| <<<Back 1 day (to 2011/10/16) | 2011/10/17 |
kens | ping chrisl | 09:10.02 |
chrisl | kens: yo! | 09:19.37 |
kens | chrisl do you have a 64-bit version of Windows installed somewhere ? | 09:32.30 |
chrisl | Yes, I've it in a VM and on my laptop | 09:33.04 |
kens | Could you run a quick test for me ? | 09:33.13 |
chrisl | Sure | 09:33.17 |
kens | Bug #692589 | 09:33.33 |
| The user is running on 64-bit and I can't test it that way | 09:33.49 |
| I don't believe it should make a difference, but.... | 09:34.01 |
| If you could run pdfwrite in PDFA mode on the file and send ther esult to me I would be grateful | 09:34.24 |
chrisl | Should I test with 9.04 or master? | 09:34.32 |
kens | 9.04 please | 09:34.39 |
| Though master shuld be the same | 09:34.50 |
| I want in part to look at the file structure, I'm sure he hasn't used a vanilla 9.04 and he definitely hasn't used the pdfa_def.ps file he says he did | 09:35.18 |
chrisl | Okay, I'll need to build either, so it'll take a few minutes | 09:35.18 |
kens | Ah, I pulled and installed the binary from the downloads site just to be sure ;-) | 09:35.40 |
| But there's no rush at all | 09:35.51 |
chrisl | Okay, I'll get the installer - I need to download that...... | 09:36.25 |
kens | Probably quicker than building :-) | 09:36.38 |
| going to fetch a coffee, brb | 09:36.48 |
chrisl | kens: the PDFA_def.ps we distribute doesn't work, because the ICC profile doesn't exist | 09:45.33 |
kens | You need the one from his bug report | 09:47.29 |
| Which is actually (along with the command line) in bug #692587 | 09:48.01 |
chrisl | Yeh, I just noticed that :-( | 09:48.18 |
kens | :-) I probably shouold have mentioned it | 09:48.31 |
chrisl | Well, I would suggest the reporter should have to put all the required info in the bug report.................... | 09:49.10 |
kens | He's not making any friedns so far... | 09:49.29 |
| Or friends for that matter | 09:50.28 |
chrisl | Am I expected an error? | 09:52.05 |
kens | No | 09:52.12 |
| THe file allegedly doesn't pas PDF/A pre-flight | 09:52.26 |
| His doesn't, mine does | 09:52.31 |
| Did you get an error ? | 09:52.56 |
chrisl | No, but the bug report is an error, isn't it? | 09:53.31 |
kens | I don't think so, just a pre-flight failuer | 09:53.49 |
chrisl | So, that title is not our error, then] | 09:54.16 |
kens | No, that's the pre-flight report from Acrobat | 09:54.35 |
| The title is misleading | 09:55.00 |
chrisl | Winging its way yourwards now | 09:55.04 |
kens | Thanks very much | 09:55.12 |
| My next PC will have a 64-bit OS | 09:55.29 |
chrisl | FWIW, AcroPro 9.x PreFlight lists no problems with PDF/A-1b compliance | 09:55.59 |
kens | And (surprise!) your file passes preflight too | 09:55.59 |
| stereo | 09:56.05 |
| So however he made or modified the file, we didn't break it | 09:56.23 |
chrisl | It would seem so..... | 09:56.42 |
kens | THat's using 32 and 64-bit OS's with his pdfa_def (except we know its not) and his command line | 09:56.46 |
| I'll update the thread, thanks for that | 09:57.02 |
Robin_Watts | kens: My PC will have a 64bit OS by next weekend. | 09:57.29 |
chrisl | NP. | 09:58.05 |
kens | Great, two of you to bug :-) | 09:58.18 |
chrisl | I think I'll go fetch a caffeine fix now....... | 09:59.08 |
kens | drinking mine | 09:59.14 |
Robin_Watts | I'm planning to set up win7 on the new SSD, and use my existing harddrive for bulk storage. That means I should be able to just swap the boot device and go back into 32bit win xp if I need to. | 09:59.58 |
kens | Sounds like a plan. This machine is 4 years old now so I'm thinking its getting towards time to upgrade (I *hate* upgrading...) | 10:00.37 |
Robin_Watts | Me too. | 10:00.51 |
| I bought new RAM for this machine on friday, and hunted around looking for the best possible match to what I need. | 10:01.29 |
| This motherboard is supposed to do PC2-8500, but that's a very awkward standard. | 10:02.12 |
kens | I didn't realise that | 10:02.29 |
| TBH hardware upgrades I can face, its shifting my installed environment I dont like. | 10:02.48 |
Robin_Watts | I found some RAM that was supposed to do PC2-6400 by default, but be rated to PC2-8500 if you fiddle it in the BIOS, so I ordered 4 Gig of it, intending to buy another 4 gig if I could make it work at the faster speed. | 10:03.09 |
kens | Did it work ? | 10:03.25 |
