IRC Logs

Log of #ghostscript at

 <<<Back 1 day (to 2013/07/09)2013/07/10 
henrys marcosw:I gave back one of guillaume's bugs but if you want me to talk to him let me know.03:40.00 
kens chrisl ping07:37.56 
chrisl kens: pong07:39.12 
kens Does the pngalpha device do any kinf od anti-aliasing ?07:39.27 
chrisl I don't really know - the naming "convention" suggests so (like x11alpha), but I never really looked. I always use the command line opts07:40.28 
kens It kind of looks like it does07:40.35 
  Rather to my surprise07:40.42 
chrisl kens: if the file is made of lots of clipped shapes, that could explain the "missing" aa07:42.20 
kens Its several shading dictionaries07:42.32 
  THey are, of course, clipped to the filled path07:42.42 
  and yes, clips are never anti-aliased07:42.53 
chrisl So, a *scaled device might be the answer07:43.12 
kens Yes, and I sadi that on the stack overflow question he raised before seeing the bug report.....07:43.29 
chrisl Oh, a prat :-(07:43.41 
kens The odd thing is I cannot reproduce his output07:43.42 
  For me the output is fully anti-aliased, and at *much* lower resolution07:44.32 
chrisl Possibly a bug that's been fixed?07:45.32 
kens Its possible, I guess I shoudl try 9.0707:45.49 
  just going to take a quick look at the content of the PDF file07:46.00 
  THere's definitely a complex clip in there07:46.37 
  3 of tehm07:46.47 
chrisl Is this a PDF?07:47.02 
kens yes07:47.06 
chrisl Any transparency?07:47.14 
kens I don't think so no07:47.20 
  But the clip will prevent anti-aliasing of the shfill07:47.32 
chrisl I wondered what would happen if we rendered the shfill into a group buffer, then rendered the group buffer when anti-aliasing was happening07:49.00 
kens boggles07:49.22 
  It would probably work, I guess, but I think using a scaed driver would be easier :-)07:49.40 
  OK 9.07 does behave as described, the curretn development version doesn't07:52.07 
chrisl So whatever it is has been fixed - sounds good :-)07:52.30 
kens I'm not sure I would say it is 'fixed'07:52.41 
  The resolution is *way* too low07:52.47 
  Now the current code behaves the same...07:57.11 
  OK I'm not going to worry about that, the current code now behaves as I woudl expect...07:57.19 
  Oh I must be tored, I've been using different filenames....07:57.27 
Robin_Watts the tor twins are here!09:49.08 
Robin_Watts shoots off for a bit.09:49.45 
sebras Robin_Watts: you would guess that tor0, tor1, tor1, tor2, tor3 would also show up some time...11:01.06 
Robin_Watts kens, chrisl: Did you book flights yet ?12:21.38 
kens Robin_Watts : not yet, but I saw the email12:22.39 
chrisl Not yet, will look shortly12:23.41 
  Are you guys return on the Sunday?12:25.30 
Robin_Watts chrisl: Yes, I think that's the plan.12:25.43 
kens Well I wasn't planning on staying longer12:25.43 
chrisl I couldn't remember when the show started12:26.13 
Robin_Watts show starts sunday, according to Miles' mail.12:26.33 
  Looks like 593.4512:26.47 
chrisl Yep, I should go find my credit card......12:27.34 
kens Seems to be what mine comes up as12:28.01 
  Hmm, when I move to the next step, the far goes up by £1....12:29.48 
Robin_Watts Interesting. Spending miles (not Miles + Money), I can get economy out, PE back for 47500 miles and 286quid of taxes.12:33.24 
kens Oh great, this time verified by visa doesn't even bother to show me a dialog12:35.19 
  I guess its broken12:35.25 
Robin_Watts so, out on Thursday 5th, back on sunday 8th.12:41.14 
kens yes 11 am out 18:00 back12:41.25 
Robin_Watts 53D out12:42.57 
kens I'm still fighting wiht it12:43.33 
  Using chrome it verified my details without asking for my password.....12:44.03 
  Verified by Visa this is12:44.13 
Robin_Watts 21D on the way back.12:44.43 
kens Show off :-)12:44.52 
Robin_Watts Now I'm going to ring them to see how much it'd cost to upgrade the outbound leg to PE too :)12:45.15 
kens Robin_Watts : why did you choose such a far back row ?12:47.36 
  I can book seats as far forward as 4212:47.57 
Robin_Watts ok, 42D12:48.45 
kens Well I shall have 43D then12:49.16 
  and 42D coming home12:50.04 
  Robin_Watts : the Virgin site is offering me PE both ways for an extra £468, no miles.12:53.08 
Robin_Watts kens: yeah. It would cost me that much to get PE both ways, so I'll live with economy both ways.12:55.09 
  Apparently as well as extra legroom seats they now do 'seat plus' which is where you can pay to have the seat next to you empty.12:55.39 
