Camelia, the Perl 6 bug

IRC log for #cdk, 2010-06-15

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

All times shown according to UTC.

Time Nick Message
03:25 sneumann_ joined #cdk
04:50 sneumann_ joined #cdk
05:30 egonw joined #cdk
05:35 sneumann_ joined #cdk
05:54 CIA-47 joined #cdk
06:05 zarah joined #cdk
06:14 sneumann_ joined #cdk
06:18 egonw joined #cdk
06:59 Gpox joined #cdk
07:45 jpansanel moin
07:45 zarah saluton jpansanel
07:46 egonw hi jpansanel
07:59 jbrefort joined #cdk
08:00 jpansane1 joined #cdk
09:49 sneumann joined #cdk
11:11 maclean joined #cdk
11:13 maclean hi
11:13 zarah hi maclean
11:53 egonw joined #cdk
11:55 maclean hi egonw
11:55 egonw hi maclean
11:56 egonw the cdk-jcp code is giving me big headaches
11:56 egonw really, really difficult to maintain...
11:56 egonw very much duplicate code
11:56 egonw mixed goals...
11:56 egonw maclean: and we just have to find another way to deal with rendering parameters in screencoordinates
11:57 egonw and keep those on the renderer side
11:57 egonw away from the generators...
11:57 egonw and have the renderer translate screen coordinate-based parameters into generator parameters
11:57 maclean well, I'm always interested in alternative architectures - if you have a proposal, I can sketch out an implementation in a branch
11:57 egonw and make sure to redo that if the scale and/or zoom changes
11:58 egonw the code of the generators will become so much easier to read, when it does not have to correct for scale all the time
11:58 egonw and honoustly, I do not get why zoom is not in there too...
11:58 maclean in where?
11:58 egonw which also affects the translation of world-screen coordinates
11:58 egonw in all the renderers....
11:59 egonw if it corrects for scale, it should correct for zoom too
11:59 egonw but they don't
11:59 egonw (and they should do neither)
11:59 egonw anyway...
11:59 maclean yet, amazingly, it does actually draw stuff.
11:59 egonw just trying to fix bugs in the controllerhub.clean()
11:59 egonw blah...
11:59 egonw the code just assumes more than two atoms everywhere...
11:59 egonw causing tons of NPE
11:59 egonw :(
12:00 egonw maclean: yeah, I guess no one is using zoom
12:00 maclean no, no they are.
12:00 egonw and misuses scale for zooming instead
12:00 maclean nope.
12:00 maclean really, I did work on the JCP application with stefan for quite a while.
12:01 maclean and experimented with zoom with rendering to a bare JPanel.
12:01 maclean mom
12:01 egonw yes, the existing uses cases shows it works...
12:02 egonw and the overall result is definately going in the right direction
12:02 egonw but the code is yet again impossible to maintain
12:02 egonw outside whatever you wrote yourself
12:02 maclean damn, I don't have an online example of zoom.
12:04 maclean well the CDK currently has no rendering of molecules at all. Are we going to start again from scratch?
12:05 maclean Because I think then all outside interest in using the CDK to draw molecules with will evaporate.
12:05 egonw hell no
12:05 egonw we start from what we have right now
12:05 egonw which is really pretty damn good
12:05 maclean Right. Good to hear.
12:05 egonw I just wished I would have been stronger on the team about the world coordinates in the past :)
12:05 egonw I let that code slip in
12:05 egonw now I am pissed at myself for letting that happen
12:06 maclean But I am still not clear about how you go about avoiding it.
12:06 maclean I don't know how I would say what size of a gap I want between double bonds (say) if I have to work out what it would mean in world coords
12:06 egonw have the generators only work on world coordinates
12:07 egonw remove BasicSceneGenerator.Scale
12:07 egonw and make all generator parameters have world coordinate dimensions
12:07 maclean Yeah just remove scale completely. That will work.
12:07 egonw e.g. the gap would be 10% of the avg bond length
12:07 maclean And the avg bond length?