Robin_Watts | And when I opened my machine I found I had *exactly* the same RAM in anyway. | 10:03.27 |
kens | :-) | 10:03.36 |
Robin_Watts | I can't make it work at the 1066 FSB speed, no. | 10:03.40 |
kens | Hmm, not so good | 10:03.53 |
Robin_Watts | But it runs at 800 at 5-5-5-15 rather than the more usual 5-5-5-18. | 10:03.55 |
kens | THat's certainly odd | 10:04.13 |
Robin_Watts | Nah, like I say, it's a very awkward standard. | 10:04.27 |
kens | I lost track of RAM standards some years ago | 10:04.53 |
Robin_Watts | tor8: If you don't have a harddrive in your machine, where do you put your swap file ? | 10:05.06 |
tor8 | what swap file? 8gb of ram ;) | 10:05.24 |
Robin_Watts | Really? You disabled swap ? | 10:05.36 |
tor8 | I haven't set a swap partition in linux for over 10 years, and I disable the windows one whenever I remember | 10:06.56 |
| it's felt pointless once I started having >= 1gb of ram in my machines | 10:07.34 |
| I'd rather crash than thrash on the disk swap | 10:07.47 |
Robin_Watts | Apparently having a swap partition on an SSD is a good way to kill an SSD. | 10:07.48 |
| With a 4Meg machine (albeit effectively a 3gig one) I swap lots of the time :( | 10:08.12 |
| s/meg/gig. | 10:08.27 |
kens | smeg ? | 10:08.47 |
| :-) | 10:09.03 |
tor8 | I can't remember the last time I had a machine swap... must've been back in the 90's ;) | 10:09.27 |
| do you perchance run firefox? | 10:09.38 |
Robin_Watts | I do. | 10:09.46 |
chrisl | Robin_Watts: it shouldn't be too bad on modern SSDs with sector scattering, or whatever it's called. | 10:09.47 |
Robin_Watts | chrisl: wear balancing? | 10:09.59 |
chrisl | that's the thingy, yes | 10:10.14 |
Robin_Watts | chrisl: Yes, but even so, I think I'll avoid it. | 10:10.28 |
tor8 | you'll probably replace your SSD before it wears out | 10:10.31 |
Robin_Watts | tor8: I really hope not. | 10:10.40 |
tor8 | even so, I'd recommend turning off pointless writes, like swap and 'atime' if you're on linux | 10:11.04 |
Robin_Watts | or rather, I hope it'll last long enough that the machine is too slow to bother replacing components from :) | 10:11.07 |
tor8 | same same ;) | 10:11.15 |
chrisl | Robin_Watts: yep, I would avoid it, too, remembering what MTBF actually means..... | 10:11.20 |
tor8 | just make sure to back up important files regularly. I've heard the horror stories. also, if they fail they usually stop being writable but you can still read your data. or so I've heard. | 10:12.17 |
chrisl | tor8: at least on systems with conventional HDD, it's usually recommended to have at least a small swap partition, regardless of physical memory capacity | 10:12.34 |
tor8 | chrisl: question is why though | 10:12.51 |
| I know the debian installer likes nagging about having a swap partition | 10:13.09 |
Robin_Watts | Windows tends to swap out in advance of need though. | 10:13.26 |
tor8 | but following the 15 year old guidelines and setting up 16gb of swap for an 8gb machine, well, I don't see the point | 10:13.38 |
chrisl | tor8: usually it's said it's because there stuff that gets used at boot and shutdown, and it's pointless having that it ram all the time. Also, if you want to hibernate....... | 10:14.12 |
Robin_Watts | If you have a process that you haven't used for a while, windows will swap out during idle times, so that sudden surges of memory use can be accomodated. | 10:14.20 |
tor8 | chrisl: right. hibernation is a diff issue altogether though :) | 10:14.40 |
Robin_Watts | Which sounds really smart, BUT, I don't believe they cope with having the same page on disc and in memory at once. | 10:14.48 |
| So all it means is that process has to be read back in when you swap back to it. | 10:15.06 |
| rather than meaning you can 'instantly' dump the memory page. | 10:15.18 |
chrisl | tor8: I haven't use hibernation for quite a while - do you know if it still uses the swap space? | 10:15.33 |
tor8 | well I just checked my windows (recently reinstalled) and it had set up an 8gb partition file by default | 10:16.27 |
| so that would be my guess | 10:16.41 |
| but I really don't know | 10:16.47 |
chrisl | I seem to remember dire warnings about Windows having stability problems if it didn't have a page file available - I really hope that doesn't hold any more! | 10:17.30 |
