| <<<Back 1 day (to 2011/11/06) | 2011/11/07 |
Robin_Watts | New graphics card and USB tv stick both delivered this morning. toys! | 13:10.33 |
kens | And its not een Xmas yet.... | 13:33.22 |
Robin_Watts | Windows 7 64bit cannot be told to run untrusted drivers. | 13:37.44 |
| (unsigned drivers, that is). | 13:37.50 |
| You can put it into 'test' mode, and 'test sign' a driver, but that's it. | 13:38.06 |
kens | Prepare to be assimilated... | 13:38.06 |
Robin_Watts | so I can't use my old TV card. | 13:38.14 |
| This xps pattern draws a '+' that's supposed to be 10 pixels across. | 14:09.36 |
Robin_Watts_ | Bah. | 14:13.28 |
Robin_Watts | This xps pattern draws a '+' that's supposed to be 10 pixels across. But at 72dpi, it ends up 3.75 pixels across. So we make the pattern tile 4x4. If I shift the path by (4-3.75)/2 to keep it central, it manages to exactly miss all the pixel centres, and so renders completely blank. | 14:14.43 |
henrys | Robin_Watts:so a bmpcmp after a "lowres" run will just have lowres comparisons? | 14:25.17 |
Robin_Watts | yes. | 14:25.30 |
| bmpcmp will only test the files that differed in the last run. | 14:25.46 |
henrys | thanks | 14:25.57 |
Robin_Watts | hence if you do a bmpcmp after a "pcl lowres", you'll only see the pcl lowres results. | 14:26.08 |
| I want to add a feature to the cluster code where you can set a filter. | 14:26.23 |
| When we produce the 'jobs' list (all the files to test), I want to filter that list according to a given regexp. | 14:26.42 |
| So you should be able to say "only test pbm files" or "only test 72dpi" etc. | 14:27.14 |
henrys | yes that would be nice. | 14:27.27 |
Robin_Watts | lunch. | 14:28.01 |
kens | ping chrisl | 14:56.11 |
Robin_Watts | Anyone else having trouble getting to casper? | 14:57.25 |
chrisl | kens: pong | 14:57.32 |
kens | Haven't tried, just a moment | 14:57.36 |
chrisl | Robin_Watts: I got just in okay | 14:58.03 |
kens | chrisl, where can I set a breakpoint (during rendering) to see what is being executed for a CFF CIDFont ? I want to look at the font interpreter. | 14:58.05 |
chrisl | In our CFF code? | 14:58.26 |
kens | Robin_Watts : I can SCP to Casper | 14:58.31 |
| chrisl, I imaine FreeType, I wnt to watch the CharString being decrypted and executed | 14:58.56 |
chrisl | OKay, give me sec..... | 14:59.23 |
kens | Thanks. | 14:59.32 |
Robin_Watts | chrisl: And now I can. | 14:59.57 |
| Transient problem, clearly. Thanks. | 15:00.06 |
kens | Hmm, I may see the problem with the type1 piece codes (finally). | 15:02.31 |
chrisl | kens: I'd start in cff_slot_load(), and then follow into cff_decoder_parse_charstrings() | 15:02.58 |
kens | Thanks, I think I may see the problem though. | 15:03.40 |
| Can you take a look at gxtype1.c, line 412 | 15:03.58 |
| chrisl ? | 15:05.12 |
chrisl | Just got it up now..... | 15:05.20 |
kens | OK, this is decoding a CharString looking for SEAC glyphs. | 15:05.34 |
| Most of the numebr decoding is done just above the #define 'cnext' | 15:05.52 |
| But (another great design decision) there is a 'special case' operator, #28 | 15:06.13 |
| THis operator is the first byte of a 3-byte number sequence. | 15:06.38 |
| But it looks to me like if that happens we simply skip the number(!) | 15:06.49 |
| We should put it on the stack. | 15:06.58 |