12:07 egonw maclean: the scale would be a property of the renderer
12:08 egonw only to be used in:
12:08 egonw 1. the affine transformation
12:08 egonw 2. translating screen coordinates back into world coordinates
12:08 egonw (aka, the reverse affine transformation)
12:08 egonw maclean: yes, the average bond length... with a 1.4 Ångström if not yet defined
12:08 maclean So when I click on the screen to make a double bond, how long is this bond?
12:09 maclean Psychic :)
12:09 egonw ok, 1.4Å would be a bit off for a double bond...
12:09 egonw but that's something I can live with
12:09 egonw close enough
12:09 maclean But these are not molecular models. They are diagrams.
12:09 egonw anyway... 2D drawings are not very exact in that matter anyway
12:10 egonw maclean: MDL molfiles use Ångström scale
12:10 egonw it's not that uncommon
12:10 egonw and there is absolutely no agrument against it...
12:10 egonw the current 40 pixels is equally useless
12:10 egonw I'd say, slightly worse, as it suggest 40 pixels fits all...
12:11 egonw 1.4Å at least matching some reality
12:11 egonw but the point is that 40px means zero
12:11 egonw it works with 800x600, without zooming in, and zero scale
12:11 egonw or whatever
12:11 maclean Heh. I was scanning through all 11,000 commits on the cdk (using gitx) and found a commit comment by Christoph that said :
12:11 egonw sorry... scale=1, of course :)
12:12 maclean "40 pixels looks good on my screen, anyway"
12:12 maclean :)
12:12 egonw the source of all evil :)
12:12 egonw cool :)
12:12 maclean Around 2001.
12:12 egonw can you blog that one?
12:12 egonw with date etc...
12:12 egonw leave out committer if you like, or not...
12:12 egonw this is historic... like "640kB should be enough for anyone"
12:12 maclean Hmmm.
12:13 maclean I'm just reluctant to give up on allowing users to specify screen coords.
12:13 egonw ah, right
12:13 egonw see above
12:13 egonw we certainly do not have to give that up...
12:13 egonw but, the generator code is just not the place to do this
12:13 maclean I know that you have to tweak things to get it to look right on different screens.
12:13 egonw it is the IRenderer<IChemObject> to deal with that
12:14 egonw maclean: not just with different screens
12:14 egonw but also when trying to generate a high-resolution image!
12:14 egonw or SVG
12:14 egonw and you have to adjust font sizes accordingly...
12:14 egonw and ...
12:15 egonw so, we need a reference frame
12:15 egonw and this is the average bond length
12:15 egonw every generator parameter is relative to this
12:15 egonw this is how you draw molecules
12:15 maclean Ok, but what if you have one very, very long bond in your diagram.
12:15 egonw ... on paper too
12:15 maclean Which skews the average bond length.
12:15 egonw indeed...
12:15 egonw that is called outlier detection
12:15 egonw the avg does not have to be the mean...
12:16 egonw it could be the median
12:16 egonw or whatever
12:16 maclean ok, but there is still the risk of sudden jumps in scale.
12:16 maclean this happens with the font, actually.
12:17 maclean As the font manager converts a continuous scale param into a discrete font size.
12:17 egonw we can also opt here to set that 'avg bond length' early on, and not have it change while drawing...
12:17 maclean bootstrapping.
12:17 egonw not recalculating it all the time will also be good for performance
12:17 egonw indeed
12:18 egonw e.g. with 1.4Å
12:18 egonw when starting from scratch
12:18 maclean Hmmm. Surprisingly like the "default_bond_length" field that I had in the renderer....
12:19 egonw quite so
12:19 egonw yeah...
12:19 egonw this is one issue with the current code
12:19 egonw there is no uniform bond length
12:19 egonw some code uses 1.4 (historic code)
12:19 egonw other places uses 40px
12:19 egonw ok, but now for something completely different
12:20 egonw we have to solve this JCP stuff over the summer...
12:20 egonw and live with what we have now...