tor8 | I just turned mine off again, I'll let you know if I get any crashes :) | 10:18.04 |
kens | On an OS you don't use ? :-) | 10:18.22 |
tor8 | I have run with no page file before with no ill effects. I believe both in XP and 7. | 10:18.26 |
chrisl | Well, I suspect when I read those warnings it at NT4.0 at the newest. | 10:19.19 |
| kens: I read "Rule 34" last week - that's, um, a weird book......... | 10:21.13 |
tor8 | chrisl: rule 34 ... of the internet? | 10:21.59 |
chrisl | tor8: the Charles Stross book | 10:22.17 |
tor8 | well, then I can only assume without having read the book that he does indeed refer to rule 34 of the internet. you perv! ;) | 10:22.46 |
Robin_Watts | I saw 'Monsters' last night. I recommend it. | 10:22.50 |
chrisl | ;-) but it's a less than oblique reference to that, yes | 10:23.04 |
kens | chrisl its related to halting state | 10:23.54 |
| Some of Stross' work is a bit obscure (accelerando for instance) | 10:24.37 |
chrisl | Yeh, with Rule 34, it was the style, rather than the story - second person, present tense would usually be a quick exit for me! | 10:25.22 |
| Whereas Accerlando I felt was just a bit rushed...... | 10:25.53 |
kens | Hmm, well it (Accelerando) was based on a short story, so its not exactly rushed.... | 10:26.24 |
chrisl | It covered a *lot* of ground, in not a very long book | 10:26.56 |
kens | Yes, the short story was better I think (not that it was very short) | 10:27.26 |
chrisl | I did enjoy Rule 34 - I'm not sure if my recollection of Edinburgh geography helped or hindered the enjoyment! | 10:29.16 |
kens | Well its nice to see places you know mentioned (other than London) | 10:29.37 |
chrisl | I wasn't so sure about the use of the Edinburgh vernacular in some places - it occasionally seemed a bit forced | 10:30.27 |
kens | Well, he is English. | 10:30.39 |
chrisl | The thing is, the dialogue was okay, it was some of the other times he tried it jarred a little | 10:31.35 |
kens | Well, my knowledge of the slang is more than slightly out of date. | 10:32.08 |
chrisl | I guess mine probably is, too | 10:32.39 |
kens | :-) | 10:32.45 |
chrisl | Still know what a "schemie" is, though....... | 10:33.06 |
kens | Umm, I htink I used to. | 10:33.18 |
| Oh yes, housing estate (or scheme) | 10:33.51 |
chrisl | Yes, a "schemie" is a less than savoury character hailing from one of said council housing estates/schemes | 10:34.58 |
Robin_Watts | kens: Just got an offer email about http://www.ebuyer.com/262562-samsung-syncmaster-e2220n-tft-lcd-21-5-vga-monitor-ls22clysb-en | 11:10.47 |
kens | That is cheap Robin_Watts | 11:17.04 |
Robin_Watts | Yeah, I'm almost tempted to get rid of the old CRT I have as a spare. | 11:17.34 |
| VGA only is the only potential hitch. | 11:17.42 |
kens | :-) | 11:17.43 |
| My current display card only has VGA as the 2nd output | 11:17.58 |
Robin_Watts | And I'm using VGA for both my 1920x1200 ones :) | 11:18.16 |
kens | Its also slightly lower res than my primary display which might be a niggle | 11:18.18 |
Robin_Watts | kens: really? | 11:18.28 |
| Your existing one is 1920x1200 in a 24" ? | 11:18.49 |
kens | Its htat resolution yes | 11:18.59 |
| DOn't know about the screen size | 11:19.08 |
| I see at the top of the page the 'Galaxy Note to be aimed at female market'. So presumably they'll colour it pink.... | 11:20.02 |
Robin_Watts | kens: I'm sure your card will cope fine with it being a different ret. | 11:21.10 |
| res. | 11:21.13 |
kens | I'm sure it will, the CRT is at 1024x768 right now ;-) | 11:21.31 |
Robin_Watts | The only thing that might be a problem is if you plan to use windows that span between the 2 monitors. | 11:21.31 |
| and I *never* do that. | 11:21.44 |
kens | Well I wasn't planning to, I didn't like the look of it | 11:21.55 |
| Off to fetch lunch | 11:24.17 |
Robin_Watts | All the rest of these plank vs pamcmyk4 problems seem to be on page 10+ of pxl docs :( | 11:46.58 |
| OK. I've got one that goes wrong in page 2 of a document. | 11:51.08 |
| Anyone have any hints on cunning down pxl files? | 11:51.20 |
| pcl commands generally begin with 0x1b. Any such simple pointers for pxl? | 11:51.39 |
kens | Robin_Watts : pxldis and pxlasm | 11:59.55 |
| disassemble it, remove the btis you don't wnat and reassemble it | 12:00.08 |
Robin_Watts | Sounds great. | 12:00.16 |
kens | In toolbin maybe ? | 12:00.28 |