  they said to ring again a few days before the flight as they might be selling things off then.12:56.00 
kens seat plus sounds interesting, but what if the plane is full ?12:59.47 
kens will probably upgrade to extra legroom at the airport again13:00.13 
Robin_Watts kens: presumably they only offer seat plus if they can predict availability13:36.15 
kens I guess, it would be annoying to find they'd sold the seat next to you anyway though13:36.38 
kens thinks; I prefer extra legroom to width on the whole13:39.48 
Robin_Watts kens: I think I'd prefer 2 seats.13:42.28 
kens :-)13:42.29 
Robin_Watts but then we have different builds :)13:42.29 
  hurrah. You know it's a good sign when you write a huge block of new code, and on the first run it dies on the very first line...13:44.23 
kens just got hundreds of differences I didn't expect on my cluster run :-(13:48.22 
Robin_Watts kens: When you were doing linearisation in gs, did you have any code to dump hint tables?14:52.55 
  ha. scared him away.14:53.15 
kens The mer mention of linearisation crashes Miranda....14:53.50 
Robin_Watts yeah.14:54.00 
kens I think I did, yes, but I doubt I still have it14:54.03 
  I'll quickly rummage around14:54.22 
Robin_Watts Thanks.14:54.26 
  Table F.3: Item 2 is supposed to be the location of the first pages page object.14:54.51 
  And in pdf_reference17.pdf, as far as I can tell, it's not.14:55.03 
kens I found something called PDFHintRead, but I don't remember it14:55.07 
  It looks too primitive to be useful14:55.34 
  It will dump that, let me rebuild it14:55.59 
  my code says the location of the first page's page object is 75819781214:57.51 
  But all the numbers its dumping look wrong14:58.14 
  Ah, look slike it wants teh hint stream extracted14:58.39 
Robin_Watts kens: Also, are we using the same pdf_reference17.pdf? :)14:59.09 
kens That's a good question...14:59.19 
Robin_Watts Mine is 32472771 bytes14:59.42 
kens Hmm I'd have to decompress the hint stream before I could read it14:59.54 
  Mine says 32,431,66215:00.24 
Robin_Watts kens: Don't let me waste your time on this. feel free to mail me the code and be done with it :)15:00.29 
kens The code is really dumb, I'm sure you could write your own faster15:00.50 
Robin_Watts kens: I have written my own, that's the problem :)(15:01.02 
kens Oh, OK then I'll mail you this, but its really stupid code, I'm not sure what I was suing it for15:01.20 
Robin_Watts I suspect the file is wrong.15:01.46 
kens Yes, because nobody uses hint streams :-)15:02.01 
  OK its on its way to you, 225 k rar, hope that format is OK15:05.08 
Robin_Watts kens: I'll cope. Ta15:05.15 
  Gah. Linearised files are so broken.16:02.15 
  Page 2 of the file is at object 1, as you'd expect.16:02.32 
  It makes reference to a shared object down at 3904.16:02.34 
  3904 being a font. Where is it's font descriptor? Presumably in the shared objects too, right? No. Its font descriptor is 9, its tounicode is 10, and its encoding is 11.16:03.44 
  So quite how that's supposed to be a 'shared' object I don't know.16:04.03 
  In fact... it's not shared. That font is only used on one page.16:05.27 
kens So it shoudl be unique16:05.46 
Robin_Watts Right, but it's in the shared object section of the file.16:06.03 
kens But to be honest, I really don't think anyone produces well linearised files, because hte spec is so naff16:06.07 
chrisl I've never found a consumer that actually uses all the of the linearisation stuff either......16:07.18 
kens Again, because nobody produces decently linearised files....16:07.40 
Robin_Watts chrisl: I've never found one that uses ANY of it.16:07.44 
  but sebras informs me that poppler does.16:07.51 
kens Poppler reads it, I'm not cerrtain it *uses* it16:08.03 
chrisl As far as I could find, poppler parses the tables, but doesn't use much of it16:08.20 
Robin_Watts chrisl: but it uses some of it?16:08.39 
chrisl Robin_Watts: I only looked at the hints table (is that what it's called?) in any detail - is parses that, didn't use it16:09.23 
  Mind, that was two or three releases ago, so.......16:11.26 
sebras chrisl: hm... as far as I could see they are using the hints to locate the page object..?16:43.08 
  I might have misread the source of course, but I still believe it to be true.16:43.31 
  I'll have a closer look later today.16:43.38 
chrisl sebras: you may well be right, but I could find where the table was parsed, but I never found it being used - it ignored the result, as far as I could tell.16:45.13 