12:20 maclean Yes.
12:20 maclean The current situation is very difficult to change.
12:20 egonw but I would love to have you on the team to get the generator code rewritten to use world coords only
12:20 egonw indeed
12:20 egonw it would have been simpler if the code was cleaner from the start
12:21 egonw there I failed
12:21 egonw ok, what we should talk about now is the signatures code
12:21 maclean Ok.
12:21 maclean I have a problem with rebasing, I think.
12:21 egonw I'm meeting Lars on Friday, and both of us are keen to get your code (which I understood is now done) in the CDK
12:22 maclean I managed to screw up two of my branches trying to do this.
12:22 egonw I have not attempted rebasing the 146-signatures branch on master yet
12:22 maclean Yes, Lars texted me.
12:22 egonw that is one task (I can do that)
12:22 maclean Ok.
12:22 maclean I looked at the PMD, and fixed some imports.
12:23 egonw ok... unit tests for singature.Orbit, signature.*Signature and the signature.MoleculeFromSignatureBuilder
12:23 egonw those are missing
12:23 maclean Ummmm. I guess I should make some ore tests for o.o.cdk.signature.Orbit class.
12:23 egonw and the JavaDoc need some cleaning: http://pele.farmbio.uu.se/nightly-​signatures/ojdcheck/signature.html
12:23 zarah egonw's link is also http://tinyurl.com/25r3ay4
12:23 egonw I started to have a look at that... but I will not be able to write JavaDoc comments
12:24 egonw but could fix the @githash, @param and @return things, I guess
12:24 maclean Missing @cdk.githash tag.
12:24 egonw I'll do those
12:24 egonw but please fix the messages like: No documentation given for the method makeEdge().
12:24 maclean Sorry. I keep asking questions just after you answer them :)
12:24 egonw that requires some idea on what the code is doing...
12:24 maclean Fine, will do.
12:24 egonw you should do those
12:24 egonw I'll also to the 7 PMD warnings...
12:25 egonw mostly too short var names
12:25 egonw I'll take care of all that
12:25 maclean Too short var names is a lame warning.
12:25 maclean "i" is a perfect counter name.
12:25 egonw then there is one last task for you: update the jar to your latest lib
12:25 egonw yes, that is some collateral damage
12:25 egonw I take that for granted
12:25 maclean Ok.
12:26 maclean I don't think that the signature.jar has changed recently...
12:26 egonw at the same time, it does cover many non-counter variables with useles names
12:26 egonw that's where it pays off
12:26 maclean I still haven't implemented chirality...
12:26 egonw maclean: the version it has now had one problem you were still looking at...
12:26 egonw that was some weeks ago
12:26 egonw maclean: ah, rigth! chirality
12:26 egonw that's an important next step
12:27 maclean Ah left, chirality...
12:27 egonw let's email about that later today
12:27 maclean It is.
12:27 maclean Ok.
12:27 egonw I need to explain how you can get the R,S label
12:27 egonw and we need to see how to plug that in
12:27 egonw that also means, I need to finish the CIP patch and submit that for review
12:28 maclean Yes, and I need to stop worrying about assigning a chirality to a single atom, rather than a group of atoms.
12:28 egonw in what way?
12:28 egonw my code still relies on being told what the chiral atoms are...
12:29 maclean Well, your model (correctly, I think) assigns tetrahedral to 4(5?) atoms.
12:29 egonw but given an atom, it will calculate R,S or 'no clue
12:29 maclean Rather than just the central atom.
12:29 egonw ah, yes, indeed
12:29 egonw 5 atoms
12:29 egonw the central atom, and it's four neighbors
12:29 egonw the CIP code (not in the CDK 1.3.x series) determines R,S based on that info
12:30 egonw by prioritizing the groups based on a subset of the CIP rules
12:30 maclean But the c-signature code just assigns a label to the central atom : [C@]([Br][I][F][Cl])
12:30 egonw which is fine
12:30 egonw once we have R or S, the ligands are no longer important
12:31 maclean But when you read it back in?