| Because we don't we end up using the wrong (preceding) value for the subr number | 15:07.15 |
| ANd, basically, all hell break sloose. | 15:07.28 |
| Can you tell me if that looks like a problem to you ? | 15:07.42 |
| Of course we don't normally get subr numbers in the thousands.... | 15:07.55 |
chrisl | I'm confused - if it's a number we're parsing, we should handle it before the switch, shouldn't we? | 15:09.53 |
kens | Normally yes. | 15:09.59 |
| But this is a speical case ;-( | 15:10.06 |
chrisl | Oh, right, I see what you mean; yuck! | 15:10.31 |
kens | If you look at the type 2 charstring docs, numbers normally begin (encoded) starting at 32 | 15:10.33 |
| Operators are numbers below 32 | 15:10.40 |
| *BUT* (here's the kludge) there's a special operator called 'shortint', operator number 28, which handles a specific kind of number | 15:11.20 |
| Whoever thought that one up should be shot. | 15:11.31 |
chrisl | And all we're doing is skipping two bytes of the stream, yeh, that looks wrong..... | 15:11.52 |
kens | Exactly what I thought. Thanks, I was beginning too think I was going mad.... | 15:12.07 |
| I think I cna fix this easily now. Thanks to cffdisasm ;-) | 15:12.23 |
chrisl | Obviously a widely used operator...... | 15:12.50 |
kens | :-) | 15:13.03 |
| Its only a problem when its used as the argument to callsubr or callgsubr.... | 15:13.18 |
| Otherwise its a number we don't care about. | 15:13.28 |
| OK, that fixes the reported bug, lets see what happens with a regression test. | 15:17.54 |
| Umm, that's odd, that erverted my changes.... | 15:19.36 |
Robin_Watts | I've just tweaked part of the cluster. If it gives problems for anyone tell me, and I'll revert it. | 15:38.09 |
kens | I guess my regression test will be the first run of your tweaked code then. | 15:38.55 |
Robin_Watts | kens: looks like it. | 15:39.41 |
kens | I'll let you kniow ;-) | 15:40.05 |
Robin_Watts | mvrhel: Morning. | 15:50.34 |
| I looked into that disappearing pattern thing, and I know what's going on. Not sure I know how to solve it though. | 15:51.03 |
| As usual the test file sets up a 10x10 pattern, to draw a '+' sign. That gets scaled to be 3.75x3.75. | 15:51.44 |
kens | Hmm, new filter ghing on cluster I see | 15:51.49 |
Robin_Watts | We make a tile at 4x4. | 15:52.28 |
mvrhel | good morning | 15:52.47 |
Robin_Watts | If I move the path by (4-3.75)/2 so it stays central, it magically missed all the pixel centres, and we don't render anything :( | 15:52.55 |
| kens: I deny everything :) | 15:53.06 |
mvrhel | hmmm. Robin_Watts. It sounds like for low res, we really should be doing AA | 15:53.27 |
Robin_Watts | mvrhel: Yeah. | 15:53.43 |
| We could solve it by turning on fill adjust, but that's going to have knock on effects everywhere. | 15:54.10 |
| kens: Did your job finish ? | 15:54.19 |
kens | Robin_Watts : my cluster test just vanished, as did all yours | 15:54.19 |
Robin_Watts | I aborted mine. | 15:54.26 |
kens | Mine just vanished | 15:54.38 |
| I'll resubmit | 15:54.46 |
mvrhel | yes. I think the solution to this "bug" from the customer is that at the low res one must use AA. At print resolutions this will not be a problem | 15:54.55 |
Robin_Watts | mvrhel: So I should push my fix? | 15:55.16 |
| and note the effects at low resolution. | 15:55.36 |