  sebras: the other point being that the offsets it was reading were complete nonsense for all the files (from various sources) that I tried16:48.43 
Robin_Watts well, popplers source has pointed me to the thing I needed.16:59.35 
  turns out they byte align more often than the spec suggests.16:59.47 
  and I am now getting plausible entries out.16:59.55 
  BUT I think popplers code is broken.17:00.08 
  in that it forces a value to 0 that it then relies on being non-zero. Joy.17:00.25 
chrisl Robin_Watts: I rather assumed that poppler would start using the hint table (and fixing problems with their code, if required) when it looked like it might be useful a reasonable amount of the time.17:03.11 
Robin_Watts chrisl: chicken and egg, innit.17:05.43 
chrisl Robin_Watts: Exactly, and Adobe don't even bother to read it - and we think write nonsense anyway17:06.53 
Robin_Watts chrisl: It's not entirely nonsense...17:07.15 
  Their files are far from optimal, but I think there is enough there to help us with progressive loading.17:07.35 
chrisl TBH, if they'd wanted that kind of functionality, the format should have been streamable17:09.38 
Robin_Watts chrisl: Yes, and in an ideal world the more food I eat the more handsome I'd become, but alas we have to work with the situation we have :)17:15.22 
ray_laptop Robin_Watts: I saw in the logs that you were looking for something to dump the hint tables. We removed it (chrisl did last August), but there was gs/lib/dumphint.ps17:18.59 
  but the comment says that it was bit-rotted. 17:19.27 
sebras Robin_Watts: it was the byte alignment I was referring to the other day.17:23.30 
  Robin_Watts: there are comments in poppler about this not being in the spec.17:23.42 
  I assumed that this was the reason that you and kens got strange results.17:23.56 
ray_laptop Does anyone know where we keep the 32000 spec(s) ?17:25.13 
Robin_Watts ray_laptop: Physical copies?17:26.15 
ray_laptop Robin_Watts: no, PDF. Does someone have one up on casper17:26.45 
Robin_Watts I have pdf_reference17.pdf there, which is the latest non ISO version.17:27.05 
  I can copy PDF32000_2008.pdf there if you want.17:27.42 
ray_laptop Robin_Watts: I have the Adobe spec and the 32000 supplement from Adobe. I was looking for the actual 32000 spec17:27.43 
  Robin_Watts: is that the ISO version ? If so, please do17:29.04 
Robin_Watts copying now to my home dir.17:29.07 
ray_laptop Robin_Watts: OK. Thanks17:30.07 
Robin_Watts Or, from the horses mouth: 
ray_laptop I thought ISO doesn't allow redistribution. Did they make an exception for Adobe ? (obviously passing it around inside the company that bought it is OK)17:31.45 
Robin_Watts ray_laptop: I know not.17:33.23 
henrys so simon is now nda'd … he never comes around to IRC?17:36.04 
Robin_Watts This document you are now reading is a copy of the ISO 32000-1 standard. By agreement with ISO, Adobe Systems is allowed to offer this version of the ISO standard as a free PDF file on it’s web site. It is not an official ISO document but the technical content is identical including the section numbering and page numbering.17:36.44 
  henrys: I've never seen him here.17:36.53 
henrys so I'll set up his account and one of the mupdf'er will give him instructions on running regressions and such?17:39.22 
Robin_Watts henrys: Sure. In the email where you tell him the account details, invite him to irc, or say he can email me for help.17:39.58 
henrys 10-417:40.18 
ray_laptop Robin_Watts: the "Copyright Notice" on page 2 explains it: This document has been derived directly from the copyright ISO 32000-1 .... It is being made available from ... Adobe ... under agreement with ISO for those that do not need the official version containing the ISO logo and copyright notices.17:40.20 
  Robin_Watts: it also adds "The technical material is identical between this version and the ISO Standard"17:40.58 
  Robin_Watts: so, thanks for the link and uploading it.17:41.21 
Robin_Watts gets dejavu :)17:41.47 
ray_laptop oops. I was reading the spec, not the logs :-/17:43.19 
Robin_Watts Right. I've now got code that successfully extracts plausible looking data from the hints stream.18:14.27 
  Next step is to use that data...18:14.34 
sebras Robin_Watts: chrisl: did you already look are the sources or is there still a point in me doing so?21:03.10 
mvrhel_laptop bbiaw22:48.15 
Robin_Watts sebras: I'm sorted, I think, thanks.22:55.15 
 Forward 1 day (to 2013/07/11)>>>