12:31 egonw good question...
12:31 maclean Is it always possible to unambiguously assign the right atoms to the chiral group?
12:31 egonw I don't know the signatures format...
12:31 egonw in SMILES, the four ligands are well defined
12:32 egonw but SMILES does not use R,S and CW, CCW instead
12:32 egonw yes, with SMILES it is
12:32 egonw I would say with signatures too...
12:32 egonw that is...
12:32 maclean Hmmm.
12:32 egonw if you can reconstruct the chemical graph
12:32 egonw you know which atoms are the neighbors
12:32 maclean True.
12:32 egonw and then it only matters what the order is...
12:33 egonw btw, slightly OT... read yesterday that Morgan numbers are actually used to define atom equivalence
12:33 maclean Yes, but it doesn't work for 3-regular graphs.
12:33 egonw at a graph level (AFAIK)
12:33 egonw ic
12:33 maclean So cubane, cuneane, dodecahedrane,...
12:34 maclean There are some other examples.
12:34 egonw yeah, saw some reference to that
12:34 egonw ok, to get working on the signatures... just work with the current branch
12:34 egonw and let me work out the rebasing
12:34 maclean Ok.
12:35 egonw ok, got some offline stuff to do now...
12:35 egonw will be back at least tonight
12:35 maclean Ok. Cu.
12:35 egonw I need to go to Stockholm tomorrow, and do not know if I can be online much...
12:36 egonw but I'll keep in touch by email
12:36 maclean Alright.
12:36 maclean I'll try not to spill any more water on my laptop.
12:37 egonw :)
12:38 egonw I just sent you the script I use to calculate R,S starting from SMILES
12:39 maclean Ok.
12:40 maclean Wow. Google ads gets really confused by my mail :
12:40 maclean Surface Topography Tools - www.Polytec-Ltd.co.uk - 3D Non-Contact Surface Measurements Roughness, Texture, Steps and Form
12:40 maclean I guess due to a Jmol mail about 3D orbital topograpgy...
13:45 egonw joined #cdk
14:19 maclean joined #cdk
14:19 carsten joined #cdk
14:36 maclean Huh. Why on earth are the SmilesGeneratorTest and the SmilesParserTest in the o.o.cdk.GUI.smiles package?
14:36 maclean Legacy, I guess.
14:41 egonw hi maclean
14:41 maclean hi
14:41 zarah ni hao maclean
14:41 egonw maclean: ah no...
14:42 egonw those are two alternatives that give GUI feedback
14:42 egonw the pure JUnit versions are where you would expect them
14:42 maclean ah right.
14:43 egonw maclean: quick question...
14:43 maclean ?
14:43 egonw OrbitTest.setUp() must be run before each test
14:43 egonw ?
14:43 maclean mom
14:43 egonw or only before the first test is run?
14:43 maclean first time I've used setUp
14:44 egonw ok, then use @BeforeClass instead of @Before
14:44 maclean So the answer is "maybe".
14:44 egonw then -> if so
14:46 egonw maclean: shall I add the missing @TestMethod annotation for signature.Orbit ?
14:46 maclean Er, sure.
14:47 maclean I'm still trying to get git checkout to work over here.
14:47 maclean fucking git.
14:47 maclean Er..heh. I was referring to the program, of course :)
14:48 egonw sure
14:49 egonw how can I assist?
14:49 maclean hmm not sure.
14:50 egonw what is the current state of the git repos?
14:50 egonw your local one, that is
14:50 maclean well, I pulled from 155-signautures.
14:51 maclean but it is not listed in git branch -r
14:51 maclean which is what I would expect
14:51 maclean and when I say : git checkout -b 155-signatures egon_patches/155-signatures
14:51 maclean it complains.
14:51 maclean oh, where :
14:51 maclean .git remote -v show
14:51 maclean egon_patchesgit://github.com/egonw/cdk.git
14:51 egonw ouch...