mvrhel | Robin_Watts: yes I am thinking so. The only thing I wish we had was a better fix for when we do XPS to PDF and that the PDF rendering matched the XPS rendering better | 15:56.07 |
Robin_Watts | mvrhel2: When we do XPS to PDF, the resultant PDF (when rendered) will use the "any part of a pixel" rule. | 15:56.41 |
| Given that xps uses "pixel centres" we are never going to be able to have the two match. | 15:57.14 |
kens | Robin_Watts : resubmitted my test, it hasn't appeared.... | 15:57.17 |
Robin_Watts | but I think we do as best as we can regardless. | 15:57.23 |
mvrhel | ok | 15:57.28 |
kens | ANd I'm not getting any emails about it either | 15:57.30 |
Robin_Watts | kens: OK. gimme a mo. | 15:57.41 |
mvrhel | Robin_Watts: ok then I would do your commit | 15:57.54 |
| I am going to be out for a couple hours this morning | 15:58.13 |
Robin_Watts | ok. | 15:58.31 |
| kens: OK. I've undone my fiddling. | 15:58.40 |
kens | OK will try again | 15:58.52 |
Robin_Watts | kens: Are you submitting via git cluster ? | 15:58.55 |
kens | gitpush.sh | 15:59.03 |
Robin_Watts | right. I've fiddled there too, so if this doesn't work, I'll revert that too. | 15:59.35 |
kens | I haven't pulled a new version of gitpush r in the last few hours | 16:00.02 |
| OK problem | 16:00.11 |
Robin_Watts | kens: gitpush pushes via a secret location with some scripts there. | 16:00.28 |
kens | "Rename from '.git/index/lock' to '.git/index' failed. SHoudl it try again ? (y/n)" | 16:00.42 |
| retry succeeded | 16:01.00 |
| I can see my job in q | 16:01.12 |
Robin_Watts | kens: Your job is done. | 16:32.36 |
kens | Yes, its OK | 16:37.18 |
Robin_Watts | Well, that may be working. | 16:46.45 |
| In theory you should be able to add lines of the form: filter=something,something,something to the end of your clusterpush lines (be they done directly, or done via gitpush). | 16:47.40 |
| git cluster bmpcmp xps filter=xpsfts-a4,ppmraw,72 filter=xpsfts-a4,pgmraw,72 | 16:48.39 |
| That would test only the XPS fts files at 72dpi with either ppmraw or pgmraw. | 16:49.02 |
kens | Sounds good | 16:52.47 |
Robin_Watts | hey marcosw_ | 17:11.31 |
| I've just had my fingers in the cluster code... | 17:11.38 |
marcosw_ | yes? | 17:11.48 |
Robin_Watts | You can now clusterpush gs filter=ppmraw,72 | 17:12.01 |
| and clusterpush filter=ppmraw filter=pgmraw | 17:12.12 |
| etc | 17:12.13 |
| Coo. git logg just listed me an 8 digit hash. Does that mean the 7 digit hash clashed? | 17:14.33 |
kens | Time to go night all | 17:31.05 |
ray_laptop | Robin_Watts: easy enoung to tell I suppose: git logg | grep _____ where _____ is the 6 or 7 digit hash | 17:47.09 |
henrys | is anyone looking at the segfault in the regressions? 692217 test file? | 18:10.34 |
| oh nvm it is reading pdf sorry | 18:12.10 |
Robin_Watts | tor8: Any news on what you're doing in Miami? Is it just you? | 20:21.33 |
| Miles had a cunning plan; arrive, hire car at airport, drive to hotel. The shuttle from the airport is $34 per person, so taxis presumably even more. Got to be cheaper to get a car and drive, and solves a lot of stuff. | 20:22.29 |
| There is a Hertz place 1 block north of the hotel to drop it back at. | 20:22.44 |
mvrhel | Robin: wow I wonder if I should get a car | 20:28.06 |
| my wife just started reading a book call swamplandia which is a fictional book set in the everglades | 20:29.11 |
Robin_Watts | mvrhel: http://www.miami-airport.com/hotel_shuttles.asp | 20:29.41 |
