| <<<Back 1 day (to 2012/06/08) | 2012/06/09 |
Gigs | Robin_Watts: yeah it seemed to be reading the pngs fine now | 00:11.10 |
Robin_Watts | fab. | 00:11.23 |
Gigs | Robin_Watts: you might want to fix those build issues in git | 00:11.24 |
Robin_Watts | what build issues? | 00:11.34 |
| bmpcmp is supplied without a makefile etc. | 00:11.43 |
Gigs | need to mv gs/libpng/scripts/pnglibconf.h.prebuilt gs/libpng/pnglibconf.h | 00:11.46 |
Robin_Watts | right, that's the gs specific pnglib. | 00:12.02 |
| I'm not messing with that. | 00:12.11 |
Gigs | I guess you could update the comment at the top with the build instructions at least | 00:12.42 |
| there's two files that no longer exist that are mentioned in there | 00:13.13 |
Robin_Watts | If we ever do a proper release of bmpcmp with projects/makefiles etc, then certainly. as it is, I'd never personally built bmpcmp with png support until yesterday :) | 00:13.29 |
Gigs | ok, well thanks for your help | 00:13.57 |
Robin_Watts | but you're right, I should fix that comment. | 00:14.10 |
| Do you remember which two files offhand are missing? | 00:14.23 |
| s/missing/surplus/ | 00:14.31 |
Gigs | I don't remember exactly | 00:14.44 |
| something gcc and then the same thing without gcc | 00:14.54 |
| let me see if I can get to that server from here | 00:15.10 |
Robin_Watts | pnggccrd.c ? | 00:15.32 |
Gigs | gs/libpng/pnggccrd.c | 00:16.25 |
Robin_Watts | pngvcrd.c too ? | 00:16.44 |
Gigs | yeah I think those were the two | 00:16.52 |
| I haven't closed the window at work so it's not in the bash history yet | 00:17.10 |
Robin_Watts | Thanks. | 00:17.13 |
| Gigs: Damn. I should have said thank you in the commit message. | 00:18.48 |
| but I forgot, and I've pushed it now. | 00:18.58 |
| sorry. | 00:19.20 |
| bedtime for me. Have a good weekend. | 00:19.28 |
Gigs | heh np | 00:19.33 |
| you too | 00:19.35 |
kens | ping Robin_Watts | 09:01.50 |
Robin_Watts | pong | 09:02.46 |
kens | Hi robin, got a problem trying to do a git bisect | 09:02.56 |
| When I try to bisect between HEAD and ghostpdl-9.05 I get a message about a merege needing tested | 09:03.23 |
| So I go ahead nayway | 09:03.28 |
| Then when I say 'git bisct bad' | 09:03.36 |
| It tells me the problem lies between two commits but goes no further | 09:03.53 |
| (apparently) | 09:03.58 |
Robin_Watts | right, so what commits does the problem lie between? | 09:04.10 |
kens | One sec | 09:04.29 |
| f62ce181e3a68d7f652d9e903c70bd1a67423782 and 63a26299f47bb9931a2b47772c95c38594535041 | 09:05.33 |
| If I try bisecting those, then it tells me : | 09:06.03 |
| Some good revs are not ancestor of the bad rev. | 09:06.03 |
| git bisect cannot work properly in this case. | 09:06.03 |
| Maybe you mistake good and bad revs? | 09:06.03 |
Robin_Watts | 63a26299 is origin/master | 09:06.56 |
kens | Yes | 09:07.05 |
| Its a reasonably recent fix | 09:07.11 |
Robin_Watts | git logg f62ce181 | 09:07.13 |
| sorry, wrong window. | 09:07.23 |
kens | :) | 09:07.28 |
| The problem was fixed between 9.05 and HEAD | 09:07.43 |
| Originally I did git bisect 'bad ghostpdl-9.05' (having previously set good to be the HEAD) | 09:08.20 |
Robin_Watts | $ git name-rev f62ce181 | 09:09.15 |
| f62ce181 tags/ghostscript-9.04rc1~18 | 09:09.17 |
| so the good commit is 18 before 9.04rc1 | 09:09.38 |
kens | THat's not right | 09:09.46 |
| Its definitely post 9.05 | 09:09.52 |
| Let me go back to square one | 09:09.59 |
| I'm currently on HEAD, 'update the exampe build line....' | 09:10.31 |
| So I do 'git bisect start' | 09:10.45 |
| then 'git bisect good' | 09:11.10 |
| then 'git bisect bad ghostpdl-9.05' | 09:11.33 |
| Which takes me to 33f7bcb2 | 09:11.54 |
Robin_Watts | Right. f62ce18 is the branch point where ghostscript-9.04 separates from the master. | 09:11.58 |
kens | So I'm doing a clean and rebuild now. | 09:12.25 |
Robin_Watts | git is definitely not behaving the way I'd expect it to here. | 09:13.43 |