14:52 egonw (git pull)--
14:52 maclean git pull egon_patches 155-signatures
14:52 egonw always 'fetch'
14:52 maclean ah oops.
14:52 egonw git pull will pull in patches into your active branch
14:52 egonw that will indeed give quite a mess
14:52 maclean arrgh.
14:54 egonw ok, what is the active branch?
14:54 egonw let's fix things ...
14:54 egonw git reset is your friend now
14:54 egonw e.g.
14:54 egonw if your current branch is master, do:
14:54 egonw git reset origin/master
14:55 maclean it is. git reset --hard?
14:56 egonw not sure
14:56 maclean hmm untracked files. git checkout --
14:56 egonw git clean -fd ?
14:56 egonw make sure there is nothing untracked that you want to keep
14:57 maclean there wasn't.
14:57 egonw good
14:58 maclean stackoverflow tends to be a good source for git advice
14:58 egonw yeah, for many things... I love it for LaTeX advice too
15:03 egonw maclean: permission to do this: line 85-87 (etc) in AtomSignature, replace that with  {@inheritDoc}, as in:
15:03 egonw @Override /** {@inheritDoc} */
15:03 egonw public int[] getConnected(int vertexIndex) {
15:04 egonw and in all classes where appriopriate
15:05 maclean Ok
15:07 maclean yeah, dammit git that's really annoying
15:07 maclean if I do "git fetch egonw 155-signatures" then "egonw/155-signatures" doesn't appear in git branch -r
15:07 egonw make that:
15:08 egonw git fetch egonw
15:08 maclean but if I do "git fetch egonw" then I get all
15:08 egonw ah
15:08 maclean and it does appear.
15:08 egonw well, you're disk is not that small?
15:08 maclean "Git : completely intuitive and consistent repository management!"
15:09 egonw I mean... some 10 extra patches for those 6 branches I have
15:09 maclean Yes, but I don't want all your branches.
15:09 egonw compared to the 11k of the whole repos?
15:09 egonw ah, you don't get them
15:09 egonw not checked out anyway
15:09 maclean I don't want every branch in every repository for everyone I work with.
15:09 egonw just the references
15:09 egonw they will not end up as local branches
15:09 egonw merely as references
15:09 egonw but I get your point
15:09 maclean "Receiving objects: 100% (1401/1401), 9.35 MiB | 392 KiB/s, done."
15:09 egonw is this pollute your list of references
15:10 egonw do I have that much in my branches??
15:10 egonw mmm...
15:10 maclean It's just annoying.
15:10 egonw oh, that must be the rdf patch
15:10 egonw that pulls in the Jena libs
15:10 maclean 73-rdf?
15:10 egonw ah, sorry about that
15:10 egonw right
15:11 maclean It's not your fault. Unless you are actually Linus? I've never met him, so you might be.
15:11 egonw well, that will likely happen at some moment anyway, as this branch is up for review
15:11 egonw no, never had the pleasure of meeting him in person yet
15:11 maclean Oh, no, wait. I watched his git talk.
15:12 egonw masak just gave me this link: http://www.mail-archive.com/dri-devel​@lists.sourceforge.net/msg39091.html
15:12 zarah egonw's link is also http://tinyurl.com/kl6ue7
15:15 egonw I love the quote: "See? All the rules really are pretty simple." :)
15:16 maclean "For true distributed development that supports proper merging, published branches should never be rewritten."
15:16 maclean From the git manual.
15:17 egonw indeed
15:17 egonw which I violate every time I upload a new CDK-JChemPaint patch :)
15:17 egonw but not in this case...
15:18 egonw which is why I used a new branch name
15:18 egonw instead of (re)using 144-signature
15:19 egonw maclean: I fixed (and pushed) all but the first three warnings at: http://pele.farmbio.uu.se/nightl​y-signatures/pmd/signature.html
15:19 zarah egonw's link is also http://tinyurl.com/3xfajs9
15:20 egonw the first three you need to look at... that constructor indeed looks weird with those three parameters simply ignored...
15:20 maclean it does.