Robin_Watts | tools, it seems, thanks. | 12:01.02 |
kens | IN "/ghostpdl/tools" | 12:01.04 |
Robin_Watts | python ? | 12:01.05 |
kens | yes | 12:01.10 |
Robin_Watts | cool. | 12:01.13 |
kens | I run them on Linux, couldn't get them to work on Windows | 12:01.28 |
Robin_Watts | That's a pain. | 12:07.09 |
kens | WHich ? | 12:07.18 |
Robin_Watts | Not being able to run 'em under windows. | 12:07.28 |
kens | Well, you might be able to, I couldn't. | 12:07.39 |
Robin_Watts | I can run them, but round tripping a file gives something that fails. | 12:07.58 |
| I bet it's linefeed related. | 12:08.10 |
kens | Oh, thats bad | 12:08.13 |
Robin_Watts | So, edit on one computer, run script on another, test on 3rd :) | 12:14.44 |
kens | boggles | 12:14.53 |
Robin_Watts | Another dumb question, to which I'm sure I should know the answer... | 12:55.53 |
| do we have a device in ghostscript that just traces out all the operations done to it? | 12:56.10 |
kens | I think so | 12:56.23 |
| But I don't recall for certain | 12:56.37 |
| gdevtrac.c ? | 12:57.03 |
Robin_Watts | looks good, thanks. | 13:07.10 |
| but doesn't seem to actually build anything in. odd. Lunchtime. | 13:15.11 |
| gawd. It's going wrong in a rendering thing called from a text call :( | 14:12.20 |
kens | what sort of area ? | 14:12.52 |
Robin_Watts | It presumably calls code to render a glyph, and that's the call that's going wrong. | 14:13.54 |
| the trace stuff doesn't get that far down. | 14:14.05 |
kens | So that's through to FAPI (probably) | 14:14.07 |
| Either chrisl or I can maybe help | 14:14.15 |
Robin_Watts | So would that be in the 'build_char' proc? | 14:14.20 |
kens | Depends on the font type | 14:14.30 |
| font->FontType will tell you that | 14:14.41 |
Robin_Watts | pl_tt_build_char | 14:14.50 |
kens | Oh, this is PCL ? oops | 14:14.58 |
| So its a TrueType font | 14:15.04 |
| Therefore no BuildChar | 14:15.10 |
| as such | 14:15.18 |
henrys | kens:oh is there a portability problem with the xl assembler and dissassembler? | 14:15.24 |
Robin_Watts | So... what am I stepping through? :) | 14:15.26 |
kens | Given this is PCL that means its still using our TT interpreter not FT | 14:15.41 |
Robin_Watts | henrys: Yes. It uses stdin/stdout, which introduces CRLF problems on 'doze. | 14:15.55 |
kens | henrys, for me the Python doesn't run on Windows, Robin gets it to run bu the resutl is broken | 14:15.57 |
| Robin_Watts : where are you in the code ? | 14:16.14 |
| in pl_tt_build_char ? | 14:16.22 |
Robin_Watts | yes. | 14:16.27 |
kens | BuildChar has a particular meaning in PostScript, I presume hte term was reused in PCL | 14:16.41 |
| TT fonts don't have a BuildChar even in PostScript. | 14:16.58 |
| Robin_Watts : so whereabouts is the problem ? | 14:17.43 |
Robin_Watts | OK. So that draws it to a bitmap. | 14:17.54 |
| and the bitmap then gets displayed as an image. | 14:18.01 |
henrys | buildchar is where we do setcachedevice and render the character - have to do that ;-) | 14:18.02 |
kens | Yes, and should put it in the cache | 14:18.05 |
Robin_Watts | It's presumably that display call that's going wrong. | 14:18.15 |
kens | Hmm, it ought to be cached first, unless its really big | 14:18.30 |
Robin_Watts | Maybe. I'm ignoring the caching stuff. | 14:18.43 |
kens | Well it normally gets drawn from the cache, so it is a different path if that's not what is heppning | 14:19.06 |
| What actually is hte problem ? (silly question time) | 14:19.32 |
Robin_Watts | ROPs are going wrong. | 14:20.08 |
| (I assume it's ROPs) | 14:20.15 |
kens | Could be, you can ROP a bitmap from the cache certainly. | 14:20.33 |
Robin_Watts | pamcmyk4 is giving me a dfferent resylt to plank. | 14:20.34 |
kens | Ah.... | 14:20.41 |
Robin_Watts | I'm sure the problem isn't in the text code. | 14:20.49 |
kens | OK, I doubt I can offer any insight then. | 14:21.04 |
Robin_Watts | It's going to be in the planar code, but I need to catch it being called... | 14:21.04 |
henrys | there is a define in build char CACHE_TRUETYPE_CHARS, it might be useful to undefine that and see if the problem works - that would isolate it to the cache. | 14:21.23 |
kens | The text code will pull the bitmap from the cache and render it | 14:21.25 |
| Assuming its cached. | 14:21.32 |
| I don't recall exactly where that happens I'm afraid | 14:21.46 |