kens | AH, so its not just me, that's something of a relief | 09:13.54 |
| Give me a moment and I'll tell you what happens with this revision & bisect | 09:14.58 |
Robin_Watts | If you can give me the good/bad revisions along the way, that would be good. | 09:15.23 |
kens | Yep it won't take long | 09:15.40 |
| The revision (33f7bcb2) has the comment 'Bug 692704: Don't reset PageSpotColors after every page'. That's the on I'm about to test. | 09:16.39 |
| Oh, forgot to say that after doing 'git bisect bad ghostpdl-9.05' It sadi 'Bisecting a merge base must be tested' before tellng me which commit it was going to | 09:17.49 |
| OK that commmit still renders wrong, so I do 'git bisect bad' | 09:18.26 |
| Which then says: | 09:18.56 |
| The merge base 4e25d12238406245ade09eaeb3dd815933f7bcb2 is bad. | 09:18.56 |
| This means the bug has been fixed between 4e25d12238406245ade09eaeb3dd815933f7bc | 09:18.56 |
| b2 and [63a26299f47bb9931a2b47772c95c38594535041]. | 09:18.56 |
Robin_Watts | 33f7bcb2 is not a valid revision for me. | 09:19.20 |
kens | Full revision is: 4e25d12238406245ade09eaeb3dd815933f7bcb2 | 09:19.51 |
Robin_Watts | kens: The convention is to give the FIRST 7 chars, not the last! | 09:20.56 |
kens | Oh well, I don't use it very often | 09:21.09 |
Robin_Watts | Try: git name-rev ghostpdl-9.05 | 09:22.08 |
kens | ghostpdl-9.05 remotes/origin/gs905 | 09:22.38 |
Robin_Watts | Right, so it's not a revision in your local tree. | 09:22.49 |
kens | Hmm, why not >? | 09:23.10 |
Robin_Watts | And indeed, the ghostpdl tags aren't in the main branch,right? | 09:23.12 |
kens | I have no idea, beats me | 09:23.23 |
| It worked last time when I used 9.00 | 09:23.37 |
Robin_Watts | chrisl makes a new branch and does the branch release on there. | 09:23.41 |
| Yeah, he changed the way he worked at 9.02 or something | 09:23.53 |
| So... | 09:23.55 |
kens | SO I can't use labels any more, oh well | 09:24.03 |
| I guess I need to find a revision close to the release then | 09:24.19 |
Robin_Watts | Yes. I'm just trying to see how to do that. | 09:24.31 |
kens | looks in gs-cvs | 09:24.42 |
Robin_Watts | a2390c3 ? | 09:25.06 |
| I use 'git logg' then press / then type 9.05 RETURN | 09:25.27 |
kens | Was looking for the change of the product string to 9.06 pre release | 09:25.30 |
Robin_Watts | * 5236973 Increment Ghostscript version on master to 9.06. | 09:26.07 |
kens | That'll do, thanks | 09:26.18 |
| Getting the 'not ancestor' message again with 5236973 | 09:27.57 |
Robin_Watts | What's the exact error? | 09:28.52 |
kens | $ git bisect bad 5236973 | 09:29.15 |
| Some good revs are not ancestor of the bad rev. | 09:29.15 |
| git bisect cannot work properly in this case. | 09:29.15 |
| Maybe you mistake good and bad revs? | 09:29.15 |
Robin_Watts | Right. | 09:29.32 |
| git bisect requires 'good' to be the earlier one in the history, and 'bad' to be the later one. | 09:29.56 |
kens | Huh, I wish the help woudl *say* that | 09:30.05 |
| It certainly isn't clear | 09:30.12 |
Robin_Watts | It's hunting for where something breaks - hence it must have worked (been good) in the past, and have failed (been bad) later on. | 09:30.45 |
kens | I would have said it should be hunting for when something *changed* | 09:31.01 |
Robin_Watts | Yes, quite why it can't keep track I don't see. | 09:31.05 |
| kens: It assumes monotonicity in the change though. | 09:31.18 |
kens | It seems to me the software should be capable of inverting the sense if the 'good' is later than the 'bad' | 09:31.27 |
Robin_Watts | indeed. | 09:31.35 |
kens | From a human poitn of view, I know one revision is 'good' and one 'abd' I want to find where that change occured | 09:32.07 |
| OK that looks like its working finally | 09:32.52 |
| OK thanks for the help Robin_Watts I found the relevant commit | 10:40.08 |
Robin_Watts | fab. | 10:42.05 |
kens | Right, I'm off for the weekend, bye all | 11:02.11 |
| Forward 1 day (to 2012/06/10)>>> | |