how I seem to be forgetting things I say only a couple of hours ago |
| 21:06 |
|
kowey |
mornfall: don't forget the process lib issue, by the way... it'd be good if we could get darcs+hashed-storage on GHC 6.8 :-) |
| 21:07 |
|
|
copumpkin joined #darcs |
| 21:07 |
|
dcoutts |
kowey: thanks! |
| 21:08 |
|
mornfall |
kowey: Well, it's a pretty stupid <whatever> bug that it doesn't work. |
| 21:08 |
|
kowey |
you mean the process thing? |
| 21:08 |
|
mornfall |
kowey: I guess a flag is in place, yes. But it will have to wait till 0.3.1 anyway. |
| 21:08 |
|
mornfall |
Yeah. |
| 21:09 |
|
kowey |
OK |
| 21:09 |
|
mornfall |
Upgrading process on 6.8 screws your haskell installation? Wtf? : - ) |
| 21:09 |
|
kowey |
:-) |
| 21:09 |
|
kowey |
I'm still trying to read Darcs.Gorsvet.unrecordedChanges and understand it |
| 21:10 |
|
kowey |
hopefully early this week, I'll have time to think about SoC plans (just letting you know where I am), but things may be a bit crazy for a while :-/ |
| 21:10 |
|
mornfall |
I'll hopefully fix unrecordedChanges tomorrow. :) |
| 21:11 |
|
mornfall |
It's got some stupid rename-related bugginess in it. |
| 21:11 |
|
mornfall |
(Cf. my reply to Reinier's review.) |
| 21:11 |
|
kowey |
hmm, maybe I should start with the thinking about the SoC stuff then |
| 21:11 |
|
kowey |
and then look at the fixed version instead |
| 21:12 |
|
mornfall |
Hm. I could maybe just look at it right away. |
| 21:12 |
|
kowey |
I'll pass on two minor notes while I'm at it: what happens with old-fashioned repos? |
| 21:12 |
|
* mornfall |
looks. |
| 21:12 |
|
mornfall |
They should work just fine (using the index as well). |
| 21:12 |
|
kowey |
and also "should unfold be called something else so that it doesn't sound like Data.List.unfold?" |
| 21:12 |
|
kowey |
(sorry for the silly question) |
| 21:12 |
|
mornfall |
Well, the index is sort of redundant for the old-fashioned repos. |
| 21:13 |
|
mornfall |
But we may want to make a case for having an index for pristine as well, that'd maintain hashes in it. Hard to tell. |
| 21:13 |
|
mornfall |
Probably not worth the hassle. |
| 21:13 |
|
kowey |
right, but does hashed-storage know to read non-hashed pristine? |
| 21:13 |
|
mornfall |
Sure. |
| 21:13 |
|
kowey |
oh ok :-) |
| 21:13 |
|
mornfall |
It knows how to read working, so... |
| 21:13 |
|
mornfall |
Well I mean, it can read plain filesystem trees. |
| 21:14 |
|
mornfall |
So it's trivial to make it read plain pristine, too. |
| 21:14 |
|
mornfall |
(So I have.) |
| 21:18 |
|
kowey |
more randomness: I'll bet fixing the witnesses is just a matter of Repository C(r u t) in Darcs.Gorsvet |
| 21:18 |
|
mornfall |
Yes, sounds like it is. |
| 21:19 |
|
kowey |
(right now, my staging repo is set to refuse to push patches if anything goes wrong, including witnesses breaking, so I'd like to see that fixed before I push :-) ) |
| 21:19 |
|
mornfall |
Tomorrow (I'm on battery power and I want to try sleeping in some 10 minutes anyway). I'm about to get unrecordedChanges fixed. |
| 21:19 |
|
kowey |
same here... I'll continue trying to understand |
| 21:21 |
|
mornfall |
Any chance about suggesting a name for unfold? |
| 21:21 |
|
kowey |
ya know, all my attempts sounded silly, so I gave up |
| 21:21 |
|
kowey |
I was pursuing ideas like "bud" or "blossom" or "bloom" |
| 21:21 |
|
Heffalump |
what's the type? |
| 21:22 |
|
mornfall |
: - ] |
| 21:22 |
|
mornfall |
Rename Stub to Bud and unfold to blossom. Interesting. : - ) |
| 21:23 |
|
mornfall |
I'm fairly sure that beats the Slurp metaphor by almost an order of magnitude. |
| 21:23 |
|
mornfall |
Slurpy, whatever. :) |
| 21:25 |
|
kowey |
:-) OK, bedtime! (Heffalump: Tree -> IO Tree ; I think it replaces hashes with actual trees) |
| 21:25 |
|
mornfall |
Well. It replaces Stub's with actual (Sub)Trees. |
| 21:26 |
|
tux_rocker |
a boring suit-and-tie name could be "expand" |
| 21:26 |
|
tux_rocker |
but i like "blossom" more :-) |
| 21:27 |
|
mornfall |
A Stub is (IO Tree), and is just an explicit way to keep the to-be-(maybe)-executed IO action (like unsafeInterleaveIO, but under explicit control). |
| 21:27 |
|
mornfall |
Well, it's really just deferred IO. |
| 21:27 |
|
Heffalump |
so is it reading, or writing? |
| 21:27 |
|
mornfall |
Nothing unsafe about it (unlike unsafeInterleaveIO which *is* unsafe.) |
| 21:27 |
|
mornfall |
Heffalump: Just reading. |
| 21:28 |
|
Heffalump |
I'd probably use a name involving 'read' or 'input' then, but 'expand' or 'blossom' are fine too. I agree 'unfold' is confusing. |
| 21:28 |
|
Heffalump |
And this could turn into a bikeshed, so that's my last word on the subject :-) |
| 21:29 |
|
mornfall |
:) |
| 21:40 |
|
* gwern |
groans at this whole discussion. you guys! |
| 21:42 |
|
mornfall |
Goodnight. :) |