15:20 egonw why?
15:21 maclean It does look strange, yes - i agree :)
15:21 egonw why trick people in providing parameter values and then ignoring them?
15:21 maclean BEcause I am evil. Muhahahhaa.
15:22 maclean Alternatively, I made a mistake. If only there was some way to automatically test classes and methods...
15:23 maclean Ah. bollox. You can't call super() then this(). hmmmm.
15:32 maclean Arrgh. Dammit. I just fixed the Orbit integer thing in the patch I just pushed.
15:38 maclean Hmmm. Also : rewriting others code without a good reason is irritating. Where "an automated tool told me to" is not a good reason...
15:40 * egonw feels save hiding behind CDK standards :)
15:40 egonw seriously... this *is* what code review is about...
15:41 egonw if you like to see the PMD tool came up with differently solved, please suggest an alternative
15:41 evilScientist or you could have it done right in the first place :)
15:44 * maclean gets back to unit tests
15:45 maclean egonw: http://github.com/gilleain/cdk/commit/d​c7652785fddb371195451f6108274ff803d8cb0
15:45 zarah maclean's link is also http://tinyurl.com/3ylz37v
15:46 maclean vs: http://github.com/egonw/cdk/commit/c28​28e562b295121427897086c331cd1bcb9e327
15:46 zarah maclean's link is also http://tinyurl.com/32e68or
15:46 maclean sorry :(
15:46 maclean (note Orbit change in both)
15:49 maclean Aargh. That's what you meant by fixed and pushed.
15:50 * maclean is not thinking right today.
15:50 egonw maclean: I do not need to check those links anymore?
15:51 egonw was afk for a few minutes
15:51 maclean you probably won't want to.
15:52 maclean as I've just fixed one of the same PMD errors as you.
15:53 maclean egonw : sorry, my fault. I could revert perhaps.
16:05 egonw maclean: pushed another commit
16:05 maclean ok.
16:05 egonw with the {@inheritDoc}
16:05 egonw that should solve a lot of the OpenJavaDocCheck warnings
16:06 egonw will now do the missing @cdk.githash
16:06 egonw (you know what those are for?)
16:06 maclean Actually, no.
16:06 * egonw is not surprised... it's not particularly obvious :)
16:07 maclean But don't worry - I just assumed they were necessary :)
16:07 egonw it allows our JavaDoc generation to make links to the source in the git repository
16:07 maclean Ah, ok.
16:09 egonw pushed to github
16:47 sneumann joined #cdk
16:52 maclean Orbit now fully tested.
17:09 jbrefort joined #cdk
17:09 sneumann joined #cdk
17:33 sneumann joined #cdk
17:42 egonw maclean: shall I copy in your new patches?
17:42 maclean mom
17:42 egonw I'll use git cherry-pick to keep the history linear
17:45 maclean The history is non-linear?
17:45 maclean I believe you, I'm just not quite sure what you mean
17:46 egonw it has a few parallel tracks :)
17:46 maclean pushed some more tests
17:46 egonw ok, I just pushed five earlier ones
17:47 egonw and will now pull on those new ones
17:47 egonw (using git fetch + get cherry-pick)
17:47 egonw ok, got one new one
17:47 egonw "Molecule From Signature Builder tests "
17:47 maclean yes
17:48 egonw OK, I will rerun Nightly now then
17:48 maclean ok
17:50 sneumann joined #cdk
17:50 egonw ok, bbl
17:51 maclean ok
17:51 egonw in some 15 mins or so
18:14 sneumann joined #cdk
18:15 egonw back
18:16 * maclean is also back
18:18 egonw there was an unused import in one of the tests that made compiling of test-signature fail
18:18 egonw rerunning Nightly now
18:18 maclean unused imports cause failures?!
18:19 maclean oh, right - cross-module problem.
18:30 sneumann joined #cdk
18:33 maclean that's weird. I just fetch/merged the commit that removed the unused import, but it didn't change things. hmmm.
18:33 egonw could be...
