Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2016-03-24

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

All times shown according to UTC.

Time Nick Message
00:11 Heffalump dino-: if you do optimize reorder it'll put the tag before the patches that aren't in it
00:15 sm ah! much more sensible
00:45 dino- Heffalump: excellent
01:18 dino- Also I know that patches having no particular need to be ordered is a feature of the algebra, not a bug. :)
02:09 Igloo joined #darcs
02:32 Riastradh joined #darcs
02:47 ilbot3 joined #darcs
02:47 Topic for #darcs is now http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.10.3 is out http://darcs.net/Releases/2.10
02:50 mizu_no_oto joined #darcs
04:03 Big_G joined #darcs
04:04 Igloo joined #darcs
04:15 mizu_no_oto joined #darcs
04:34 mizu_no_oto joined #darcs
04:54 mizu_no_oto joined #darcs
07:12 pointfree joined #darcs
07:46 alexei joined #darcs
09:55 mal`` joined #darcs
10:32 mal`` joined #darcs
11:24 gal_bolle joined #darcs
12:32 BitPuffin joined #darcs
13:09 Riastradh joined #darcs
13:17 Big_G joined #darcs
13:45 xauth joined #darcs
15:29 kaol joined #darcs
16:10 teratorn_ joined #darcs
16:54 Heffalump dino-: nonetheless the ordering does matter in the UI (and being able to reorder freely is a feature that can help with UI)
17:11 teratorn im not sure if this is anything you guys can fix from a UI perspective, but as a relatively old darcs user (but not so much in recent history i started using it again for dot files) I just blew away some files by accident... I darcs add -r <some dir> and noticed it pulled in a lot more files than i wanted... I wasn't sure what to do, so 'darcs revert' (i looked for unadd, but no), and yeah, blew away untracked files on disk just when all iwanted was to u
17:11 teratorn them
17:11 teratorn 'darcs remove' is the right thing? I guess?
17:12 teratorn I hadn't recorded anything, so I didn't expect it to make any working tree changes
17:32 alexei joined #darcs
18:19 alexei joined #darcs
19:00 alexei joined #darcs
19:10 Riastradh joined #darcs
19:15 pointfree ganesh/ssh tests depend on libssh2-hs and it just so happens that libssh2 apparently does not support the curve25519-sha256@libssh.org kex.
19:25 pointfree However, darcsden does not use libssh2-hs so it appears to get past the kex conversation when I just test with darcsden. Although something is still not done properly in the kex because decryption fails thereafter. My suspicion is the wire format of the data types is not correct.
19:54 pointfree It works! After some more modifications I can push repos to my darcsden instance with curve25519-sha256@libssh.org Huzzah!
19:57 pointfree Now the code just needs to be cleaned up, and the other new changes and factorings that have been going on with haskell ssh need to be integrated. Also need to look into switching to cryptonite.
20:09 pointfree ...except signature verification sometimes fails (but usually doesn't). I think this is because I need add padding to the signature when it is less than the expected length.
21:04 sm teratorn: ouch.. darcs unrevert may have you covered
21:05 sm if even it does, I think that's a pretty nasty UI issue
21:05 teratorn sm: kind of yeah...
21:06 teratorn sm: how can unrevert help? these files were never recorded?
21:06 teratorn (in any case, i recovered them already)
21:06 sm I don't know the answer.. ok good
21:06 sm I've done something similar in git I believe
21:06 teratorn oh well i would *expect* git to have horrible UI edge cases
21:07 sm right. darcs is supposed to be the simple & safe one
21:07 teratorn since it's basically a VCS construction kit not a streamlined experience based on sane defaults and conventions
21:08 sm I think darcs remove is the opposite of darcs add
21:08 teratorn remove just sounded bad at first glance, I mean I don't want to run an 'rm'-sounding command if i don't want files to be deleted
21:09 sm yeah, I'm always doubtful what remove will do. Same with git
21:09 teratorn is there a reason there is no unadd? just because it's covered by something else.. ?
21:09 pointfree untrack
21:10 sm remove was probably thought more intuitive based on other VCSs
21:10 teratorn I think an unadd would be ok even it it's just an alias possibly with different default options for anotehr command
21:11 teratorn it's just the obvious thing.
21:11 teratorn for someone who doesn't have `man darcs' memorized
21:11 sm cvs remove, svn remove, git rm => darcs remove
21:11 dolio I think I brought this up before, and what I was told then was that revert is supposed to require you to ascent to reverting something that adds all the lines in the file.
21:12 dolio Which should give you pause.
21:12 * sm wonders how to list all command aliases darcs understands
21:12 dolio Although I seem to recall it didn't in the case I encountered.
21:13 dolio revert is a command that changes the working directory, though.
21:13 sm pointfree: nice
21:14 dolio I think the oddity is that reverting an add isn't just undoing the add, it's undoing all the changes to the file before it can undo the add.