| Especially not with ROPs involved. | 14:22.13 |
henrys | gx_image_cached_char | 14:22.30 |
kens | Thanks henrys | 14:23.04 |
Robin_Watts | The image stuff is being called from image_bitmap_char | 14:23.26 |
kens | That sounds correct. | 14:23.36 |
Robin_Watts | so I'm guessing no cache involved. | 14:23.37 |
kens | Its always a possibility. | 14:23.51 |
| And might explain your difference if the other device is using the cache | 14:24.03 |
henrys | of so it is a bitmap char or you are doing artificail emboldenting. | 14:24.05 |
kens | coffees | 14:24.17 |
Robin_Watts | I'm doing artificial emboldening. | 14:24.50 |
| (I was interested to see the 'smear' comment in there. I wrote code for my previous job to do 'glows' and used the term 'smear' to refer to the boldening operation I did on that) | 14:25.41 |
| OK, so this is using the 'slow case, masked, not-orthogonal' path through the code... | 14:26.53 |
| and we end up in gx_dc_pattern_fill_rectangle | 14:28.54 |
kens | That seems odd, is this a pattern colour space ? | 14:29.48 |
Robin_Watts | pattern color space, with rop, with a clip, with non orthogonal axes, plus anything else you can think of to make it tricky. | 14:31.14 |
kens | Oh god. | 14:31.31 |
Robin_Watts | Yes. | 14:31.40 |
kens | Well if its a pattern space then I guess that makes sense to be there | 14:31.42 |
Robin_Watts | But without clist :) | 14:31.51 |
| So it could be worse. | 14:31.57 |
kens | Small mercies | 14:32.02 |
henrys | I guess the issue with pxldis.py is the binary "open" | 14:35.12 |
| the assembler must use binary. | 14:36.04 |
| no big deal I can check it later - but I'd like it to be portable. | 14:36.59 |
kens | It would be nice not to have to boot Linux | 14:37.21 |
Robin_Watts | dis is fine. | 14:37.40 |
| it's the reassembly that's fine. | 14:37.49 |
| I can dis on windows, as long as I asm on linux. | 14:38.02 |
| It is, I believe, that asm is outputting to stdout. | 14:38.18 |
henrys | okay I'll boot windows and have a look - after the coffee has brewed ;-) | 14:40.33 |
Robin_Watts | henrys: It's not urgent. | 14:40.52 |
| I am happy to use an ubuntu window. | 14:41.01 |
| http://code.activestate.com/recipes/65443-sending-binary-data-to-stdout-under-windows/ | 14:41.36 |
| Actually, that makes it look dead easy to fix. | 14:41.48 |
| I'll give that a whirl in a mo. | 14:41.54 |
henrys | oh thanks Robin_Watts | 14:43.45 |
Robin_Watts | Yeah, that fixes it. | 14:44.23 |
| Mr Google to the rescue again. | 14:44.42 |
| Morning mvrhel2 | 14:45.33 |
mvrhel2 | good morning | 14:45.46 |
Robin_Watts | mvrhel2, henrys: In this particular plank vs pamcmy4 problem, the pattern tile seems to only have data in the first plane. | 14:46.53 |
| I wonder if that's because pxl needs to know about planar pattern tiles? | 14:47.18 |
| px_paint_pattern perhaps? | 14:55.31 |
henrys | mvrhel2? | 15:10.21 |
Robin_Watts | It's a class 4 image. | 15:10.46 |
henrys | sorry I was offline for a bit. px_paint_pattern should just paint an image, why would it need to know about tiles? | 15:13.24 |
Robin_Watts | AIUI, pxl can be fed a 'pattern' as a block of image data, right? | 15:14.00 |
| It then feeds that image data through the standard image kit, and and the end, we are left with a memory imprint of it. | 15:14.43 |
| We then turn that into a pattern tile, and put it in the cache. | 15:14.52 |
henrys | mvrhel2:there will be a trip to Company 'C' nov 2 ray is the obvious choice location wise but you have more of a history with them? Who do you think should go? | 15:14.53 |
Robin_Watts | The tile I'm getting produced only has information in the first plane. | 15:15.32 |
| Until recently, pattern tiles were always chunky, so there was only ever one plane. | 15:15.56 |
henrys | it is no different than a postscript or pdf pattern with an image inside. | 15:15.57 |
| is there something in the pattern template that is new? | 15:16.19 |
Robin_Watts | I'm not 100% sure - mvrhel2 changed this bit, I think. | 15:16.47 |
| but I don't think so. | 15:16.53 |
| The problem may be that we have an image clipper and/or a raster op here. | 15:17.11 |
henrys | thought emboldening was the current problem | 15:17.59 |
mvrhel2 | henrys: since ray_laptop is right there it would make sense for him to go. | 15:20.06 |
henrys | so the issue if any is in render_pattern, when we make the pattern. | 15:20.23 |