| That makes it sound like $32 for a taxi (plus charges). But the hotel website said $34 for a shuttle, so... | 20:30.17 |
mvrhel | I wonder if kids are $34. That seems really | 20:31.30 |
| high | 20:31.32 |
Robin_Watts | mvrhel: If you're getting a car, let me know, cos it'll mean we don't need to get 2 minivans (minivan + 2 cars may be more flexible depending on what people want to do after the everglades) | 20:31.33 |
mvrhel | I had not really planned to get one but I will look into it | 20:31.58 |
Robin_Watts | That link makes it sound like $19 per person on a shuttle bus. | 20:31.58 |
mvrhel | right | 20:32.15 |
Robin_Watts | I hadn't planned on getting a car either, but given there are 3 of us arriving on the same flight, it doesn't seem stupid. | 20:33.08 |
mvrhel | I wonder if there is a extra drop off fee to return the care to the nearby hertz place | 20:35.32 |
Robin_Watts | $10, miles says. | 20:35.42 |
mvrhel | oh wow | 20:35.46 |
| that is nothing | 20:35.48 |
| that may make the most sense | 20:36.03 |
| I will look into that | 20:36.11 |
| this shuttle/taxi stuff looks a bit expensive | 20:36.29 |
| are you ken chris and helen arriving at the same time? | 20:36.44 |
Robin_Watts | Ken is coming wednesday. | 20:36.54 |
| I believe Chris is on the same flight as us. | 20:37.00 |
mvrhel | ah ok | 20:37.01 |
| and you are coming Tuesday? | 20:37.09 |
Robin_Watts | yeah, | 20:37.12 |
mvrhel | morning? | 20:37.13 |
Robin_Watts | evening. | 20:37.21 |
mvrhel | ok | 20:37.23 |
Robin_Watts | land at 15:55. | 20:37.51 |
mvrhel | ok. I need to get cracking on that patch you gave me | 20:39.06 |
| did you have an update that fixed some warnings? | 20:39.31 |
| if not that is ok | 20:39.36 |
| I need to check but I am thinking we can close the xps bugs | 20:44.39 |
tor8 | Robin_Watts: We are thinking of coming in on wednesday and join your everglades tour, depart on sunday. | 20:44.44 |
| everglades on thursday right? | 20:44.59 |
mvrhel | need to grab some lunch | 20:45.31 |
| I think that is the plan | 20:45.31 |
Robin_Watts | tor8: That's the plan. So 2 of you? | 20:58.36 |
tor8 | yes. the alligators cinched the deal :) | 20:58.59 |
Robin_Watts | mvrhel: Sorry, no update ;) | 20:59.03 |
| tor8: Fab. | 20:59.12 |
mvrhel | you did a lot of work | 21:07.04 |
| Robin_Watts: so is the only piece that is missing is getting the halftone data into the full buffer before the copy_planes | 21:08.29 |
| in gxht_thresh.c | 21:09.40 |
| and basically replace the commented out copy_plane with the appropriate copy_planes? | 21:13.21 |
Robin_Watts | mvrhel: exactly. | 21:16.19 |
mvrhel | ok | 21:17.00 |
| dusting off the neural cobwebs now | 21:17.08 |
| hmm. what I wrote for the color thresholding makes no sense to me. it seems that the same threshold array is going to get applied to all the colorants | 21:42.13 |
| oh | 21:42.37 |
| nevermind I see | 21:42.56 |
| brb | 21:54.18 |
| whew this is trickier than I thought it would be | 22:13.15 |
tor8 | Robin_Watts: bad news. the image zooming is actually in the scaling code :( | 23:30.30 |
| I was getting thrown off by build skew problems in Xcode when tracking it down | 23:30.44 |
| it only linked against the new libraries every other compile :( | 23:30.59 |
| -3s/zooming/zooming bug/ | 23:31.32 |
| Forward 1 day (to 2011/11/08)>>> | |