21:14 sm yes.. I wonder why it was done that way
21:14 teratorn dolio: yeab but if the files were never recoded and hence NOT part of the repo, it shouldn't jst wack them without warning
21:14 teratorn *recorded
21:15 dolio Or rather, if the file is getting added, there's no need to revert the changes before undoing its added status.
21:15 dolio Whereas for a previously added file, it makes more sense to undo the changes.
21:15 dolio teratorn: It did warn you, unless the bug I mentioned wasn't fixed.
21:15 teratorn and one other thing was the first line item reuiqring confirmation in the revert UI was an 'unadd' operation iirc so i assumed it would all be safe and yet 'a'.
21:16 teratorn s/yet/hit/
21:17 dolio Hmm, looks like the bug may not be fixed.
21:17 teratorn cool, glad people care :)
21:17 dolio You say yes to 'unadd' and it automatically decides you said yes to the 'delete all lines' part.
21:18 dolio Whereas you clearly would have said no.
21:18 teratorn yeah I suppose
21:19 dolio At least, I assume you would have been skeptical of that.
21:19 teratorn oh yeah i think so
21:19 teratorn but it's still not completely explicit that unrecorded files will be whacked
21:19 teratorn sure i want to remove these lines from my REPO
21:20 teratorn and yes i know "revert" is fundamentally about changing the working tree, but...
21:20 dolio I think the better logic would be to have reverting an add not do anything to the file.
21:21 pointfree sm, do you think we should plan to keep diffie-hellman-group1-sha1 support, or force everyone to upgrade by removing that line from ~/.ssh/config for the sake of security and sloc? (after I fix the presumptive signature padding issue)
21:21 teratorn since there wasn't an 'unadd' and 'remove' just didn't sound like the right thing at first glance... I reached for 'revert' and the UI confused me, so, not saying it wasn't my fault - just that it was unfortunate.
21:21 dolio But it's also backwards.
21:21 teratorn dolio: I would think so to at first glance.
21:21 dolio At _least_ undoing the add should depend on removing all the lines, and not vice versa.
21:22 teratorn unless you want to add an unadd command and perhaps an explict DELETE THIS FILE? for files which were never recorded?
21:22 dolio So that undoing the add doesn't automatically remove all the lines.
21:22 teratorn adding more commands should be heavily justified - we dont want git.
21:23 dolio I don't think adding commands is what makes git bad.
21:23 teratorn well it's just one minor thing - yeah
21:23 dolio More like they don't add commands, they add flags to existing commands that do entirely different things, but are implemented similarly.
21:24 teratorn hahaha, but there is still a million and one commands
21:24 teratorn and YES there is that too
21:26 sm pointfree: if the ssh libs retires an insecure encryption method, I'm fine with that
21:27 sm I mean, it'll probably be a hassle, but.. what does openssh do ?
21:27 sm gtg, bbl
21:28 dolio Anyhow, I'm not a darcs developer, so you still need to get the attention of someone who can fix this.
21:28 castlelore joined #darcs
21:29 teratorn oh darn I thought you were already writing the patch!
21:29 dolio Nah. I just reported the same thing a while back.
21:31 dolio I guess I didn't bug them enough about it.
21:57 sm dolio: is there an issue ?
21:57 dolio Don't recall.
22:00 castlelore joined #darcs
22:00 kaol joined #darcs
22:02 castlelore joined #darcs
22:34 Heffalump dolio: I think reverting an add has to delete the file, otherwise unrecord+revert wouldn't be the same as obliterate
22:39 dolio Then the decision-making procedure on the patches is backwards.
22:44 dolio I'm not sure why unrecord+revert being different to obliterate would be a bad thing, though.
22:46 Heffalump well, it'd break my intuition :-) Perhaps it's not essential though.
22:49 dolio I guess I haven't encountered this problem again, because I set --look-for-adds.
22:49 dolio So I never add anything.
22:53 dolio I wouldn't be surprised if a lot of people don't automatically expect revert to have any connection to obliterate, though.
22:53 dolio Because they're for two conceptually different things.
22:55 dolio One is for undoing things I've done since the last commit, and one is for eliminating patches.
23:10 Heffalump obliterate removes things from the working copy and from the recorded set; unrecord removes things from the recorded set; revert removes things from the working set.
23:11 Heffalump I see what you mean about people not thinking of it that way, but to me it would feel odd for them not to be connected.
23:39 dolio Okay, well, your intuition deletes the only copy of my file without asking.
23:39 dolio And mine just leaves you with some extra files to manually delete.
23:39 dolio :)
23:40 dolio Another way of putting it is, I think, I think of obliterate as the opposite of pull, largely.
23:48 dolio It's even categorized the same as pull in the help, and revert is not.

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