mvrhel2 | I am fine going to though if it would help | 15:21.00 |
| off to take kids to school... | 15:21.51 |
Robin_Watts | henrys: No. | 15:22.36 |
henrys | I think it is wasteful unless happen to want to go to Newport Beach, you can go surfing with the kids ;-) | 15:22.48 |
| well the pattern paint proc in xl just uses the usual image painting code. It must be something with the device in the graphics state the pattern accumulator or something otherwise all images would be broken. | 15:26.28 |
| guess I'm stating the obvious. | 15:26.51 |
| kens: are you and I going to do support while marcos is gone? Or do you want to give somebody else a turn? | 15:27.56 |
Robin_Watts | henrys: Yes, I can't see anything in the pxl specifics. | 15:28.53 |
henrys | ray_laptop you are being signed up for a meeting Nov 2 with Company 'C' will you be around? | 15:29.09 |
kens | henrys that's fine, no problems there | 15:29.19 |
ray_laptop | henrys: yes | 15:29.32 |
| Scott called me about it last week to find out where to fly into | 15:30.00 |
henrys | well we were discussing whether you or michael should go scott thought michael would be better because of his history with them, I don't think it makes much difference so if you don't feel strongly about it I'd rather you just go. | 15:31.31 |
ray_laptop | be nice to see if we can get them un-stuck, and find out what, if any testing they have done. The contact between mvrhel2 and MQ was a bit irritating that they just did some testing and never talked to us | 15:31.32 |
| henrys: it's only a 25 min drive for me, so even if mvrhel2 goes, I'd like to be there just in case we need local follow up. | 15:32.33 |
henrys | yes | 15:32.57 |
ray_laptop | but I don't think they had any specific technical issues on color, halftoning, etc. that would require mvrhel2's expertise | 15:33.38 |
henrys | true. | 15:34.24 |
ray_laptop | we probably managed to convince the submitter of bug 692587 that there isn't much point to having a contract -- hard to imaging getting more attention than he did. | 15:38.44 |
henrys | he did contact scott long before the bug discussion. | 15:39.50 |
kens | He still has another bug, and it is a real if obscure one | 15:39.55 |
| I'm ignoring it for now, especially as I have a customer bug to work on | 15:40.08 |
henrys | hard to make a deal there - small number of units and gs is not a significant chunk of the functionality. | 15:41.14 |
kens | Hmm adding PDF/A output is quite significant. | 15:42.23 |
| Its becoming more and more of a requierment | 15:42.34 |
ray_laptop | kens: I agree | 15:42.39 |
kens | Oddly I don't seem to be able to mark the bug reopened, so I've just made it assigned | 15:43.51 |
henrys | in his view it is small because he only is doing it for a small percentage of his customers at least that is what scott was told. | 15:44.15 |
kens | Yes, I'm sure that's what he says ;-) | 15:44.30 |
henrys | but by all means hold off on the bug customers come first. | 15:44.44 |
kens | The only reason I've looked at them is because we *do* haev a lot of paying customers doing PDF/A conversion so if there's a problem we need to fix it. | 15:45.18 |
| But in his case he's taking 2 PDF files, converting them both to PDF/A, merging them together somehow and then converting the newly merged file back to PDF/A which is kind of daft | 15:45.53 |
| I don't think we'll have commercial customers doing that, so I'm leaving it while I have this other bug to look at, at least | 15:46.24 |
ray_laptop | frankly, I'd like to see some of the 'enhanclements' for pdfwrite before obscure bugs like this from a "free" user -- like compressed objects and xref for PDF 1.5 and writing linearized PDF's directly | 15:51.09 |
kens | Hey, I just did text rendering modes ! :-) | 15:51.32 |
ray_laptop | a balance between development and bug fixing | 15:51.44 |
kens | I was kind of hoping to get back to the text extraction at some point | 15:52.09 |
ray_laptop | kens: yes, and now that the formats are correct, it is working. | 15:52.17 |
| thanks. | 15:52.29 |
kens | Silly mistake on my part ;-) | 15:52.41 |
| Thanks for commiting the fix | 15:52.49 |
ray_laptop | I guess I need to open a bug to request UTF-8 instead of 16-bit | 15:53.01 |
kens | I know I want to do it. | 15:53.10 |