18:34 egonw you had a patch to remove imports...
18:34 egonw perhaps it had lost a line?
18:34 egonw it would normally have warned me
18:34 egonw and maybe it did, and did I just not spot it
18:34 egonw Nightly is still running...
18:35 * egonw has to figure out a way to specify externally which modules to test on...
18:40 maclean would be good.
18:44 egonw for the cdk-jcp patch, I have done this by hacking the .xml files...
18:44 egonw but there I can freely rebase...
18:44 maclean rewriting history!
18:44 egonw yes
18:45 egonw beats fixing patches anytime
18:48 maclean hmmm. yes, it seems that my fetch/merge applied your commit that removed the import, then applied another that added it back again.
18:48 maclean I have a feeling that I am doing something wrong here.
18:48 egonw :)
18:49 egonw you could try rebasing on my branch
18:49 egonw but likely not anymore...
18:49 egonw btw, this is why you publish your patch online in patch branches...
18:49 egonw and not in the clone of the upstream branch
18:49 egonw well, not that that is technically any different :)
18:50 maclean Am I not publishing in a patch branch?
18:51 maclean Because I would rather do things right at the start, rather than fix them later.
18:58 egonw not sure what you mean
18:58 egonw OK, new results are out: http://pele.farmbio.uu.se/nightly-signatures/
18:58 zarah egonw's link is also http://tinyurl.com/34krd5c
18:59 maclean Well, as it happens, I am not sure what you mean either :)
18:59 maclean So, I mean this : "Is my git workflow wrong in some way?"
19:00 egonw PMD clean; JavaDoc: 13 left missing; JUnit: 14 left missing
19:00 egonw yes and no
19:00 egonw we did not really decide about a main branch
19:00 egonw so, I kept mine clean
19:00 egonw so, instead of merging in patches, I uses cherry-pick
19:01 maclean Ok.
19:01 egonw you kept up by repeatedly merging in changes from my branch
19:01 egonw nothing at all wrong with that
19:02 maclean Ok ... but...
19:02 maclean (I suspect :)
19:02 egonw no but
19:02 maclean So what was the "no" of "yes and no".
19:02 egonw oh, the no is just practical...
19:03 egonw no, as you have the feeling it doesn't feel exactly right :)
19:03 egonw that's all :)
19:03 maclean Well it's confusing, but then git IS confusing.
19:03 egonw there are different approaches, you have yet to find the one you are most happy with, or so it seems
19:03 egonw you have read 'Git from the bottom up', right?
19:04 maclean I change something, push it to a branch in my cloned repository, you pull from that, I merge from somewhere else.
19:04 maclean It makes things confusing. Doesn't matter how clever git thinks it is.
19:05 maclean I have read various git manuals, yes.
19:05 egonw well, git is not clever...
19:05 egonw but it allows clever things...
19:05 egonw just the ability to cherry-pick patches is just brilliant
19:10 egonw but it also is so powerful that you can easily mess things up...
19:10 maclean That was really my point.
19:10 maclean It is far too easy to mess things up with git.
19:12 maclean And therefore far too difficult to do fairly normal stuff - pull, push.
19:12 maclean And I think that this probably applies to all distributed version control systems.
19:14 egonw actually, given the right commands used, it works well
19:14 egonw git fetch upstream master
19:14 egonw git rebase upstream/master
19:14 egonw git push upstream master
19:14 egonw tune upstream and master
19:14 egonw but there are many ways to work...
19:15 egonw so, what you read in tutorials is often different
19:15 maclean And therefore confusing. It's the perl of the rcs world.
19:16 * egonw quickly checks if masak is around :)
19:16 maclean I thought there was an interactive merge thing.
19:17 maclean No, wait. I wanted this, I think : http://plasmasturm.org/log/530/
19:17 zarah maclean's link is also http://tinyurl.com/2exasux
19:17 maclean patch splitting.
19:21 maclean got to go : dinner.
21:29 egonw joined #cdk
21:56 egonw joined #cdk

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