| But if you want an enhancement bug then by all means go ahead | 15:53.24 |
| Or you could add it to the still open 'want a text extraction class' bug | 15:53.38 |
ray_laptop | I think if we had that, then we could finally retire ps2ascii.ps | 15:53.42 |
kens | Its next on my list to do. | 15:54.05 |
| I *thought* the Tr stuff wouldn't take long. 4 weeks later.... | 15:54.20 |
ray_laptop | I'm down to 1 customer bug (we'll see how the customer for 692158 takes my analysis and closing of that one -- keep you fingers crossed) | 15:59.17 |
| off to get coffee... | 16:01.41 |
mvrhel2 | looks like we are down to 21 customer bugs | 16:08.07 |
kens | Heading off, night all | 16:10.36 |
mvrhel2 | good night kens | 16:10.41 |
Robin_Watts | Found it, I think. Was a problem in tile_by_steps. My fault. | 16:52.25 |
| going to be hard to fix thoguh. | 17:05.03 |
| mvrhel2: You here? | 17:11.17 |
| Actually, I have to pop out to get Helen from the station. | 17:11.35 |
| If you get a spare mo, could you look at tile_by_steps for me please and sanity check me? | 17:11.58 |
| In 'tile_colored_fill' we check first to see if (source == NULL)... and if it is we do the top bit of the if. | 17:12.24 |
| Otherwise we do the else clause... where source is never referenced.... is that right? | 17:12.43 |
| I can see that it would make more sense if it was rop_source we were testing there... | 17:28.48 |
henrys | is the tree reasonable stable now wrt the last release? | 17:30.04 |
Robin_Watts | I know of no reason why it isn't. | 17:32.37 |
| mvrhel2: In fact, as far as I can tell in that code, rop_source is only ever set to be the same as source. | 17:33.04 |
| So why have a separate rop_source record ? | 17:33.12 |
| I have to go. bbs. | 17:33.22 |
ray_laptop | found some really ugly bugs related to starting the interpreter that depend on GC not happening and moving i_ctx_p !!! | 17:36.14 |
| in psi/idsp.c and psi/imain.c -- both the same problem. | 17:37.18 |
| and I will take this opportunity to mention (again) that I HATE some of the widespread macros that hide which pointer is used (osp and gs_imemory are the ones that confused someone here) | 17:38.52 |
| I'm going to search for similar usages to see if there are some other places. (e.g. gs_main_run_string called followed by something that uses i_ctx_p WITHOUT updating it from minst->i_ctx_p) | 17:41.24 |
| looks line gs_main_finit also does | 17:42.12 |
henrys | 100 year old runs the toronto marathon, I had a bunch of friends in that race who didn't want to get beaten by him ;-) I suggested they stay with him until the last 800 meters and then take him in final sprint. | 18:04.01 |
| there was an eighty year old that was about 1/2 hour faster than my best marathon time, wow! | 18:05.28 |
Robin_Watts | mvrhel2: Ping ? | 19:01.18 |
| Actually, henrys might be the person to ask... | 19:03.44 |
| henrys: You here? | 19:03.48 |
| Back in 1998 (so, clearly this this will be fresh in your mind) you introduced the tile filling stuff. | 19:04.19 |
| Oh, that was the original import :( | 19:04.47 |
| I can see no purpose in having separate rop_source and source entries in gxp1fill.c - anyone object to me removing rop_source please speak soon. | 19:06.11 |
| mvrhel2, henrys: I think I've hit a bit of a showstopper here, I'd like to discuss it with you... | 19:46.26 |
| For the purposes of allowing the fast threshold based halftoning stuff, we introduced a new device procedure 'copy_plane', that basically does the same as copy_color, but only operates on a single plane at a time. | 19:48.00 |
| This is fine, except for the 'rop source device'. The rop source device basically does whatever drawing operation is requested of it, but combining it with a preset source and lop. | 19:49.22 |
| As we've discussed before, we can't 'copy_plane' with a lop (because planes aren't necessarily independent). | 19:50.16 |
| (we'd be OK with r,g,b, but not with c,m,y,k). | 19:50.34 |
| Suppose we have a contone pcl image... we halftone the c/y/k planes with michaels new code, and call copy_plane to put that on the page - but if the pcl has a rop set, we'll be in trouble, right? | 20:01.31 |
henrys | Hmm we agreed we were going to to rop after halftoning which we know is approximate. And I recally we agreed on some policy for the general issue of cmyk rop of 000 vs. 0001 but I'll be damned if I remember what we agreed to, but I think both flavors of black should be converted to 000. What did I forget? | 20:21.09 |
| sorry my numbers are rgb change 0 to 1 above | 20:50.19 |
Robin_Watts | I think regardless of what representation we choose we have big problems. | 20:56.11 |
| copy_plane itself is fine. It's the use of copy_plane with the rop source device that gives us problems. | 20:57.25 |
| because we can't work on just one plane in isolation. | 20:57.47 |
| I'm tempted to think about removing copy_plane entirely, but that impacts on mvrhel2s thresholding code. | 20:59.42 |
henrys | why can't the rop source device get all the planes? | 21:00.52 |
Robin_Watts | How hard would it be for him to buffer up n planes worth and then do 1 'copy_planes' instead of n 'copy_planes' ? | 21:01.01 |
| henrys: Because the copy_plane entry point only gets given a planes worth of data. | 21:01.18 |
henrys | well I think we do want to buffer the planes. | 21:01.55 |
Robin_Watts | Right. So copy_plane should become copy_planes. Need to check with mvrhel2 for how much work that is for him. | 21:02.35 |
henrys | the other solution would be to do the rop before halftoning right in the image code and not use the device at all. | 21:02.42 |
Robin_Watts | henrys: Yes. | 21:03.56 |
| However that's hard. | 21:04.14 |
henrys | yes copy_planes is the right thing to do. | 21:04.33 |
Robin_Watts | The image code might be (like in this case) producing rectangles that are diagonal. | 21:04.52 |
| I'm off to watch tv. will check back in later to see if mvrhel2 reappears. | 21:05.51 |
henrys | okay thanks for fixing the assembler, BTW! | 21:06.09 |
mvrhel2 | Robin_Watts: buffering the planes would not be an issue for me | 22:02.20 |
| sorry I was out a big part of today taking advantage of some good weather. my father turned 77 today so I took him up on a hike | 22:03.33 |
sebras | or8 chrisl Robin_Watts: slighty off topic, but anyway... SSD sector scattering is called wear levelling. more over the kind you really want is static wear levelling (since unchanged data has ~10year retention on NAND flash). I implemented this at my previous employer's. a b0rked SSD either has inconsistent internal state (likely), i.e. sector data is not mapped where the SSD thinks it should be, or it ran out of spare space when writing new dat | 22:37.49 |
| how on earth did I omit the initial t of tor8!? I pasted that! darn screen... | 22:38.38 |
| a (unlikely), i.e. the accumulated number of NAND erase blocks have risen over the number spare blocks. either way I'd suggest _not_ power-cycling the system and trying to get the data off the SSD as fast as possible. (hm.. irssi didn't mention that the last part got truncated as well...) | 22:42.50 |
Robin_Watts | sebras: I plan to install the core system on the SSD (OS, tools etc), and keep all my 'volatile' data on a harddrive. | 23:08.25 |
| That way if the SSD dies, I should be able to get away with a reinstall (onto a new SSD) | 23:08.57 |
| In any case, the important stuff I do tends to be backed up in git repos etc :) | 23:09.23 |
| mvrhel2: About the buffering - that's fab. | 23:09.54 |
| About your dad - congrats. | 23:10.14 |
| henrys: I don't suppose there is a pcl equivalent for pxlasm/dis ? | 23:18.36 |
mvrhel2 | so Robin_Watts: so we have a bit of a chicken and egg thing going on then. I suppose you could perhaps add in the command to do all the planes at once, then I could fix my end? | 23:32.18 |
Robin_Watts | mvrhel2: Yeah... | 23:32.43 |
| Let me try and find a fix for the current issue that's availing me. | 23:32.58 |
| (which basically will be avoiding calling copy_plane somewhere where I'm calling it now) | 23:33.19 |
| Then that *should* leave the only caller as being you. | 23:33.39 |
| I'll recast the code and get you a patch to look at. | 23:33.51 |
mvrhel2 | sounds good. going to beat on my halftone code to output data so use ray_laptops linearize_threshold operation | 23:51.41 |
Robin_Watts | Just tried my intended fix, and it's produced lots of differences on peeves :( | 23:52.15 |
| Will look into it tomorrow. | 23:52.22 |
| Bedtime. Night all. | 23:52.27 |
LaoLang_cool | My up time is your bed time?! | 23:57.55 |
Robin_Watts | LaoLang_cool: I'm in the UK. | 23:59.49 |
| Forward 1 day (to 2011/10/18)>>> | |